Xaira Therapeutics
AI 原生药物发现引擎——投入要素异常强、科学进展真实,但定价纪律仍未解决
继续研究:Xaira 的科学基础、团队和资本足以支撑继续尽调,但公开证据不足、定价未知,还撑不起按高溢价私人估值承销。
封面要素
公司概况
Xaira 于 2024 年 4 月带着超过 $1 billion 的承诺资本启动,使命是把先进 AI 研究、大规模生物数据生成和疗法产品开发接到同一套操作系统里。公开产物包括 Orion 这一大型 perturb-seq 数据集、Pisces,以及首个虚拟细胞版本 X-Cell。公司更像平台 + 管线一体化的生物科技公司,而不是纯软件供应商;最可见的外部牵引力也在开放科学,而不是已披露的商业采用。
- 成立时间
- 2024-04-23
- 创始人
- David Baker, Hetu Kamisetty
- 创立地点
- San Francisco Bay Area
- 总部
- South San Francisco, CA
- 产品
- Xaira 不是在做一个已经完成的企业软件 SKU,而是在搭一套内部使用、也面向合作伙伴的发现引擎。公开可见模块包括 Orion perturb-seq 图谱、Pisces 数据和 X-Cell 虚拟细胞模型产物;私下里,公司把自己定位成一套应能把模型与数据转化为差异化疗法项目的系统。
- 客户
- 当前公开证据集中在内部平台用户、开放科学研究者和潜在生物科技 / 制药合作方,而不是已披露的付费客户。
- 商业模式
- 可能的货币化路径是内部管线价值创造,加上少数高价值合作或与平台绑定的战略伙伴关系。公开定价和合同结构未披露。
- 阶段
- Private, post-launch
- 融资情况
- 2024 年 4 月启动时拥有超过 $1 billion 的承诺资本。公开投后估值和融资条款仍未披露。
执行摘要
主要优势
- 超过 $1B 的承诺资本,让 Xaira 这家新成立的 AI 生物技术公司具备少见的融资韧性。
- 管理层和董事会密度在行业里少见,拼合了前沿 AI、蛋白设计、转化医学和大公司运营经验。
- Orion 和 X-Cell 给了 Xaira 比典型隐身 techbio 更多的公开科学证据,也显示研究社区已有真实外部兴趣。
- AI、数据和疗法一体化架构一旦跑出可重复的合作或资产证据,就可能支撑溢价。
主要风险
- 估值条款、每股价格和优先权结构都未公开,无法直接用公开证据判断入场纪律。
- 开放科学牵引力还没有转化为具名付费客户、披露定价或合作经济性。
- 公开证据仍未显示,虚拟细胞和扰动数据资产已经清晰转化为患者结局或资产级证据。
- 安全、合规和监管准备度材料公开得太少,撑不起成熟企业或受监管用途的承销判断。
- AI 赋能药物发现的公开可比公司仍在低个位数十亿美元区间;没有私下证据时,可证明的溢价空间受限。
未决问题
- 价格、投后估值和完整条款清单经济性
- 完全稀释股权结构表和清算优先权瀑布
- 烧钱分配和基于里程碑的现金跑道计划
- 具名合作管线、定价逻辑和合同结构
- 平台到资产的转化指标和合规尽调包
目录
01公司概况
1.1 身份定位、范围与商业模式
Xaira Therapeutics 是一家总部位于 South San Francisco 的 AI 原生生物科技公司,定位是端到端药物发现与开发平台,而不是单点软件供应商。官方材料反复强调三根整合支柱:先进机器学习研究、大规模数据生成、疗法产品开发。公司目标是让生物学“更可计算”,用前沿模型找到正确生物学与分子,最终缩短从实验室洞见到药物的路径,尤其针对过去被认为难以或无法成药的靶点。落到运营上,Xaira 想同时做模型构建者、湿实验数据生产者和管线公司。Goldman Sachs 主办的发言和启动报道也强化了这一点:公司试图把靶点识别、分子设计和临床开发接进同一技术栈。这个定位重要,因为它意味着资本强度和执行范围都高于典型平台型生物科技公司;但如果模型开始产出差异化资产,Xaira 也能更控制价值沉淀的位置。公开来源支撑其身份和野心,但收入、估值和客户指标仍未披露。[CO001, CO003, CO004, CO005, CO006, CO018]
| 指标 | 数值 / 状态 | 日期 | 置信度 | 缺口或注意事项 |
|---|---|---|---|---|
| 启动资本 | >$1B 承诺资本 | 2024-04 | 高 | 已披露承诺资本;提款安排未披露 |
| 总部 | 700 Gateway Blvd, 4th Floor, 南旧金山 | 2026-05 | 高 | 官方地址;运营足迹已延伸至总部之外 |
| 创新中心 | 西雅图和伦敦 | 2026-03 | 高 | 当前站点级人数未披露 |
| 启动时员工数 | ~50 | 2024-04 | 中 | Endpoints 的启动期数字,不是官方备案 |
| 后续披露员工数 | 总计约 80 人;西雅图约 15 人 | 2024-08 | 中 | 未找到当前 2026 年员工数更新 |
| 公开科学里程碑 | X-Atlas/Pisces 上的 X-Cell(4.9B 参数;25.6M 细胞) | 2026-03 | 高 | 模型 / 论文里程碑,不是临床证明 |
| 私募估值 | 2026-05 | 中 | 已审阅公开来源未披露同期估值 | |
| 收入 / 客户数 | 2026-05 | 中 | 未披露公开收入或客户指标 | |
| 临床阶段资产 | 未公开披露 | 2026-05 | 中 | 管线被描述为正在搭建;未找到具名 IND 阶段资产 |
快照表将已验证的公开事实与显式空值并列;截至 runDate,公开证据不足以支持的指标均标为空。
[CO003, CO018, CO019, CO020, CO021, CO038]1.2 创始人与科学起点
Xaira 的创立故事把顶级科学源头与重磅风投孵化拼在一起。官方履历列出 David Baker、Marc Tessier-Lavigne 和 Hetu Kamisetty 为联合创始人;启动报道也把 ARCH 的 Robert Nelsen 和 Foresite 的 Vik Bajaj 描述为组建公司的风投操盘人。科学内核显然来自 Baker 位于 University of Washington 的 Institute for Protein Design:RFdiffusion 和 RFantibody 都出自那里,Xaira 多名早期研究者也曾在此训练。启动材料还称,Xaira 纳入了从 Illumina 剥离的功能基因组学能力,以及来自 Interline Therapeutics 的蛋白质组学团队,因此生物数据生成能力比纯蛋白质设计初创更宽。这解释了为什么管理层反复说,Xaira 要横跨生物学发现、类药物物质设计和临床开发,而不是只盯抗体设计。结果是一家公司从第一天起就具备真正的跨学科宽度,但创立叙事也横跨多个利益群体;公司扩张时,治理和运营对齐必须更谨慎。[CO002, CO007, CO008, CO009, CO010, CO011]
| 人员 | 角色 | 公开披露背景 | 贡献 | 依赖 / 尽调角度 |
|---|---|---|---|---|
| Marc Tessier-Lavigne | 联合创始人、董事长兼 CEO | 前 Genentech CSO;前 Stanford 和 Rockefeller 校长 | 科学领导力、公司建设、外部可信度 | 关键人物和声誉风险仍然重大 |
| Hetu Kamisetty | 联合创始人兼 CTO | 前 Meta;前 Baker 实验室博士后;ML 博士 | AI 模型架构与平台搭建 | 需要证明研究人才之外的规模化产品化能力 |
| David Baker | 联合创始人兼科学顾问 | UW Institute for Protein Design 主任;2024 年 Nobel 获得者 | 蛋白 / 抗体设计可信度和招聘磁石 | 顾问而非运营角色,可能限制日常杠杆 |
| Robert Nelsen | 联合创始人 / 董事 | ARCH Venture Partners 创始人 | 资本形成和战略网络 | 经济权利 / 控制权未公开披露 |
| Vik Bajaj | 联合创始人 / 董事 | Foresite Labs CEO;Foresite Capital 董事总经理 | 孵化、融资、转化策略 | 需要了解治理权和未来融资影响力 |
| Debbie Law | 首席科学官 | 前 BMS SVP;前 Merck/Jounce/Ablynx 高管 | 生物制剂发现和转化专长 | 近期加入,平台输出仍处早期证明阶段 |
| Paulo Fontoura | 首席医学官 | 前 Roche SVP / 多个治疗领域全球负责人 | 临床开发和以患者为中心的开发设计 | 尚无公开临床项目可衡量其影响 |
| Bo Wang | SVP 兼 Biomedical AI 负责人 | U of Toronto/UHN/Vector Institute;scGPT 先驱 | 虚拟细胞和多模态生物学模型领导力 | 执行取决于专有数据规模和湿实验闭环质量 |
| Jeff Jonker | 总裁兼 COO | 前 Belharra、Ambys、NGM、Genentech、Wilson Sonsini | 运营扩张和合作经验 | 需要评估运营节奏能否匹配资本强度 |
| Rachel Lane | SVP,业务发展与运营 | 前 Belharra CBO;曾任 Versant 和 Inception 角色 | 交易达成和制药合作伙伴形成 | 业务发展计划仍早期,尚未由公开交易衡量 |
表格聚焦与尽调最相关的创始人和当前运营领导者;角色和背景来自官方简介和公司新闻稿。
[CO007, CO008, CO009, CO010, CO013, CO014]1.3 领导班子、董事会与治理覆盖
对一家尚无已披露临床阶段资产的公司来说,Xaira 当前领导层资历异常深。Marc Tessier-Lavigne 有 Genentech 和学术机构领导经验。Hetu Kamisetty 坐镇 AI/ML 技术栈;Debbie Law、Paulo Fontoura、Bo Wang、Jeff Jonker 和 Rachel Lane 在 2024–2026 年陆续加入,补强科学、医学、AI、运营和合作能力。董事会与科学顾问阵容纸面上同样强,Scott Gottlieb、Alex Gorsky、Carolyn Bertozzi、Richard Scheller、Robert Nelsen 等人覆盖监管、大药企、风投和诺奖级科学。这种宽度让 Xaira 在治理、招聘和未来合作上,比多数早期生物科技公司更有覆盖。但公司外部可信度也高度集中在少数明星名字上。Marc 和 David Baker 仍是两个最显眼的身份锚点;其中任何一人脱钩或信誉受损,对招聘、融资和外部信任的冲击都可能不成比例。高规格班底是真优势,但不能消除关键人物或监督风险。[CO008, CO009, CO013, CO014, CO015, CO016]
| 利益相关方 | 角色 | 重要性 | 影响证据 | 开放尽调问题 |
|---|---|---|---|---|
| ARCH Venture Partners 与 Robert Nelsen | 领投方和联合创始赞助方 | 资本锚点和战略赞助方 | 最大初始 ARCH 承诺;提及 >$200M ARCH 出资 | ARCH 持有哪些所有权和控制权? |
| Foresite Labs / Vik Bajaj | 领投方和联合创始赞助方 | 共同孵化方和转化策略伙伴 | 联合孵化和董事会席位 | 承诺资本中有多少可随时间调用? |
| Scott Gottlieb | 董事 | 监管和政策可信度 | 启动时进入董事会,并列于官方团队页 | 董事会在治理 / 合规上的监督有多活跃? |
| Alex Gorsky | 董事 | 大型制药和运营网络 | 启动时进入董事会,并列于官方团队页 | Xaira 是否会利用董事会关系做合作或招聘? |
| Carolyn Bertozzi | 董事 / 科学可信度 | Nobel 级科学和 Stanford 生态触达 | 列于董事会和团队页 | 顾问深度是否能转化为项目选择质量? |
| Richard Scheller | 董事 | Genentech / BridgeBio 疗法开发可信度 | 列于董事会和官方团队页 | 董事会如何评估平台到管线的转化? |
| Scientific Advisory Board | 非董事专家网络 | 延伸 AI、生物学和转化触达 | 官方团队页列出 Baker、Barzilay、Anandkumar、Weissman 等 | SAB 多频繁影响项目组合决策? |
| Parker Institute for Cancer Immunotherapy(机构) | 投资者 / 生态利益相关方 | 释放肿瘤学和转化网络访问信号 | 列为启动支持方 | 是否存在项目化合作,还是纯投资人支持? |
| 西雅图 IPD 生态 | 人才管线和技术依赖 | 把蛋白设计经验输入 Xaira | GeekWire 记录西雅图团队由 IPD 校友搭建 | Xaira 对 UW/IPD 招聘优势有多依赖? |
映射影响 Xaira 超出执行团队之外的资本、治理和生态利益相关方。
[CO010, CO021, CO022, CO023, CO026, CO027]1.4 资本基础、地域布局与已披露规模
Xaira 启动时拿到的是生物科技初创史上罕见的大额初始融资承诺:来自 ARCH、Foresite 和一批知名风投的超过 $1 billion 承诺资本。Bob Nelsen 另外告诉 Endpoints,仅 ARCH 就打算投入超过 $200 million,并把这笔钱称为“硬钱”,还暗示初始资本只是起点而非上限。这样的资产负债表支撑 Xaira 从一开始就建设宽口径内部能力,而不是狭窄地依赖合作。公开披露的规模数字仍然稀少,但启动报道称公司 2024 年约有 50 名员工;GeekWire 后来报道约 80 名员工,其中约 15 人在 Seattle,London 还有少数员工。官方页面和 2026 年公司材料现在称,Xaira 总部位于 South San Francisco,并在 Seattle 和 London 设有创新中心。公司还宣布在 South San Francisco 内部迁总部,另一篇报道指向 73,075 平方英尺的 San Francisco 空间扩张。合在一起,这些披露支持一个真实的多地点运营建设,即便当前精确员工数和现金消耗仍未公开。[CO018, CO019, CO020, CO021, CO022, CO023]
| 站点 / 规模信号 | 公开披露细节 | 来源日期 | 运营含义 |
|---|---|---|---|
| 南旧金山总部 | 700 Gateway Blvd, 4th Floor;BioMed Realty Gateway of Pacific III 园区 | 2024-12 to 2026-05 | 把 Xaira 锚定在湾区生物技术枢纽,靠近人才、投资者和合作伙伴 |
| 西雅图创新中心 | 2024 年约 15 人;分子设计和 AI 团队位于 Lake Union 附近 | 2024-08 | 让公司贴近 Institute for Protein Design 人才基础 |
| 伦敦创新中心 | work-with-us 页面和 2026 年公司材料正式披露 | 2026-03 to 2026-05 | 增加欧洲人才触达,但具体职能和规模仍未披露 |
| 启动时员工数 | 西雅图和加州合计约 50 名员工 | 2024-04 | 显示公司启动时有真实运营基础,不只是空壳 |
| 后续披露员工数 | 约 80 名员工,多数位于湾区 | 2024-08 | 显示 2025/2026 年领导层补强前,早期招聘已快速推进 |
| 旧金山足迹扩张 | 计划 73,075 平方英尺,约 2025 年 7 月入驻 | 2024 年二级报道 | 暗示长期实体扩建,与大型数据和湿实验运营一致 |
办公和员工数披露来自官方页面、公司新闻稿和一份二级本地开发报道;当前 2026 年逐站点人员配置仍未披露。
[CO003, CO018, CO019, CO020, CO024, CO025]1.5 里程碑与当前运营姿态
公开里程碑显示,Xaira 正从隐身组建走向可见的平台证明点,而不是走进临床。Xaira 于 2023 年注册成立,2024 年 4 月公开启动,之后十八个月主要补齐高管班子和物理布局。科学侧的主要公开里程碑,是 2025 年 6 月发布 X-Atlas/Orion,当时被称为最大公开可用的全基因组 Perturb-seq 数据集;随后是 2026 年 3 月推出 X-Cell,一个用 25.6 million 个细胞的 X-Atlas/Pisces 数据集训练、拥有 4.9 billion 参数的虚拟细胞模型。这些发布重要,因为它们把 Xaira 的“AI 药物发现”叙事转化成至少部分公开技术产物和开放科学信号。与此同时,公司仍未公开披露具名临床候选物、IND 时间或首次人体试验日期。因此,里程碑轨迹支持这样一个判断:Xaira 是一家资金充足、正在从保密转向选择性科学披露的临床前平台公司,还不是一家已有临床验证的疗法公司。[CO001, CO015, CO016, CO017, CO024, CO031]
| 日期 | 事件 | 类型 | 状态 / 金额 | 参与方 | 重要性 |
|---|---|---|---|---|---|
| 2023-05 | 公司以 Orion Medicines 名义隐身注册 | 治理 | 隐身组建 | Xaira 创始团队 | 显示运营建设早在公开发布前就已开始 |
| 2024-04-23 | Xaira 公开发布 | 创立 | >$1B 承诺资本公告 | ARCH、Foresite、Marc Tessier-Lavigne、David Baker 等支持方 | 立即形成规模,并设定异常宏大的范围 |
| 2024-08-14 | 西雅图实验室侧写发布 | 规模 | 约 80 名员工;西雅图约 15 人 | GeekWire 和西雅图团队 | 发布后首次具体披露运营规模 |
| 2024-10-17 | Debbie Law 和 Julia Tran 获任命 | 治理 | 新增 CSO 和 CPO | Xaira 领导层 | 释放科学和人才运营建设信号 |
| 2024-12-11 | Paulo Fontoura 和 Hetu Kamisetty 宣布出任高管;总部搬迁披露 | 治理 | CMO 和 CTO 角色;总部搬迁 | Xaira 领导层 | 加深临床和技术班底,同时正式确立南旧金山基地 |
| 2025-04-03 | Bo Wang 加入并领导生物医学 AI | 治理 | SVP 招聘 | Xaira 及 U of Toronto/UHN/Vector 背景 | 补强虚拟细胞和多模态生物学建模领导力 |
| 2025-06-17 | X-Atlas/Orion 数据集公开发布 | 产品 | 8M-cell Perturb-seq 数据集 | Xaira 科学团队 | 公司首个重要开放科学资产 |
| 2025-07-09 | Jeff Jonker 加入,担任总裁兼 COO | 治理 | 运营和规模化招聘 | Xaira 领导层 | 补充资深运营者和潜在合作对手方接口 |
| 2026-03-17 | X-Cell 虚拟细胞模型发布 | 产品 | 基于 25.6M-cell 数据集的 4.9B-parameter 模型 | Xaira、Ci Chu 与 Bo Wang | 迄今最具体的 AI 平台论点证明点 |
| 2026-03-26 | Rachel Lane 加入,推动业务发展和运营 | 合作 | SVP 招聘 | Xaira 领导层 | 暗示准备把平台转化为外部交易,同时推进内部管线工作 |
时间线捕捉创立、规模、治理、产品和合作准备度上对尽调最重要的公开里程碑。
[CO001, CO016, CO017, CO020, CO024, CO041]时间线日期采用公司和独立网站报道的公开发布日期。
[CO001, CO016, CO017, CO021, CO041, CO042]1.6 反向信号与尽调缺口
反向材料中有两个问题最突出。第一,Xaira 的领导层故事绕不开 Marc Tessier-Lavigne 2023 年从 Stanford 辞职引发的争议。KQED 和 Retraction Watch 都记录称,独立审查没有认定 Tessier-Lavigne 本人存在欺诈,但发现重要缺陷、其实验室其他人的数据问题,以及未能果断纠正科学记录。这不推翻 Xaira 的技术投资逻辑,但会反复成为治理和声誉层面的讨论点。第二,尽管 Xaira 拥有巨额资金,公司在经济指标和开发时间表上仍明显不透明。本报告审阅的公开来源没有披露当前股权估值、收入、客户数、账面现金或首个临床候选物时间。启动和画像报道也保留了科学家与投资人的怀疑:de novo 抗体生成和以 AI 为先的生物科技执行仍是早期学科。因此,Xaira 带着异常强的资本和人才进入后续尽调,但治理韧性、经济披露,以及平台多快能产出人体测试资产,仍是未解问题。[CO031, CO033, CO034, CO035, CO036, CO038]
1.7 展示材料
02市场分析
2.1 Xaira 实际切入的市场
不能把 Xaira 当成一家出售单一现成 AI 软件产品的公司来分析。官方材料描述的是一个端到端系统,把 AI 研究、自有数据生成和疗法产品开发结合起来。2026 年独立报道进一步显示,公司正在主动建设炎症和免疫管线,最初以抗体疗法为中心,同时仍把 AI 平台视为核心引擎。这意味着 Xaira 至少同时触及三层经济池:AI 药物发现平台支出、抗体发现与生产基础设施,以及免疫学和炎症疾病的最终疗法收入池。区分这些层次很重要,因为每一层的买方、估值逻辑和采用节奏都不同。平台预算由 R&D 和 BD 团队控制,可以随着技术证据推进;疗法收入则取决于多年临床转化、报销和医生采用。因此,本章最稳妥的边界是多视角定义,而不是单一万能 TAM。[CM001, CM002, CM004, CM005, CM021, CM022]
| 细分 / 类别 | 纳入支出 | 排除支出 | 买方 / 付款方 | 与 Xaira 的相关性 |
|---|---|---|---|---|
| AI 赋能药物发现平台市场 | 与靶点识别、设计和预测相关的软件、模型访问、发现服务、工作流工具,以及部分平台合作支出 | 已获批药物销售、一般 CRO 收入、通用云支出,以及未作为 AI 平台一部分销售的湿实验服务 | 大型制药和生物技术 R&D / BD 预算;部分学术平台预算 | 如果 Xaira 销售或合作开放其 AI 与数据栈,这就是最相关的近期外部变现视角 |
| 炎症性疾病疗法市场 | RA、IBD、银屑病及相关慢性炎症疾病等炎症和免疫介导疾病的药物收入 | 诊断、器械、手术、一般医院服务和无关治疗领域 | 医疗系统、保险方、政府项目和医院渠道 | Xaira 公开披露 I&I 管线方向的最佳广义终端市场视角 |
| 免疫学药物市场 | 用于自身免疫或免疫性疾病的免疫调节药物,包括 mAbs、融合蛋白、免疫抑制剂及其他疗法 | 很多非免疫疗法、一般健康消费,以及广义医院服务 | 与上文相同的下游支付方体系,同时受专科医生处方影响 | 有助于观察相邻终端市场,覆盖更广的自身免疫与免疫学收入池 |
| 抗体生产 / 发现生态 | 与抗体发现和生产相关的研究、开发、制造、耗材和软件 | 不包括成药销售和非抗体生物药模态 | 药企、生物技术公司、CDMO 和研究机构 | Xaira 初期披露的模态重点是抗体,因此这块是平台与产品之间的桥梁 |
| 排除的相邻市场与现状替代方案 | 常规湿实验发现、通用 AI 工具、CRO 服务,以及 Xaira 公开重点之外更广的生物药或制药收入池 | N/A | 多类 | 这些可作为对照项或替代预算,但若没有明确论证,不应并入单一 Xaira TAM |
各行不能相加。它们代表平台支出、模态基础设施和最终治疗价值上彼此重叠但并不等价的几种口径。
[CM001, CM004, CM006, CM011, CM015, CM019]2.2 用多个不可等同的视角衡量机会
平台支出视角最窄,也最接近当下。Mordor Intelligence 估计,AI 药物发现市场 2026 年为 $3.25 billion,到 2031 年以 25.94% CAGR 增至 $10.29 billion。Worldmetrics 报告的高增长区间大体相近但并不相同;McKinsey 引用的制药 AI 市场预测更大,但范围比单纯发现更宽,因此不应当成同口径比较。疗法收入视角大得多,变现也慢得多。第三方对炎症疾病、免疫学和抗炎药市场的估计,2026 年集中在 $122 billion 到 $141 billion 区间;2033–2035 年预测约在 $228 billion 到 $293 billion。模态支持视角介于两者之间:Precedence 估计 2026 年抗体生产市场为 $31.71 billion;Coherent 的更宽口径抗体市场达到 $323 billion,因为它覆盖的疾病领域远超 Xaira 当前披露的重点。正确结论不是把这些数字平均,而是把它们分别保留为平台收入、模态基础设施和终端疗法价值的不同镜头。[CM006, CM007, CM008, CM011, CM013, CM015]
| 发布方 / 来源 | 年份 | 地区 | 市场规模 | CAGR | 方法 | 置信度 | 核心限制 |
|---|---|---|---|---|---|---|---|
| Mordor Intelligence | 2026 | 全球 | $3.25B (2026) → $10.29B (2031) | 25.94% | 按组件、应用、终端用户和地区划分 AI 药物发现市场的自有分段方法 | 中 | 包含较宽泛的 AI 药物发现类别,不是 Xaira 特定平台经济性 |
| Worldmetrics | 2026 更新 | 全球 | $2.3B (2023) → $6.2B (2028);备选:$1.5B (2020) → $10.9B (2030) | 21.9%–24.8% | 汇编多来源统计摘要 | 低 | 属于汇编,不是单一且透明的一手分析师方法 |
| McKinsey | 2025 | 全球 | >$4B (2025) → $25.7B (2030) | n/a | 面向高管的制药 AI 市场增长讨论 | 低 | 口径覆盖更广的制药 AI,并非纯 AI 药物发现估算 |
| Precedence Research | 2026 | 全球 | $133.5B (2026) → $241.34B (2035) | 6.80% | 按疾病、药物类别、给药途径、渠道和地区估算炎症性疾病市场 | 中 | 衡量治疗终端市场收入,而非平台支出 |
| Fortune Business Insights | 2026 | 全球 | $123.05B (2026) → $228.18B (2034) | 8.02% | 按药物类别、适应症和渠道划分免疫学市场 | 中 | 免疫学口径更宽,与炎症性疾病有部分重叠但不完全一致 |
| Global Market Insights(机构) | 2026 | 全球 | $141.3B (2026) → $293.4B (2035) | 8.5% | 按药物类别、治疗方式、给药途径和渠道划分抗炎药市场 | 中 | 采用抗炎药口径,与免疫学部分重叠但并不相同 |
| Coherent Market Insights(机构) | 2026 | 全球 | $122.16B (2026) → $280.35B (2033) | 12.6% | 按药物类别、适应症和渠道估算免疫学市场 | 低 | 边界似乎更宽,部分描述性文本还混入移植相关口径 |
| Precedence Research | 2026 | 全球 | $31.71B (2026) → $93.76B (2035) | 12.83% | 按产品、流程、类型和最终用途划分抗体生产生态 | 中 | 基础设施市场,而非终端治疗收入 |
| Coherent Market Insights(机构) | 2026 | 全球 | $323.04B (2026) → $764.71B (2033) | 13.1% | 覆盖多个疾病领域和终端用户的广义抗体市场 | 低 | 口径过宽,不能直接作为 Xaira 已披露 I&I 重点的 SAM |
各行刻意保留不等价定义。没有明确转换逻辑前,治疗收入池、平台支出市场和支持生态不应混成一个数字。
[CM006, CM007, CM008, CM011, CM013, CM015]Xaira 的三层叠加视角:大型疗法价值池、中等规模抗体基础设施层,以及短期较窄的 AI 平台支出层。
这些层级不能相加,也不是字面上的 TAM/SAM/SOM 瀑布;它们代表由不同来源类别支撑、逐步更接近短期且更窄的变现视角。
[CM006, CM011, CM019, CM021, CM047, CM049]免疫 / 炎症终端市场的独立预测区间显示,2033–2035 年机会集很宽,但规模明确很大。
所有数值都是以 $B 表示的疗法市场预测估算。它们方向上可比,但方法上并不完全一致,因为各来源的类别边界和预测期限不同。
[CM011, CM013, CM015, CM017]2.3 买方、用户与支付方分层
Xaira 近期的经济买方最可能是大型制药和大型生物科技公司的 R&D 或 BD 组织,它们想获得差异化靶点识别、机制或患者筛选能力,但不想把每个组件都内建。Mordor 的终端用户拆分支持这一看法:制药和生物科技公司主导当前支出,学术机构虽重要,但作为直接经济买方处于次级位置。Xaira 自身在重要意义上也是买方,因为自有资产策略要求公司在 AI、湿实验和开发之间做内部资本配置。下游的用户和支付方结构则完全不同。候选物进入试验和商业化后,临床医生、转化研究者和患者成为相关用户;但医院渠道、保险公司和政府支付方决定疗法价值能否转化为收入。Xaira 公开报道和 Bo Wang 访谈还暗示了一个采用顺序:从数据生成,到虚拟细胞预测,再到靶点和分子假设,最后进入临床开发。这个顺序关键,因为它解释了为什么技术采用可能比获批药物带来的财务回报早多年发生。[CM009, CM010, CM020, CM021, CM022, CM023]
| 细分市场 | 买方 | 用户 | 支付方 | 工作流 | 预算负责人 | 采用触发因素 |
|---|---|---|---|---|---|---|
| 大型药企 / 顶级生物制药合作伙伴 | 外部创新负责人、CSO、BD 管理层 | 疾病领域科学家、计算生物学家、转化团队 | 中央 R&D 与合作预算 | 评估模型 / 数据优势 → 试点或尽调项目 → 共同开发 / 期权 / 授权结构 | CSO、R&D 负责人、BD 委员会 | 明确证据显示 Xaira 能提升靶点质量、MoA 洞察、患者分层或发现速度 |
| 新兴生物制药 / 平台合作方 | CEO、CSO 或业务发展负责人 | 小型转化团队和外包发现合作伙伴 | VC 或公开市场资金支持的运营预算 | 借助 Xaira 获取差异化靶点,或减少内部计算 / 湿实验建设 | CEO、CFO、CSO | 比内部搭完整技术栈更便宜或更快地降低风险 |
| Xaira 内部管线团队 | Xaira 管理层和管线委员会 | Xaira AI、湿实验、发现和开发团队 | Xaira 资产负债表资金和已承诺资本 | 生成数据 → 构建模型 → 提名靶点 / 分子 → 推进内部项目 | CEO、COO、CSO、CTO | 内部证据足以支持把项目推进到高成本临床前或临床工作 |
| 学术 / 转化合作方 | PI 和平台负责人 | 实验台科学家、计算生物学家、受训人员 | 科研资助预算、联盟资金、机构研究支持 | 基准测试模型、验证生物学、访问开放数据 / 工具子集 | PI / 资助办公室 | 独特数据访问、发表价值或转化合作机会 |
| 临床医生 / KOL / 试验研究者 | 医学和临床开发团队招募他们,而不是直接向他们销售 | 医生、研究者、患者 | 初期由申办方试验预算支付;获批后由支付方承担 | 生物标志物假设 → 试验设计 → 站点选择 → 临床证据生成 | CMO / 开发管理层 | 机制可信度,加上临床可执行的患者选择逻辑 |
| 医疗系统 / 保险方 / 医院渠道 | 药品路径上的处方集和覆盖委员会;不是平台本身买方 | 患者、输注中心、医院药师、专科临床医生 | 商业保险方、Medicare / Medicaid 类机构、政府系统 | 获批 → 指南支持 → 处方集覆盖 → 配药和报销 | 药房 / 医疗预算委员会 | 差异化疗效和安全性,同时净价与报销条件可接受 |
这张地图把平台销售路径和自有药物路径拆开看,因为即使底层科学栈共享,买方和支付方也有实质差异。
[CM009, CM020, CM021, CM022, CM023, CM024]基于证据的序数图,展示今天谁会为类似 Xaira 的能力付费,以及谁只有在药物上市后才重要。
序数评分反映基于引用来源的尽调判断:1=低,2=中,3=高。它们不是来自调研的量化指标。
[CM009, CM020, CM022, CM023, CM024, CM025]从 Xaira 的数据和模型栈走到最终疗法报销,需要多阶段推进,并跨过差异很大的买家群体。
[CM002, CM004, CM005, CM021, CM024, CM042]2.4 增长驱动与采用约束
AI 赋能生物制药的乐观情景很直接:行业 R&D 生产率偏弱,新药上市成本仍然巨大,而平台团队现在能用到的算力、数据和实验室自动化能力,都比几年前更好。Deloitte 的 2025 年调查显示,实验室现代化带来了实实在在的吞吐提升和错误减少;McKinsey 和 Accenture 也都把数字化与 AI 工具视为应对资本效率低下的必要手段。Xaira 正处在这个逻辑里,因为它的模式依赖生成自有扰动数据,并把数据回灌到模型开发。但约束情景同样真实。McKinsey 称,制药行业还没有看到开发时间线或成功率的系统性改善。Mordor 强调可解释性、数据碎片化、人才短缺和责任不确定性。ACS 和 GEN 都保留了残酷基线:药物发现仍是十年以上流程,临床候选物中只有约 1/10 能获批,多数分子仍会在商业成功前失败。即便 Xaira 的科学有效,免疫学定价、不良反应风险和生物类似药压力,也意味着庞大的终端市场不会自动转化为不受约束的定价权。[CM026, CM027, CM028, CM029, CM030, CM031]
| 驱动因素 / 约束 | 方向 | 时间维度 | 对 Xaira 的影响 | 尽调问题 |
|---|---|---|---|---|
| 生物制药 R&D 生产率压力和成本通胀 | 增长驱动 | 近期且结构性 | 如果能提高命中质量或缩短周期,差异化 AI / 数据平台具备经济吸引力 | 相比标准发现工作流,Xaira 声称可量化 ROI 的具体环节在哪里? |
| 自有多模态湿实验数据和实验室在环学习 | 增长驱动 | 近期 | 如果 Xaira 数据规模保持独特,就可能形成可防御的模型优势 | 相比同业,X-Atlas、X-Cell 和未来扰动数据集的自有性和可复现性有多强? |
| 规模大且仍在增长的免疫学 / 炎症性疾病价值池 | 增长驱动 | 长期 | 如果 Xaira 能把平台洞察转化为获批资产,最终治疗上行空间很大 | Xaira 会优先推进哪些 I&I 适应症和患者亚群? |
| 免疫学领域内生物药和抗体占主导 | 增长驱动 | 近期和长期 | Xaira 已披露的抗体重点,与当前治疗支出本就集中的方向匹配度更高 | Xaira 能否展示针对现有玩家难以成药靶点的抗体设计? |
| 可解释性、审计追踪和监管文档要求 | 采用约束 | 近期 | AI 系统若影响候选物选择或临床决策,举证门槛会被抬高 | 面向内部和合作伙伴使用,Xaira 如何记录模型血缘、验证和决策边界? |
| 数据碎片化和 AI / 生物学人才稀缺 | 采用约束 | 持续 | 平台扩张需要跨职能人才和治理良好的数据管线,二者都稀缺 | Xaira 的优势集中在少数领导者身上,还是已经固化到制度化系统和流程中? |
| 临床转化风险和漫长药物开发周期 | 采用约束 | 持续 | 即使发现结果很强,也可能需要十年以上才能转化为获批药物或经济性证明 | 在获批之前,投资者或合作伙伴应期待 Xaira 给出哪些中间证明点? |
| 免疫学中的定价、报销、不良反应风险和生物类似药压力 | 采用约束 | 持续 | 限制名义 I&I 市场中有多少能转化为新进入者的溢价定价 | 对未来任何 Xaira I&I 资产,哪些净价和报销假设才现实? |
这些驱动和约束来自公开市场、咨询和技术来源中最反复出现的主题,并非穷尽所有生物制药变量。
[CM012, CM014, CM016, CM026, CM027, CM028]2.5 市场证据对 Xaira 意味着什么
市场证据支持对 Xaira 作更细的解读。公司把自己放在一个极大的疗法问题集和快速增长的平台类别里;早期 I&I 加抗体的取向,也让它有了具体切口,而不是泛泛讲“AI 赋能生物学”。与此同时,本报告审阅的公开来源没有量化虚拟细胞模型、抗体设计和炎症疾病疗法交叉处的精确可服务市场。这不是小的建模麻烦,而是真实尽调缺口。后续章节因此不应假装 Xaira 的 SAM 已经被清楚测量。竞争分析应测试买方究竟把 Xaira 视为平台供应商、特定模态生物科技公司,还是未来一体化药物公司。估值工作也应把平台期权价值与临床资产价值分开,而不是给整个故事套一个笼统倍数。[CM004, CM021, CM026, CM042, CM043, CM045]
2.6 展示材料
03竞争格局
3.1 哪些竞争者真正影响 Xaira
Xaira 不能只拿另一个 AI 生物科技公司当唯一参照。已披露模式把前沿机器学习、自有扰动数据生成和疗法产品开发结合起来;2026 年报道还特别指向炎症和免疫抗体疗法这个第一切口。因此,相关竞争版图至少有四层。第一层是直接的以 AI 为先的疗法平台,例如 Generate、insitro、Recursion/Exscientia、Isomorphic Labs 和 Absci,它们公开把差异化模型与实验系统或内部资产创建结合。第二层是相邻的生物制剂设计专家,例如 Chai Discovery 和 Nabla Bio,它们争夺同一买方集合里的抗体和蛋白质设计预算。第三层是替代工具链,例如 Schrödinger,其物理 + AI 软件和合作项目让买方可以不购买 Xaira 式一体化平台,也能解决同一发现任务的一部分。第四层是大型制药公司的内部自建,Lilly、Novartis、Sanofi、Roche 等可以把自有数据与多个外部伙伴组合起来。因此,竞争问题不是 Xaira 有没有同行,而是未来 2–3 年哪一层最直接约束其定价权和交易条款。[CP001, CP002, CP003, CP004, CP005, CP024]
基于证据的序数定位:X 轴为专有生物数据 / 湿实验室整合(1=以计算为主,10=非常强的一体化数据引擎),Y 轴为商业-临床验证(1=早期 / 未披露,10=广泛合作与临床证据)。
评分是定性但有证据支撑。它们反映截至 2026 年报告运行日的公开披露,用于比较相对位置,不量化内在价值。
[CP003, CP006, CP009, CP012, CP015, CP021]3.2 直接同行与相邻对手
直接同行可以分成两种模式。Generate、insitro、Recursion/Exscientia 和 Absci 都公开描述了一条整合闭环:自有数据和湿实验系统持续改善模型,并把结果输送到合作项目或内部项目。Isomorphic 的重点不同:公开叙事更明确围绕建立在 AlphaFold 之上并超越 AlphaFold 的前沿预测和生成模型质量;商业证明则来自非常大的制药合作,而不是已披露的内部管线深度。Xaira 在战略上似乎最接近整合数据公司,但它已宣布的技术旗舰仍是虚拟细胞和扰动数据栈,而不是公开临床资产或具名商业合作。在这个核心周围,还有范围更窄但仍相关的抗体和蛋白质挑战者。Chai Discovery 推广具有原子级精度的 de novo 抗体设计;Nabla Bio 强调 AI 加人类相关湿实验测试,用于针对困难靶点的抗体和其他蛋白质疗法。这些公司未必是 Xaira 的直接全栈匹配,但会争夺一部分相同的生物制剂买方注意力,也让抗体设计不再是空白市场,而是拥挤切口。[CP003, CP004, CP006, CP009, CP012, CP015]
| 竞品 | 类别 | 规模 / 融资信号 | 目标客群 | 差异化 | 局限 |
|---|---|---|---|---|---|
| Generate:Biomedicines | 直接同业 —— 一体化 AI 生物药平台 | 140k+ sq ft 场地;已生成 / 构建 / 测试 42,000 个蛋白;Novartis 交易带来 $65M 首付款和 >$1B 里程碑;Fierce 提到 Amgen 最高 $1.9B,以及两个临床资产 | 大型药企生物药团队和内部蛋白疗法管线 | 生成式生物学平台,跨蛋白模态持续运行生成—构建—测量—学习闭环 | 主要聚焦蛋白 / 生物药;公开定价只在定制合作中可见 |
| Isomorphic Labs | 直接同业 —— 前沿 AI 药物设计引擎 | Alphabet / DeepMind 支持;Lilly + Novartis 合作据报道合计接近 $3B | 大型药企战略发现合作 | 在 AlphaFold 之外延展,公开发布的结构、亲和力、口袋和抗体界面主张较强 | 关于内部临床管线深度或已披露湿实验规模,公开证据很少 |
| insitro | 直接同业 —— 因果生物学 + ML 平台公司 | 2025 年公司发布称累计融资 >$700M;2026 年新闻材料提到约 $800M 融资和约 $150M 合作收入 | 大型药企合作,加上代谢、肿瘤和神经科学内部管线 | 将人类临床数据和细胞数据与 ML 整合;可谈判灵活权益结构 | 公开经济条款仍是定制化,内部临床验证仍有限 |
| Recursion / Exscientia | 直接同业 —— 全栈技术生物平台 | >50 PB 自有数据;Sanofi 交易从 $100M 起,可超过 $5.2B;Bayer 交易最高 $1.5B;合并在签约时新增精准化学能力和约 $850M 现金 | 大药企平台买方,加上内部罕见病和肿瘤管线 | 已披露组合最宽,覆盖数据生成、表型组学、患者数据和小分子化学 | 近期项目裁减和资本纪律压力是有意义的反向信号 |
| Absci | 直接同业 —— AI 生物药 / 抗体设计 | 77,000+ sq ft 湿实验室;SoluPro 中有数十亿细胞;ACE 测定吞吐量 >4,000x;6 周学习闭环 | 药企生物药团队和内部 / 合作生物药项目 | 抗体从头设计,加上湿实验吞吐量和反向免疫学靶点发现 | 生物药范围较窄,公开定价信息稀少 |
| Chai Discovery | 相邻专才 —— 抗体设计初创公司 | 官网强调 Chai-2 访问;Fierce 报道近期完成 $130M Series B 轮 | 聚焦针对困难靶点的抗体从头设计团队 | 围绕原子级精度抗体设计的定位高度具体 | 公开商业模式、经济条款和管线范围仍稀少 |
| Nabla Bio | 相邻专才 —— 生成式蛋白设计 | 2024 年 $26M Series A 轮,加上 >$550M 合作;2025 年 Takeda 交易增加数千万美元首付款和 >$1B 成功付费 | 面向困难膜蛋白和抗体的药企蛋白疗法团队 | JAM 基础模型,加上针对困难靶点的人体相关湿实验测试 | 在模态广度和公开管线透明度上似乎窄于 Xaira |
| Schrödinger | 替代方案 / 既有工具链,加上合作管线 | 约 800 名员工;多个合作项目覆盖发现到 3 期;Lilly 免疫学合作最高 $425M,另有特许权使用费 | 需要计算设计和企业信息学的药企发现组织 | 物理 + AI 平台,小分子深度已获验证,并具备一定生物药能力 | 定位不同于 Xaira 这种因果细胞、湿实验优先的平台 |
| 大型药企内部自建 | 现状 / 内部自建替代方案 | 自筹 R&D、拥有自有临床和临床前数据,并能并行运行多个外部合作 | 与 Xaira 想切入的顶级药企买方群体相同 | 可以把内部资产与多个供应商的外部工具组合,而不是绑定单一外部平台 | 构建前沿能力可能更慢,但降低对任何单一初创供应商的依赖 |
行项并不穷尽。覆盖的是与 Xaira 当前披露平台和生物药策略最相关的公开竞品和替代方案。
[CP003, CP004, CP005, CP006, CP009, CP012]| 采购标准 | Xaira | Generate | Isomorphic | insitro | Recursion / Exscientia | Absci | Schrödinger |
|---|---|---|---|---|---|---|---|
| 自有扰动 / 细胞系统数据引擎 | 强 —— X-Atlas / X-Cell 因果扰动栈已公开强调 | 部分 —— 蛋白数据闭环强,但未被呈现为虚拟细胞平台 | 未知 / 部分 —— 模型能力公开,湿实验数据规模披露较少 | 强 —— 平台整合人类临床 + 细胞数据 | 强 —— 表型组学、组学、ADME 和患者数据规模达 >50 PB | 部分 —— 有生物药训练数据和湿实验,但不是广义细胞系统平台 | 部分 —— 强调软件和模拟;自有生物数据引擎不够明确 |
| 抗体 / 蛋白从头设计 | 部分 / 起步 —— 抗体疗法被披露为早期切入口 | 强 —— 定制蛋白、抗体、酶和其他模态 | 部分 —— 公开证据包含抗体-抗原建模能力 | 部分 —— 具备生物药能力,但不是公开叙事重点 | 部分 —— 平台范围宽于生物药设计本身 | 强 —— 抗体从头设计和生物药是核心产品叙事 | 部分 —— 支持生物药设计,但不是核心市场叙事 |
| 小分子发现广度 | 未知 —— 查阅来源中未公开披露 | 有限 / 未知 —— 公开叙事集中在蛋白疗法 | 强 —— 小分子合作是已披露 GTM 的核心 | 强 —— 公开强调 ChemML 和 ADMET 建模 | 强 —— Exscientia 合并和多个小分子项目拓宽范围 | 未知 / 有限 —— 公开重点是生物药 | 强 —— 核心业务仍是小分子计算发现 |
| 一体化湿实验验证 | 强 —— 数据生成是三大核心要素之一 | 强 —— 生成—构建—测量—学习闭环明确 | 未知 —— 公开技术证据强,但湿实验整合细节有限 | 强 —— 体外和临床数据生成处于核心位置 | 强 —— 自动化湿实验室,每周数百万次实验 | 强 —— ACE 测定和 SoluPro 构成高通量验证闭环 | 部分 —— 平台支持发现,但公开定位不太以湿实验为中心 |
| 公开披露的药企合作 GTM 经济条款 | 未知 —— 未找到具名外部平台合作条款 | 强 —— Novartis 和 Amgen 经济条款已有公开讨论 | 强 —— Lilly / Novartis 后端经济条款已有公开报道 | 强 —— Lilly、BMS 和其他结构已有公开描述 | 强 —— Sanofi 和 Bayer 条款公开 | 弱 / 未知 —— 能力公开,但经济条款披露少得多 | 强 —— Lilly 和其他合作经济条款公开 |
| 公开临床 / 项目可见度 | 查阅来源中未披露 | 公开提到两个临床候选物 | 未找到已披露的内部临床资产组合 | 管线公开,但临床阶段证据仍有限 | 多个 I/II 期和合作项目公开 | 内部和合作项目公开,但阶段更早 | 自有 1 期项目和合作 2/3 期项目公开 |
该矩阵刻意保守。缺少支撑的单元格标为“未知”,而不是从类别标签或营销语言推断。
[CP027, CP028, CP029, CP030, CP041, CP043]横跨五个竞争维度的相对能力压缩视图。未知表示审阅来源未明确确认该能力。
条目为定性判断,并基于引用来源。该图刻意比详细表更窄,用来呈现模式差异,而不是重复每个表格单元。
[CP021, CP022, CP025, CP031, CP041, CP043]3.3 市场公开定价与打包方式
这个同行群体的公开货币化数据指向同一个方向:定制研究合作、期权结构、里程碑、版税和偶发股权是常态,透明标价基本缺席。Generate 与 Novartis 的合作包含 $65 million 首付款、股权、十亿美元级里程碑和版税。Recursion 与 Sanofi、Bayer 的合作,在已披露后端经济条款上规模更大。Isomorphic 的首批 Lilly 和 Novartis 交易据报道合计接近 $3 billion,尽管没有公开内部临床管线。insitro 的打包方式更灵活:在部分 Lilly 项目中,它保留全球权利,由 Lilly 提供赋能技术或获得里程碑和版税;insitro 与 BMS 的合作则支付靶点选择里程碑。Nabla 显示,即便规模更小的抗体设计初创,只要战略制药买方相信平台能打开困难靶点,也能拿到千万美元级首付款和十亿美元级后端经济条款。相比之下,Xaira 尚未公开披露伙伴定价、具名平台客户或任何临床项目经济条款。这个缺口不证明公司弱,但意味着财务核算必须依赖可比合作结构,而不是 Xaira 自身的定价证据。[CP007, CP011, CP013, CP014, CP016, CP023]
| 公司 / 包装 | 公开披露经济条款 | 包含能力 | 未知项 / 缺口 | 含义 |
|---|---|---|---|---|
| Xaira | 所审阅来源未发现公开定价、合作伙伴经济条款或具名外部平台合同 | AI 研究、数据生成、虚拟细胞研发与疗法叙事一体化 | 无公开标价、无具名商业伙伴、无公开项目经济条款 | 最难直接做投资测算的变量;后续章节必须用可比交易替代 |
| Generate / Novartis | $65M 首付款现金(含 $15M 股权);>$1B 里程碑;分层销售分成最高达低两位数 | 用 Generate Platform + Novartis 生物学 / 开发能力做多靶点蛋白疗法发现 | 靶点数量和具体疾病领域未披露 | 高溢价生物制剂平台标杆,合作方付费意愿清楚 |
| Isomorphic / Lilly + Novartis | 媒体报道称,Lilly 首付款 $45M + 最高 $1.7B;Novartis 首付款 $37.5M + 研究经费 + 最高 $1.2B | 借助 Isomorphic 引擎开展多靶点 AI 赋能小分子发现 | 官方公开条款细节仍少于媒体摘要;内部资产经济性未披露 | 即使没有公开的内部临床验证,前沿模型可信度也能支撑十亿美元级后端收益 |
| insitro / Lilly 与 BMS | BMS 靶点扩展触发 $10M 里程碑;Lilly 的架构让 insitro 在部分项目保留全球权利,而 Lilly 获得里程碑 / 销售分成或提供技术 | ALS 靶点、siRNA 递送、抗体,以及 ADMET / 小分子模型构建 | 不同项目的经济条款各异,许多首付款仍未披露 | 平台价值足够差异化时,可以打包出灵活且有利于生物技术公司的条款 |
| Recursion / Sanofi + Bayer | Sanofi $100M 首付款 + 最高 $5.2B 加销售分成;Bayer 最高 $1.5B 加销售分成 | 靶点发现、精准设计、先导优化和更广的平台合作 | 单项目分成、排他性和实际兑现价值未完全公开 | 该赛道已披露规模最大的全栈 AI 发现可比组 |
| Nabla / Takeda + 2024 年合作伙伴组合 | 2025 年 Takeda 交易:数千万美元首付款 + 研究成本付款 + >$1B 成功付款;2024 年合作包 >$550M 加销售分成 | 借助 JAM + 湿实验室做从头设计抗体、多特异性分子和其他定制疗法 | 具体靶点数量和里程碑时间未披露 | 即便规模更小的生物制剂设计初创公司,也能拿到很大的后端经济条款 |
| Schrödinger 与 Lilly | 发现、开发和商业里程碑最高 $425M,加低个位数至低两位数销售分成 | 计算设计与发现项目;另有 LiveDesign 企业软件业务 | 公开来源未披露软件席位定价或实际定价组合 | 替代供应商主要仍靠带里程碑的合作来变现药物创造工作 |
在审阅的同业中,没有公司公布 AI 药物发现访问权的透明标价。公开经济条款几乎都基于合作。
[CP007, CP011, CP013, CP014, CP016, CP023]紧凑评分板,展示 Xaira 相比审阅同业最影响决策的公开准备度和拥挤度指标。
所有数值都是有证据支撑的计数或直接公开披露。部分负向差值反映市场拥挤或缺少公开证据,而不是业务失败。
[CP025, CP026, CP027, CP041, CP043]3.4 切换成本、多平台并行与护城河耐久性
公开记录显示,AI 药物发现的护城河形成仍然更少来自传统软件锁定,更多来自数据所有权、工作流整合、内部实验基础设施和谈判获得的资产权利。这对 Xaira 是双刃剑。如果它的扰动数据引擎和虚拟细胞模型确实能比同行更好地发现靶点或患者假设,由此形成的工作流与资产整合成本可能有意义。但大型制药买方似乎不会只押一个平台。Lilly 与 Isomorphic、insitro 和 Schrödinger 合作;Novartis 与 Isomorphic 和 Generate 合作;Sanofi 和 Bayer 与 Recursion 合作;Nabla 等早期生物制剂挑战者也在赢得项目。买方多平台并行是理性的,因为多数合作只披露部分经济条款、排他性不清,对长期项目结果的证据也有限。这里的反向证据很重要。Recursion 与 Exscientia 合并拓宽了技术栈,但随后出现管线削减、持续聚焦收缩,以及投资人对烧钱的担忧。这是重要的反证信号:更多数据、更多项目和更多资本,并不会自动变成持久执行力或定价权。[CP024, CP031, CP032, CP033, CP034, CP038]
| 护城河主张 | 威胁 | 严重性 | 缓释措施 / 尽调问题 |
|---|---|---|---|
| 因果虚拟细胞与扰动数据护城河 | Recursion 和 insitro 已经运营大型专有数据闭环,公开证据尚未把 Xaira 的因果技术栈与它们做正面对标 | 重大 | 要求提供与同业平台正面对比的证据:靶点提名质量、未见生物学泛化能力,以及从命中物到先导物阶段的影响 |
| 生物制剂设计差异化 | Generate、Absci、Chai 和 Nabla 已在抗体与蛋白疗法周围挤满 AI 生物制剂切口 | 高(近期) | 厘清 Xaira 的优势究竟是因果靶点选择、患者匹配或跨模态设计,而不只是抗体 |
| 合作定价权 | 大型药企买方似乎愿意同时与多家 AI 供应商多栖合作,削弱排他性和议价能力 | 重大 | 要求提供排他性、工作流嵌入或权利结构证据,说明 Xaira 一旦接入后更难被替换 |
| 资本作为护城河 | Recursion 并购后的管线削减表明,规模和现金不保证执行可持续,也不保证烧钱更低 | 重大 | 在激进的数据生成、实验室扩张和内部管线情景下建模 Xaira 的烧钱强度,而不是假设初始弹药足够 |
| 信任 / 监管姿态 | AI 药物发现走向成熟后,买方可能要求模型可审计性、可复现性和治理,而不只是基准测试主张 | 中等 | 要求提供验证方案、数据来源文档,以及用于合作伙伴沟通的任何治理材料 |
| 内部自建替代 | 大型药企可以把内部数据与 Isomorphic、Schrödinger、insitro、Recursion 等合作伙伴组合起来,而不必标准化到一个外部平台 | 重大 | 识别哪些用例中,Xaira 能比内部 + 合作伙伴混合栈明显更快、更精准或更省资本 |
严重性评级是定性尽调判断。高表示威胁可能在约 2–3 年内损害合作伙伴经济条款;重大意味着约 3–5 年内有实质影响;中等表示需要密切跟踪,但证据仍不完整。
[CP028, CP031, CP032, CP033, CP034, CP038]3.5 竞争版图对 Xaira 意味着什么
同行集合说明,Xaira 不需要证明 AI 赋能疗法平台存在需求。Generate、Isomorphic、insitro、Recursion、Nabla 和 Schrödinger 披露的制药合作数量与规模,已经让需求可见。Xaira 仍需要公开证明的是自己落在光谱的哪个位置。它最强的已披露技术主张,是 X-Cell 的规模及其背后的扰动数据集;这指向围绕因果细胞生物学的可信科学切口,而不只是生成式序列设计。这一点重要,因为生物制剂设计市场已经被 Generate、Absci、Chai 和 Nabla 挤满。如果 Xaira 在因果靶点选择、患者匹配或跨模态疗法设计上显著更强,它可能配得上更接近大型全栈平台的可比公司组。否则,买方很可能一边等待更清楚证据,一边在更成熟同行之间多平台并行。后续财务和估值工作因此应追问,管理层能否展示伙伴兴趣、排他性或内部资产进展,把 Xaira 从“科学上令人印象深刻”推进到“商业上可对标”。[CP027, CP028, CP040, CP041, CP042, CP043]
3.6 展示材料
04财务情况
4.1 收入模型:合理但尚未获公开验证
Xaira 官方材料描述的是一家围绕 AI 研究、大规模数据生成和疗法产品开发建设的公司。这足以识别合理的收入路径,但不足以证明其中任何一条已经启动。本报告审阅的公开来源没有披露 Xaira 收入、年经常性收入(ARR)、合同价值、合作伙伴研究经费、里程碑收款或产品销售。结合商业模式和竞争集合,近期最可信的货币化路径不是药品销售或软件订阅,而是合作收入、里程碑付款、版税,以及可能围绕内部生成资产的期权或对外授权交易。可比 AI 生物制药平台今天就是这样变现的。Generate、Isomorphic、insitro、Recursion 和 Nabla 都展示了里程碑重、带版税的结构:制药公司在产品获批前为差异化科学买单。Xaira 未来可能开发自有疗法并获得产品销售,但本报告审阅的公开来源没有显示具名临床阶段 Xaira 资产,也没有显示外部客户已经为平台付费。因此,收入质量还不能被视为经常性、多元化,甚至已经开始。[CI003, CI004, CI005, CI006, CI020, CI021]
| 收入流 | 机制 | 单位 | 当前价值 / 状态 | 质量 | 尽调问题 |
|---|---|---|---|---|---|
| 平台合作 / 研究经费 | 药企为访问 Xaira 的模型、数据、工作流或共同开发能力付费 | 每年或每项目 $ | Xaira 未披露公开金额;如存在,这是近期最可能的变现路径 | Xaira 尚未验证;同业案例显示可行,但高度定制 | 要求提供任何已签的平台或共同开发协议、年度研究经费和收入确认处理 |
| 里程碑付款 | 靶点提名、候选物选择、IND、临床进展或批准触发的项目专项付款 | 每个里程碑 $ | Xaira 未披露公开里程碑包 | 潜在价值高,但二元性强且后端集中 | 要求提供里程碑时间表、触发定义和概率加权时间假设 |
| 销售分成 / 利润分成 | 合作资产销售的下游分成 | 净销售额百分比或利润分成 | Xaira 未披露公开销售分成经济条款 | 周期长,且高度依赖合作伙伴和临床成功 | 要求提供分成层级、保留权利、地域范围和期限 |
| 内部资产对外授权 / 期权交易 | 第三方在早期风险降低后购买或获得 Xaira 原创资产的期权 | $ 首付款 + 里程碑 + 销售分成 | 未披露公开资产交易 | 如果 Xaira 在完全自商业化前优先推进平台生成资产,这一路径可行 | 厘清管理层按模态和阶段更偏好资产出售、共同开发还是保留所有权 |
| 内部疗法销售 | Xaira 出资把项目推进到临床开发,并最终销售获批药物 | $ 产品净销售额 | 未披露 Xaira 公开临床阶段产品或收入 | 周期极长,风险最高 | 要求提供管线阶段图、目标产品画像,以及各项目的任何商业化意图 |
该表把可能的变现逻辑与已披露财务现实拆开。公开证据支持逻辑,但不支持当前已实现收入。
[CI003, CI004, CI005, CI006, CI020, CI021]| 价格 / 单位 / 合同 | 标价与实际定价 | 折扣 / 未知项 | 来源 / 含义 |
|---|---|---|---|
| Xaira 平台 / 合作 | 未披露公开价格或合同结构 | 未披露标价、实际价格或已确认收入 | 公开沉默意味着估值必须使用同业可比,而非公司特定定价证据 |
| Generate / Novartis:$65M 首付款 + >$1B 里程碑 + 销售分成 | 交易条款只披露到摘要层面,但实际兑现时间取决于靶点数量和进展 | 靶点数量和疾病领域未披露 | 为高溢价生物制剂平台变现定标 |
| Isomorphic / Lilly + Novartis:报道首付款 $45M 和 $37.5M,后端经济条款大得多 | 经济条款来自媒体报道,并非公开价目表 | 官方条款细节仍少于媒体摘要 | 说明前沿模型可信度可以通过药企战略交易变现 |
| insitro / Lilly + BMS:灵活权利结构 + 里程碑 | 实际经济条款因项目而异,部分架构让 insitro 保留全球权利 | 许多具体首付款仍未披露 | 暗示 Xaira 未来可能拥有多种变现模板,而不是一个标准合同 |
| Recursion / Sanofi + Bayer:$100M 首付款 + 最高 $5.2B;Bayer 最高 $1.5B | 实际收入和里程碑取决于进展和确认规则 | 单项目分成和排他条款不清楚 | 规模化全栈 techbio 模型的上沿标杆 |
| Nabla / Takeda:数千万美元首付款 + >$1B 后端潜力 | 仍基于合作,而非标价销售 | 时间和靶点数量未披露 | 只要科学有差异化,更早期的生物制剂设计公司也能很好变现 |
没有一家已审阅公司公布 AI 药物发现访问权的透明标价。同业变现仍高度定制,且里程碑占比高。
[CI020, CI021, CI029, CI032, CI033, CI034]Xaira 当前活动可能如何转化为收入,以及为什么这座桥在公开信息中仍主要是假设。
[CI004, CI005, CI020, CI021, CI032, CI039]4.2 成本结构与资本强度
现有公开证据指向资本密集型成本底盘。Xaira 启动时的使命,就是同时资助机器学习研究、数据生成和疗法开发。官方 X-Cell 材料称,公司正从 25.6 million 个受扰动单细胞转录组,走向覆盖原代细胞、类器官和体内扰动的更广数据集。Drug Discovery Trends 引述 Bo Wang 称,AI 成功的三根支柱是人才、算力和数据,并称 Xaira 有资金同时追这三项。GeekWire 报道 2024 年约有 80 名员工,多数在 Bay Area,15 人在 Seattle;Intelligence360 的地产信息则报道了 73,075 平方英尺的 San Francisco 扩建项目。Fierce 引述 COO Jeff Jonker 称,整合 R&D 平台和临床测试计划需要多年时间,可能要花费 $1 billion 或更多。换句话说,Xaira 的成本结构更接近临床阶段 techbio 公司,而不是精简软件初创。人才、湿实验运营、算力以及临床前 / 临床项目支出,几乎肯定是主要现金消耗项。[CI002, CI007, CI008, CI009, CI010, CI011]
| 指标 | 数值 / 空值 | 置信度 | 重要性 | 尽调问题 |
|---|---|---|---|---|
| 公开外部收入 | 未知 / 未披露 | 低 | 决定 Xaira 是否已用合作伙伴现金抵消烧钱,还是仍完全靠股权资金 | 要求提供 2025 年和 2026 年截至目前的合作收入或其他经营收入 |
| 员工数 / 场地规模代理 | 启动时约 50 名员工(Endpoints),2024 年约 80 名员工,分布于湾区、西雅图和伦敦;报道称旧金山建设面积 73,075 sq ft | 中 | 衡量人员、实验室和租用场地成本强度的实用代理 | 要求提供按职能划分的当前员工数,以及所有有效实验室 / 办公室租赁承诺 |
| 资本密集型支出组合 | 高;公开来源显示,人才、算力、数据生成、湿实验室和疗法开发是主要支出项 | 中 | 解释为什么不应按纯 AI 软件初创公司的烧钱或利润率来对标 Xaira | 要求提供 2025 年 AI、数据生成、平台运营和疗法方面的支出分配 |
| 估算年度烧钱代理 | $120M–$260M 每年(基于 Xaira 规模信号和同业公开结果的极低置信度估算) | 低 | 在缺少当前现金披露时,这是情景建模资金续航期的必要输入 | 要求提供截至 2026 年 Q1 的实际月度烧钱、现金经营费用和年化运行率 |
| 已实现收入毛利率 | Unknown | 低 | 如果 Xaira 通过研究经费或里程碑变现,最终毛利率可能较高;如果把资产内部化,毛利率路径会长得多,也更耗资本 | 要求提供各变现路径下的收入组合假设和销售成本结构 |
| 项目级开发成本 | Unknown | 低 | 内部资产支出是稀释风险和资金续航期压缩的最大摇摆因素 | 要求提供每个主要项目从当前阶段推进到 IND 并完成 Phase 1 的成本 |
该表有意明确列出空值,因为 Xaira 的私营状态拿掉了生物技术投资测算通常依赖的季度和年度财务披露。
[CI007, CI008, CI009, CI012, CI019, CI024]Xaira 单位经济性链条中公开可见的环节;私有披露仍必需的地方明确标注未知。
该图是定性的,因为 Xaira 最重要的数字输入——当前现金、收入和实际烧钱——并未公开。它映射单位经济性链条中哪些环节有证据支撑,哪些仍未知。
[CI012, CI019, CI022, CI024, CI036, CI037]Xaira 现金可能流向何处,以及各支出桶在短期压力和披露质量上的对比。
评级是定性且有证据支撑。「未知」表示公开证据没有量化该支出桶,不代表成本不存在。
[CI007, CI008, CI009, CI010, CI011, CI012]4.3 资本充足性与现金跑道
最难的财务判断,是 Xaira 的巨额启动轮到 2026 年是否仍留下充足现金跑道。答案方向上是肯定的,但也仅是方向上。启动时超过 $1 billion 的承诺资本是罕见优势,应当给管理层足够灵活性去建设人员、数据和平台基础设施,而不必马上回到融资市场。但启动资本不等于当前账面现金,本报告审阅的来源也没有披露今天还剩多少。这迫使任何现金跑道分析都进入基于同行的情景推演。Recursion 在 2025 年底有 $753.9 million 现金,此前经营现金流出 $371.8 million,但仍只预计现金跑道到 2028 年初。Schrödinger 在 2025 年底有 $402.3 million 现金,经营费用为 $309.5 million,并有真实软件和药物发现收入支撑。Relay 在 2025 年 Q1 即便削减成本,仍有 $710.3 million 现金;Absci 在 2025 年 Q3 有 $152.5 million,现金跑道只到 2028 年上半年。这些数字给出一个宽泛但有用的 Xaira 烧钱参照:每年低数亿美元。这个区间不意味着马上承压,但意味着资本充足性最终会取决于商业化前能否实现合作货币化或内部管线证明。[CI001, CI013, CI014, CI015, CI016, CI017]
| 资本输入 | 公开数值 / 估算 | 置信度 | 重要性 | 尽调问题 |
|---|---|---|---|---|
| 启动时承诺资本 | 2024 年承诺资本 >$1B | 高 | 为私营 AI 生物制药建设提供异常庞大的起步现金池 | 确认其中多少立即到位、多少随时间承诺,以及是否仍有未调用部分 |
| 当前账面现金 | 公开未知 | 低 | 融资风险和资金续航期的最重要单一变量 | 要求提供截至最近月末的资产负债表现金、等价物和有价证券 |
| 月度烧钱代理 | 年度烧钱估算区间隐含每月 $10M–$22M | 低 | 把启动资本转换为实际执行时间窗口 | 要求提供当前总烧钱、扣除合作方流入后的净烧钱,以及按季度的烧钱趋势 |
| 资金续航月数 | 公开不可得;从 2026 年运行日期起,取决于剩余现金和实际烧钱,粗略情景区间约 24–72 个月 | 低 | 界定 Xaira 多快需要合作变现或另一次融资事件 | 要求管理层在基准、内部管线偏重和合作偏重情景下的资金续航模型 |
| 计划资金用途 | AI 模型开发、数据生成、多模态疗法开发,以及数据 / 实验闭环扩张 | 高 | 解释为什么收入可见前现金部署可能加速 | 要求提供 2026–2028 年按平台、数据和项目组合划分的资本配置计划 |
| 下一轮 / 战略触发点 | 可能是重大合作经济条款或内部资产验证,而非近期产品收入 | 中 | 决定稀释或战略合作压力何时变得尖锐 | 要求提供董事会层面的融资计划,以及管理层认为重返资本市场前必须达到的里程碑 |
| 债务 / 项目融资义务 | 公开未披露 | 低 | 未披露固定义务可能实质改变资金续航期分析 | 要求提供债务、租赁、云 / 计算承诺,以及任何设备融资时间表 |
该表有意区分已披露事实、情景估算和无法获得的私营数据。启动资本不能替代当前现金披露。
[CI001, CI013, CI014, CI015, CI016, CI017]Xaira 年度现金消耗估算的低 / 基准 / 高情景,以公开同业数据和 Xaira 披露规模为基础。
这些是低置信度、证据支撑的情景估算,不是公司指引。低情景锚定 Absci 等较小同业,基准情景锚定 Xaira 披露的多站点和数据野心,高情景锚定 Recursion 等更大全栈临床同业和后期临床支出模式。
[CI019, CI023, CI024, CI025]4.4 公开牵引力缺口与私有指标依赖
缺失的信息比已有信息更重要。Xaira 不发布财务报表、SEC 文件、季度现金余额、合作收入、伙伴合同、债务时间表、租赁义务或项目级 R&D 分配。公司是私营企业,这并不意外,但会带来真实分析后果。Schrödinger 等上市同行有监管文件和年度财务;Recursion、Relay 和 Absci 至少会发布定期业绩或现金跑道声明。Xaira 没有。因此,核心单位经济问题仍无答案:月度烧钱是多少?支出中核心 AI、湿实验和内部疗法各占多少?是否已有伙伴对话带来研究经费?租赁、算力或制造承诺是否形成固定义务?没有这些答案,哪怕是看似基础的结论——例如 Xaira 花钱是否足够激进、能跑赢同行,或是否足够保守、能避免早期融资需求——仍有一部分是推测。本章仍可形成方向性判断,但不是经审计的判断。[CI003, CI022, CI028, CI035, CI036, CI037]
| 缺失指标 | 影响 | 精确尽调路径 |
|---|---|---|
| 当前现金、等价物和有价证券 | 没有该项,资金续航期仍是情景练习,而非资产负债表结论 | 要求提供最新月度现金桥表,以及经审计或董事会审阅的现金状况 |
| 按季度和职能划分的实际烧钱 | 无法有把握建模 Xaira 的烧钱更像 Absci、Relay、Schrödinger,还是更接近 Recursion 规模 | 要求提供季度经营现金流、现金经营费用,以及 AI、数据生成和疗法之间的支出拆分 |
| 具名合作伙伴合同和经济条款 | 阻碍判断平台变现是否已经开始,以及收入质量如何 | 要求提供已执行的条款书或已签协议,并附收入确认和里程碑时间表 |
| 项目级支出和阶段图 | 无法预测内部管线野心会多快吸收资本 | 要求提供当前项目清单、阶段、模态,以及达到下一重大里程碑的预期支出 |
| 债务、租赁、算力承诺和固定义务 | 即使表面现金看起来充足,也可能实质改变资金续航期 | 要求提供租赁时间表、云 / 计算合同、设备融资,以及任何或有义务 |
| 当前商业牵引力指标(客户、合同、使用量、试点) | 缺少商业牵引力证据时,估值只能依赖科学承诺和同业可比,而非商业验证 | 要求提供当前活跃平台评估、付费试点、已签合作伙伴数量,以及截至目前已实现收入 |
这里列出的所有缺口都对估值有实质影响。仅凭公开信息无法可信解决任何一项。
[CI003, CI022, CI028, CI034, CI035, CI036]4.5 财务结论
合理的财务结论是:Xaira 尚未产生收入、资本密集,且大概率仍资金充足,但财务透明度还不足以支撑基本面投资核算。启动轮异常大,很可能买到了真实时间。这个时间重要,因为 Xaira 要构建的是差异化因果生物学栈,并最终形成疗法引擎,而不只是一个软件产品。但让故事有吸引力的同一野心,也让公司变得昂贵。如果管理层在外部货币化出现前更重押内部资产,烧钱可能很快移向同行集合中更临床、更高现金消耗的一端。如果 Xaira 先通过合作伙伴把平台货币化,业务则可能更像一家里程碑和版税型平台公司,现金跑道更长,稀释风险更低。今天的公开证据无法告诉我们哪条路径已经占优。对后续估值而言,Xaira 应被视为一个由平台货币化加内部管线创建组成的、有资金支撑的期权,而不是一家收入质量已验证、利润率可见或现金转化已披露的公司。[CI023, CI025, CI031, CI038, CI039, CI040]
4.6 展示材料
05产品与技术
5.1 Xaira 实际交付什么
公开来看,Xaira 不像一家有外售产品菜单的传统软件公司。官方叙事是一套整合型生物科技平台,把先进 AI 研究、大规模数据生成和疗法产品开发结合起来,每一层都为下一层供料。这意味着公司的“产品”更应被描述为一种运营模型:构建因果生物学数据,用这些数据训练预测模型,用实验验证预测,再把产出转成疗法项目。现有证据最强地支持这种内部平台身份。Xaira 的 2024 年启动材料和当前方法页面都强调三支柱栈。此后,公司公开交付了置于该栈内部的研究产物——X-Atlas/Orion、X-Atlas/Pisces、X-Cell、GitHub 文档和 Hugging Face 卡片——但本报告审阅的来源没有显示公开价格表、企业部署包或具名外部软件客户。放到实际工作流里,今天最可信的用户是 Xaira 自己的科学家、潜在合作者,以及评估部分开放发布的外部研究者。这个区别对尽调重要:这里确有真实技术内容,但它仍更像平台证明,而不是完成商业包装的产品。[CE001, CE002, CE015, CE031, CE037, CE042]
| 模块 / 资产 | 主要用户 | 状态 / 成熟度 | 差异化 | 尽调缺口 |
|---|---|---|---|---|
| 一体化 AI + 数据 + 疗法平台 | Xaira 内部科学家、平台领导层、潜在合作方 | 自启动以来作为公司模式运转;未作为公开 SKU 打包 | 三支柱运营模式统一 AI 研究、数据生成和疗法产品开发 | 无外部定价、合同或具名企业部署的公开证据 |
| X-Atlas/Orion + FiCS Perturb-seq | 功能基因组学、扰动生物学和模型训练团队 | 2025 年以 8M 细胞图谱和 FiCS 方法公开发布 | 工业化扰动数据生成,带有剂量依赖敲低框架和深度测序 | 除预印本和新闻稿主张外,没有公开成本、吞吐量或可复现性 SLO |
| X-Atlas/Pisces | 模型训练团队,以及评估部分公开发布的外部研究人员 | 2026 年公开宣布;范围比 Orion 更广,但对外仍只部分上传 | 覆盖 7 个筛选和 16 个生物学情境的 25.6M 个扰动单细胞转录组 | 数据集卡称上传仍在进行,查看器不可用 |
| X-Cell / X-Cell Mini | 计算生物学家和内部发现团队;发布完成后的外部研究人员 | 文档、代码仓库和模型卡已公开;权重和推理代码仍即将推出 | 基于扩散的虚拟细胞模型,主张具备多模态先验和零样本泛化 | 没有面向外部用户的公开托管端点、基准套件或企业支持界面 |
| 分子设计 / 抗体设计层 | 西雅图分子设计团队和下游疗法项目 | 内部已运行,但公开规格不如 X-Cell 详细 | 建立在 RFdiffusion、ProteinMPNN 等 IPD 根基之上,并结合高通量蛋白测试 | 无具名 Xaira 原创抗体资产或公开生产栈披露 |
| 内部疗法管线生成 | Xaira 发现和开发团队 | 战略目标;下游产出披露仍稀少 | 平台目标是把因果生物学和分子设计转化为差异化疗法 | 公开验证停留在路线图层面,而非具名 Xaira 生成临床资产清单 |
该矩阵捕捉公开来源中可见的层次。它有意把已发布研究资产与仍在内部的项目化层次拆开。
[CE001, CE002, CE003, CE006, CE007, CE015]5.2 数据引擎与 X-Cell 架构
Xaira 技术栈最具体的部分,是因果数据引擎以及建立在其上的虚拟细胞模型。X-Atlas/Orion 引入 FiCS Perturb-seq,把它作为工业化大规模数据生成流程:使用 10x Chromium 平台,声称具备高灵敏度和低批次效应,并产出一个面向全部人类蛋白编码基因、包含 8 million 个细胞的公开图谱。X-Atlas/Pisces 随后把这个基础扩展到 25.6 million 个受扰动单细胞转录组,覆盖七个全基因组 CRISPRi 筛选和 16 个生物学情境。X-Cell 是在这些数据上训练的模型层。公开文档称,它是集合级扩散 Transformer,采用四步迭代优化,通过交叉注意力引入多模态生物学先验,并提供最高 4.9 billion 参数的全尺寸模型家族。公开文档还暴露了超出营销文案的实现细节:X-Cell Mini 被记录为从 scGPT 初始化的 55M 参数变体;计划中的 API 接收 AnnData 或 .h5ad 对照细胞;模型文档披露 Mini 配置至少需要 8 GB GPU 占用。这些都是有意义的技术披露,即便公开交付体验仍不完整。[CE003, CE004, CE006, CE007, CE008, CE009]
| 层级 / 流程 / 组件 | 作用 | 关键依赖 | 证据 / 成熟度 | 风险 |
|---|---|---|---|---|
| FiCS Perturb-seq 湿实验平台 | 工业化生成大规模扰动数据 | 10x Chromium 工作流、湿实验运营、测序深度 | 官方发布和预印本摘要提供支持 | 公开的吞吐量、单位成本和可复现性指标仍然稀少 |
| X-Atlas/Orion 和 X-Atlas/Pisces 数据集层 | 整理干预型单细胞数据,用于训练和验证 | 跨细胞情境的大型实验活动 | 数据集规模和情境多样性证据强;公开文件可获得性不均衡 | 完整文件发布前,社区可复现性受限 |
| X-Cell 模型核心 | 从对照细胞预测扰动响应 | 扩散 Transformer 训练加 GPU 计算 | 模型卡、文档和发布新闻稿有详细说明 | 权重和代码发布前,公开用户无法完整审计运行时行为 |
| 生物学先验整合层 | 借助交叉注意力,把外部知识注入扰动建模 | ESM-2、STRING、GenePT、DepMap、JUMP-Cell Painting、scGPT/gene 向量嵌入 | 模型卡和文档已有公开说明 | 性能对整理后先验的依赖,外部人士可能很难拆开判断 |
| 分子设计层 | 把生物学洞察推进到设计蛋白和抗体 | IPD 衍生模型积累,加实验蛋白验证 | GeekWire 和官方定位在方向上提供较强支持,但文档并不完整 | X-Cell 输出与具名 Xaira 资产之间的直接链接尚未公开 |
| AI 验证反馈回路 | 将实验结果反馈到训练和项目决策 | 计算团队与实验室团队紧密耦合 | 官方方法页和第三方访谈均有明确表述 | 缺少公开吞吐量指标,外部人士无法量化回路速度或效率 |
| 面向开发者的发布界面 | 向外部研究者开放文档、代码骨架、模型卡和数据集卡 | GitHub、原始文档、Hugging Face,以及规划中的软件包 / API | 可见且可信,但仍不完整 | 文档先于完整交付的发布物出现,带来成熟度疑问 |
本表区分公开架构和未经验证的内部实现细节。它标出 Xaira 已展示到足以理解技术栈的部分,以及尚未展示的部分。
[CE004, CE008, CE009, CE010, CE011, CE012]Xaira 公开披露的技术栈从扰动数据生成出发,经虚拟细胞建模,进入分子设计和治疗决策。
该技术栈只纳入公开来源直接支持的组件,不假设未披露的编排或隐藏模型层。
[CE003, CE006, CE008, CE009, CE019, CE021]5.3 运行流程:预测、验证与疗法转化
Xaira 公开展示的是 AI 与湿实验室反馈闭环,不是一次性模型发布。Drug Discovery Trends 引述 Bo Wang 描述了一套系统:AI 给出预测,湿实验验证预测,实验产出再喂给下一轮模型迭代。GeekWire 对 Xaira Seattle 站点的报道让这个闭环更有运营细节:分子设计研究者生成候选蛋白,高通量实验系统测试结合与稳定性,数据再快速回灌模型。Fierce 补上公司层面的商业解读——Xaira 正在建设整合型 R&D 平台,机器学习栈先行,疗法管线随后跟进。也就是说,X-Cell 不是最终产品;它位于大规模扰动数据与下游靶点选择、机制研究、抗体设计并最终进入内部疗法之间。限制在于披露深度。Xaira 和第三方报道让整体工作流可读,但没有进一步给出具体由 Xaira 生成的抗体资产、平台吞吐指标,或该闭环带来的外部验证客户结果。[CE014, CE019, CE020, CE021, CE022, CE023]
| 用户任务 | 当前工作流 | Xaira 方案 | 主张收益 | 限制 |
|---|---|---|---|---|
| 因果靶点发现 | 运行大规模扰动实验,观察基因如何影响细胞状态 | FiCS Perturb-seq + Orion/Pisces 提供基因组规模的干预式训练数据 | 用因果细胞生物学而非纯观察数据改进靶点发现 | 尚无 Xaira 原创靶点提名进入临床的公开证据 |
| 扰动响应预测 | 科学家运行湿实验室 CRISPRi 筛选,再对未见干预建模 | X-Cell 预测未见情境下基因敲低的转录响应 | 可能减少实验负荷,并更快排序后续生物学验证 | 公开 API、权重和推理代码尚未完全交付 |
| 作用机制和患者分层假设生成 | 解读扰动后的通路响应,并把它连接到疾病场景 | Xaira 称 X-Cell 可支持作用机制识别、靶点-患者匹配和毒性预测 | 从细胞状态建模通往转化决策架起桥梁 | 公开来源中的应用仍属前瞻,而非外部验证工作流 |
| 针对难靶点的蛋白 / 抗体设计 | 用生成式设计模型加湿实验室测试,创造结合物和治疗性蛋白 | 西雅图分子设计团队建立在 IPD 衍生模型和高通量验证之上 | 可能打开困难或「不可成药」靶点,并加速迭代 | 确切生产栈、吞吐量和具名 Xaira 设计抗体仍未披露 |
| 内部疗法项目生成 | 把模型和实验洞察转化为自有药物项目 | 管理层称平台优先,管线由平台延伸出来 | 让 Xaira 更像一体化 techbio 建设者,而不是工具供应商 | 公开证据仍主要集中在数据 / 模型层,下游资产披露较弱 |
本表映射的是公司声称的用例,而非已验证的外部客户部署。收益表述严格贴近公开来源可支持的范围。
[CE014, CE019, CE020, CE023, CE024, CE025]Xaira 描述的运营闭环从扰动生成开始,延伸到预测、验证和治疗优先级排序。
图中呈现运营流程,而非商业客户旅程;公开证据首先指向内部科学家工作流。
[CE014, CE019, CE020, CE023, CE024, CE033]5.4 信任、发布成熟度与合规姿态
公开信任姿态真实但有限。Xaira 自 2025 年 1 月 1 日生效的隐私政策称,公司使用技术、组织和行政安全措施,开展欺诈防护与调试,并围绕服务收集分析和 cookie 数据。招聘页面还包含求职诈骗提醒,警告候选人避开非官方招聘渠道。这些都是真实的信任控制,但属于公司网站级控制,不是 X-Cell 或 X-Atlas 的产品级验证。在产品层面,最清楚的公开护栏其实是一个限制:模型卡称 X-Cell 旨在用于计算生物学和基因组学研究,而非临床决策。公开发布成熟度也仍然只是部分完成。GitHub 仓库、文档站点和 Hugging Face 卡片提供了可见开发者界面,但权重和推理代码仍标为“coming soon”;Pisces 数据集卡片称文件仍在准备,数据集查看器不可用。本报告审阅的公开来源没有披露这些产物的 SOC 2、ISO 27001、HIPAA、GxP、21 CFR Part 11、可用性 SLA 或企业支持承诺。对尽调而言,这意味着发布内容足够可信、可以检查,但还不完整,不能作为生产基础设施来投资核算。[CE016, CE017, CE026, CE027, CE028, CE029]
| 控制 / 政策 | 状态 | 范围 | 已验证内容 | 缺口 |
|---|---|---|---|---|
| X-Cell 预期用途声明 | 公开且明确 | 模型卡 / Hugging Face 发布 | X-Cell 被定位为用于计算生物学和基因组学研究 | 未公开声称可用于临床、诊断或受监管部署 |
| 开放发布许可 | 公开且明确 | 模型卡、代码仓库、文档和数据集卡 | 发布物以 CC BY-NC-SA 4.0 / 非商业条款发布 | 许可证说明了访问姿态,但不说明服务可靠性或商业可用性 |
| 网站隐私与安全控制 | 公开但通用 | 公司网站服务和用户数据处理 | 隐私政策提到技术、组织和管理保障,以及欺诈预防 | 不能证明 X-Cell 或 X-Atlas 发布具备产品级控制 |
| 招聘渠道安全 | 公开且明确 | Careers 和求职者互动 | Work-with-us 页面提醒防范冒名、非官方平台和付款骗局 | 对品牌安全有用,但不是软件安全或质量认证 |
| 公开合规认证 / 受监管质量声明 | 未发现 | 产品、基础设施和开发运营 | 已审阅来源未披露这些公开发布物具备 SOC 2、ISO 27001、HIPAA、GxP 或 21 CFR Part 11 声明 | 企业和受监管使用的尽调仍需直接文档 |
信任画像更多由研究用途限制和通用网站控制塑造,而不是成熟企业或受监管产品证据。
[CE016, CE017, CE026, CE027, CE028, CE029]5.5 关键依赖、差异化与路线图
如果 Xaira 的护城河最终被证明耐久,它很可能来自对自有干预数据、模型训练、湿实验验证和疗法转化之间闭环的掌控。官方叙事和第三方报道都指向这一点。Orion 和 Pisces 提供因果数据规模;X-Cell 在其上应用扩散架构和生物学先验;GeekWire 描述了闭环所需的湿实验蛋白测试基础设施;公司领导层继续描述未来将扩展到原代细胞、iPSC 衍生细胞类型、类器官、体内扰动和抗体疗法。这条路线图野心很大,方向上也一致。它同时指出了主要依赖风险所在:与 10x 相关的扰动工作流、大规模实验运营、外部生物学先验、GPU 算力、开发者发布渠道,以及仍披露较少、与 Xaira 的 Institute for Protein Design 根源相连的蛋白质设计层。公开信号显示,平台仍处于建设模式,包括 2026 年 3 月的招聘活动。但今天的公开证明仍前置在数据和模型发布上,而不是具名、由 Xaira 起源的临床或商业产出上。因此,路线图有吸引力,但仍明显跑在公开证据包前面。[CE013, CE018, CE021, CE022, CE023, CE025]
| 日期 / 阶段 | 里程碑 / 发布 | 状态 | 含义 | 来源 |
|---|---|---|---|---|
| 2024-04 发布 | 组建一家整合 AI 研究、数据生成和治疗药物开发的公司 | 已完成 | 在任何重大公开技术发布前,先确立三支柱运营模式和融资基础 | 发布新闻稿 |
| 2025-06 数据平台里程碑 | 宣布 X-Atlas/Orion 和 FiCS Perturb-seq | 已完成 | 展示首个具体公开产出,并验证 Xaira 能够工业化生成扰动数据 | Business Wire、bioRxiv 与 GEN |
| 2026-03 模型里程碑 | 基于 X-Atlas/Pisces 发布 X-Cell | 已完成 | 把数据资产转成可见的模型层,并强化虚拟细胞叙事 | Business Wire + 文档 + GEN |
| 2026-03 公开开发者发布 | GitHub 代码仓库、文档、模型卡、Hugging Face 卡,以及规划中的软件包 / API | 部分完成 | 搭出可检查的开发者界面,但尚未交付可完整运行的公开版本 | GitHub + 原始文档 + Hugging Face |
| 前瞻路线图 | 将 X-Atlas 扩展到原代细胞、iPSC 衍生细胞类型、类器官和体内扰动 | 已规划 | 表明公司希望从细胞状态建模推进到更广的因果生物学覆盖 | 官方 X-Cell 发布 + GEN + BiopharmaTrend |
| 治疗转化路线图 | 用平台生成抗体疗法和内部管线资产 | 推进中,但披露稀少 | 展示平台到产品的野心,但公开证据仍落后于公司表述 | Fierce、Drug Discovery Trends 与 GeekWire |
路线图最强的部分,是 Xaira 已发布数据和模型发布物的部分。下游治疗里程碑仍更偏方向性,公开证据尚不充分。
[CE013, CE016, CE018, CE023, CE034, CE035]Xaira 的公开技术栈依赖实验、计算和分发渠道,三者都必须协同跑通。
依赖关系仅限公开来源直接支持或强烈暗示的关系;不推断披露以外的私有供应商。
[CE004, CE019, CE021, CE032, CE035, CE041]Xaira 各模块的公开成熟度差异很大:数据和文档真实可见,但外部产品化层仍不完整。
评级只基于公开证据作定性判断。低分可能只是披露缺失,不必然意味着技术弱。
[CE016, CE017, CE022, CE029, CE033, CE035]5.6 展示材料
06客户情况
6.1 客户分层:用户先于付费方可见
Xaira 的客户基础最好分成四组,其中只有一组今天有真正有意义的公开证据。第一组是 Xaira 内部团队——发现科学家、计算生物学家和项目组合团队——它们很可能仍是平台的主导用户,因为公司一直把这套技术栈描述为生成疗法的内部引擎。第二组是开放科学共同体;它现在可直接访问 Orion,部分访问 Pisces/X-Cell,并在 Hugging Face 和 GitHub 上拥有可见社区界面。第三组是潜在商业合作者,Xaira 明确表示愿意合作,但没有点名任何对象。第四组是未来的大型制药或生物科技公司对手方;如果 Xaira 通过合作或共同开发把平台货币化,它们很可能才是真正付费买方。缺失的是最重要的商业层:本报告审阅的公开来源没有披露具名付费客户、已签约企业部署,或带有经常性收入的外部账户。因此,公开证据告诉我们谁可能购买、谁已经在实验,但没有告诉我们谁真的在付钱。[CU001, CU002, CU003, CU004, CU005, CU006]
| 分层 | 买方 / 用户 / 付款方 | 用例 | 规模 | 收入 / 战略价值 | 缺口 |
|---|---|---|---|---|---|
| Xaira 内部组合和平台团队 | 买方:执行领导层 / 组合委员会;用户:Xaira 科学家和计算生物学家;付款方: Xaira 资产负债表 | 在内部跑通因果生物学、分子设计和治疗生成闭环 | 可能是当前主导使用群体,但按职能划分的员工数未披露 | 战略价值高,因为产品证明今天最可能在这里生成 | 没有按内部团队披露公开使用率、席位数或工作流指标 |
| 开放科学社区 / 学术研究者 | 买方:未披露;用户:外部计算生物学家和基础模型研究者;付款方:科研经费 / 非商业预算 | 下载 Orion/Pisces 数据,检查 X-Cell,做基准测试或基于发布内容继续开发 | 目前唯一有量化外部采用信号的分层 | 可用于验证、引用、基准测试和发现漏斗顶端协作者,战略价值高;直接收入有限 | 公开来源中,使用这些发布内容的具名机构仍然稀少 |
| 潜在商业合作者 | 买方:CSO / 研发负责人 / BD;用户:发现、转化和计算团队;付款方:生物技术公司或制药公司研发预算 | 探索平台合作、数据访问、靶点发现或联合开发 | Xaira 已明确释放信号,但没有公开具名对手方 | 如果以合作驱动的商业化胜出,可能成为未来最重要收入分层 | 没有公开证据证明已签交易、试点或转化漏斗 |
| 大型制药公司 / 头部生物制药平台买方 | 买方:CSO / 外部创新 / BD 领导层;用户:疾病领域和转化科学家;付款方: 中央研发和合作预算 | 用 Xaira 改善靶点发现、MoA 研究或治疗设计 | 如果出现,可能是低账户数 / 高价值 | 可创造超额收入,但也会立即带来集中度风险 | 目前没有公开具名制药客户、平台交易或伙伴证明 |
| 未来开发者 / 模型用户 | 买方:个人或团队研究人员;用户:数据科学家;付款方:完整发布前不明确 | 通过开放工具运行 X-Cell 或复用 X-Atlas 数据 | 可见的社区界面已经存在,但公开软件包仍是部分交付 | 可把认知度扩展到直接合作买方之外 | 没有公开证据证明付费自助动线、企业层级或支持包 |
本表有意区分已有证据支持的外部用户和潜在商业买方。对 Xaira 而言,用户层可见度最高,付款方层最弱。
[CU001, CU002, CU003, CU004, CU005, CU006]6.2 具名客户证据是科学验证证据,不是收入证据
最强的具名外部验证并不是企业案例,而是 Orion 周围的研究社区。GEN 引述 Human Protein Atlas 联合主任 Emma Lundberg,称这次发布是虚拟细胞领域的重要贡献;Arc Institute 的 Hani Goodarzi 则称其为基础模型的重要训练资源。这些背书有分量,因为两位评论者都是技术上成熟的外部观察者。Hugging Face 给出的类用户证据更强:Orion 讨论页显示,一位真实外部用户 zboldyga 询问 Xaira 用于剂量分层的 sgRNA 计数数据,Xaira 的 Ann Huang 随后给出了具体的 Figshare 文件名。这不是客户背书电话,也不是采购记录,但确实证明外部人在使用数据集,并且 Xaira 提供了后续支持。局限也很清楚:所有证据仍停留在科学社区的使用、讨论和验证里。它不能证明付费合同、生产部署或长期商业韧性。Xaira 公开可见的外部用户和验证者是真实的,但他们仍更接近研究者和评估者,而不是带来收入的客户。[CU007, CU008, CU009, CU010, CU011, CU012]
| 客户 / 证明来源 | 分层 | 部署 / 用例 | 生产与试点 | 结果 / 信号 | 局限 |
|---|---|---|---|---|---|
| 开放科学社区(Orion 数据集下载者) | 学术 / 计算生物学 / 基础模型研究者 | 下载并分析 Orion 开源 perturb-seq 数据集 | 数据集文件进入生产分发,但不是已签约客户部署 | RDWorld 报道发布后两周内下载 >16,451 次 | 下载数没有绑定具名机构、重复使用或商业转化 |
| Hugging Face 上的 zboldyga | 独立外部数据用户 | 想获取 sgRNA 计数数据,以复现按剂量分层的分析 | 活跃外部评估 / 研究使用 | 提出了详细技术问题;Xaira 用精确 Figshare 文件名作答 | 一个具名用户互动有意义,但作为客户样本仍然很小 |
| Emma Lundberg / Human Protein Atlas 联合主任 | 学术 / 科学社区验证者 | 外部评估 Orion 作为稳健虚拟细胞建模资源的价值 | 验证 / 背书,不是产品部署 | 称该发布对社区是重大贡献 | 正向背书不能证明她或其实验室是重复用户或客户 |
| Hani Goodarzi / Arc Institute | 学术 / 基础模型社区验证者 | 外部评估 Orion 作为基础模型训练数据的价值 | 验证 / 背书,不是产品部署 | 称该数据集为整个社区提供了大量资源 | 说明领域相关性,不说明商业耐久性或合同价值 |
本表有意不把科学验证包装成付费客户证明。对 Xaira 而言,这是今天最接近的公开代理指标。
[CU007, CU009, CU010, CU013, CU014, CU015]目前最好的证明是外部科学验证和技术互动,而不是付费客户成熟度。
矩阵仅基于公开证明,对证据质量作定性评分。「低」可能意味着披露缺失,而非实际价值弱。
[CU009, CU013, CU014, CU015, CU016, CU032]6.3 采用轨迹:Orion 领先,X-Cell 跟进
采用轨迹已经看得见,但它主要还是研究分发轨迹,不是收入分发轨迹。Orion 是最强的表面证据,因为它有可衡量的开放数据下载量、Hugging Face 点赞和讨论,以及外部科学评论。X-Cell 和 Pisces 扩大了触达面,但外部证据还没有同样扎实,部分原因是模型和数据集发布仍不完整、时间也更近。Hugging Face 的 Pisces 卡片显示,上月有 80 次下载、6 个点赞,足以证明社区有一定关注,但不足以证明持久采用。Orion 的 Hugging Face 页面也重要,因为它降低了可用性摩擦:Parquet 转换和标准数据工具兼容性,让这个数据集更容易用常见分析工具查询。换句话说,Xaira 目前的外部采用曲线经过开放数据发现、技术评估和基准测试,还没有进入企业部署或签约合作的公开证明。实际结论是,Xaira 的采用证据新且有分量,但集中在科学社区,商业化漏斗仍处早期。[CU007, CU008, CU009, CU010, CU011, CU012]
| 指标 | 数值 | 日期 | 来源 | 置信度 | 含义 / 缺失分母 |
|---|---|---|---|---|---|
| Orion 开放数据下载 | 发布后 ~2 周内 >16,451 次下载 | 2025-06-26 | RDWorld | 中 | 最强的量化外部采用信号;未识别唯一机构、重复用户或商业转化 |
| Hugging Face Orion 页面互动 | 22 个点赞和 2 个公开讨论 | 截至 2026-05-12 观察到 | Hugging Face 讨论索引 | 中 | 显示活跃社区兴趣,但不代表留存或变现 |
| 具名外部支持互动 | 1 个已披露外部用户问题,并以精确文件名作答 | 2025-10-24 至 2025-11-12 | Hugging Face 讨论 #2 | 中 | 至少证明一个真实外部用户工作流;样本量极小 |
| Orion 可用性增强 | 已发布 Parquet 转换分支 | 2025-?? | Hugging Face 讨论 #1 | 中 | 改善外部可查询性;但几乎不说明活跃用户数量 |
| Pisces 公开牵引力 | 上月 80 次下载;6 个点赞 | 截至 2026-05-12 观察到 | Hugging Face Pisces 卡 | 中低 | 显示小于 Orion 但真实的后续兴趣;无机构拆分或重复使用数据 |
| 具名外部验证者 | 2 位被引用的外部专家(Emma Lundberg、Hani Goodarzi) | 2025-06-17 | GEN Edge | 中 | 可证明社区相关性;不是部署或付费证明 |
| 具名付费客户 | 0 个披露 | 截至 2026-05-12 | 根据已审阅来源推断 | 中 | 核心商业化分母仍然缺失 |
轨迹表只使用直接公开代理指标。客户数、部署数和收入数字未披露时,本表有意保持空值。
[CU007, CU008, CU009, CU010, CU011, CU012]Xaira 的可见旅程从开放科学认知开始,进入技术评估,并且只是可能走向未来商业合作。
旅程图反映公开证据模式:开放数据和社区讨论先于任何已披露商业客户证明。
[CU003, CU024, CU025, CU031, CU037]当前公开漏斗以科学社区为先:发布、评估和讨论可见,货币化部署不可见。
由于大多数转化数量未披露,这里用流程图而非量化漏斗。只有 Orion 下载量等部分公开信号已知。
[CU007, CU008, CU009, CU011, CU012, CU039]6.4 留存、扩张和集中度仍基本是空白
客户持续性是公开证据最薄弱的地方。没有任何已审阅公开来源披露 NRR、GRR、logo 流失、续约率、合同期限、账户扩张或满意度调查。仅有的重复使用代理指标都是间接的:Orion 发布数月后仍有公开讨论,Hugging Face 页面仍有点赞,Pisces 也有较小但持续的下载信号。这些信号说明关注在延续,但不是商业意义上的留存指标。扩张逻辑也仍偏前瞻。RDWorld 引述 Xaira 称,数据集对学术界免费,同时公司乐于与有合作兴趣的商业实体合作,这意味着双轨模式:先扩大开放科学触达,再靠合作变现。这条路可能走得通,但若成功,收入基础很可能集中,少数大型 制药或 techbio 对手方的重要性会高于成千上万的自助用户。采购摩擦也高:公开发布仍不完整,企业支持条款不可见,产品级合规证据缺位。合在一起,今天的留存故事应按空白处理,扩张故事则更像一个期权,而非已经验证的动作。[CU019, CU020, CU021, CU022, CU023, CU024]
| 指标 | 数值 / 空值 | 分层 | 置信度 | 尽调要求 |
|---|---|---|---|---|
| NRR / GRR | 空值 — 未披露 | 商业客户 | 高(公开值缺失) | 一旦存在任何合作或软件合同,要求提供队列收入留存 |
| 客户流失 / 续约 | 空值 — 未披露 | 所有外部账户 | 高(公开值缺失) | 要求提供任何商业或试点用户的续约率、合同期限和账户级状态 |
| 重复使用代理指标 | 部分 — 持续点赞 / 讨论和后续 Pisces 活动 | 开放科学社区 | 中低 | 要求按发布版本提供实际重复下载、引用或回访用户指标 |
| 支持响应代理指标 | 部分 — Xaira 回答了 1 个公开技术问题 | 外部数据集用户 | 中低 | 要求提供面向公开用户或合作伙伴的中位响应时间、问题单数量和支持流程文档 |
| 满意度 / 推荐质量 | 部分 — 正向专家引用,但没有正式评分 | 科学验证者 | 中低 | 要求提供案例研究、独立基准测试或具名合作者推荐,说明结果和重复使用 |
本表遵循一个原则:无支持的留存说法应变成空值或低置信度代理指标,而不是编造事实。
[CU019, CU020, CU021, CU022, CU023, CU034]| 扩张驱动 | 集中度风险 | 影响 | 尽调路径 |
|---|---|---|---|
| 开放数据集和模型可见度 | 社区触达可能无法转化为付费合作 | 高 — 强认知度若没有变现,客户质量仍无解 | 跟踪合作咨询、引用到伙伴的转化,以及任何可归因于 Orion/X-Cell 的入站商业线索 |
| 商业合作渠道 | 如果商业化出现,收入很可能来自少数大型对手方 | 高 — 立即带来收入集中和谈判杠杆风险 | 要求提供商业洽谈管线、目标账户清单,以及 1-3 个锚定交易的情景分析 |
| 制药 / techbio 买方分层 | 大票单价意味着账户数低 | 高 — 失去一个买方就可能实质改变收入前景 | 要求提供买方分层、头部账户敞口假设和条款清单历史 |
| X-Cell 类工具的企业采用 | 部分交付、支持证明有限、无公开合规栈,会抬高采购摩擦 | 中高 — 从好奇心转向生产部署会变慢 | 要求提供企业路线图、支持模式、安全 / 合规材料和部署参考 |
| 学术 / 开放科学验证者分层 | 战略上有用,但直接收入低 | 中 — 可带来影响力和引用,但不必然带来持久变现 | 要求提供开放发布带来的引用、基准测试提及和合作者介绍数据 |
对 Xaira 而言,扩张和集中度风险不可分割:最可能的变现路径,也最可能导向高度集中的客户群。
[CU024, CU025, CU026, CU027, CU028, CU029]Xaira 公开社区界面的重复使用代理队列示例。这些是分析师代理情景,不是公司披露的留存数据。
Xaira 未披露真实留存指标。这些百分比是低置信度代理情景,仅用于尽调框架,把观察到的早期社区信号转成重复使用假设。
[CU019, CU020, CU021, CU034]6.5 客户结论
正确的客户结论是:Xaira 在开放科学生态里有可信的早期外部用户证据,但还没有公开可读的商业客户基础。Orion 已经切入研究社区,下载量实在、用户问题可见,也有受尊重的外部验证者。这比只有 logo 的证明强很多。但它仍不等于付费账户基础、留存证据或多元收入。如果 Xaira 成功变现,未来客户模型大概率指向少数高价值合作,这会让集中度风险成为核心尽调问题。在公司披露具名商业用户、合作合同或续约证据之前,投资者不应把科学牵引力误当商业牵引力。因此,带入后续章节的客户结论很直接:Xaira 有市场兴趣和用户好奇心,但尚无公开证据证明其已经变现且形成持久客户采用。[CU016, CU025, CU026, CU031, CU032, CU035]
6.6 展品
07风险
7.1 监管与法律风险
Xaira 仍足够早,因此最大的监管问题不是已知执法行动,而是监管预期上升与公开准备证据稀疏之间的缺口。FDA 和 EMA 现在把药物生命周期中的 AI 定义为一套需要风险管理、文档留痕、并按使用场景界定的纪律,而不是不受约束的研究活动。EMA 反思文件明确表示,药品生命周期中的 AI 会引入新风险,必须缓释这些风险,以保护患者和临床证据完整性;FDA 2025 年指南草案则要求,当 AI 支持监管决策时,申办方要建立基于风险的可信度评估框架。EU AI Act 和 GDPR 又在欧洲增加了围绕健康、安全、基本权利和个人数据处理的约束。 Xaira 的公开材料不能证明它已经把 AI 输出用于监管申报,但确实显示,公司希望端到端连接模型、生物学和患者。因此,监管严谨度只是时间问题,不是相关性问题。当前公开合规表面很薄:网站隐私政策存在,但没有已审阅的公开 DPA、GxP 包、验证档案,也没有可供外部尽调的 AI 治理披露。法律条款同样重要。X-Cell 的公开发布使用 CC BY-NC-SA 4.0 许可,这有利于研究分发,但在没有单独权利的情况下会限制商业复用。因此,实际风险有两层:Xaira 的推进速度可能快于其公开合规叙事;其公开法律包装也可能尚未达到未来企业或受监管对手方的预期。[CR001, CR002, CR003, CR004, CR005, CR006]
| 规则 / 问题 | 司法辖区 / 范围 | 当前状态 | 可能性 | 严重性 | 缓解措施 | 剩余暴露 | 尽调路径 |
|---|---|---|---|---|---|---|---|
| AI 生命周期治理(FDA / EMA) | 药物和生物制品开发 | 已有有效指南和原则;监管预期在收紧 | 高 | 高 | 内部科学和监管人才强;仍处早期,有时间准备 | 如果 Xaira 的 AI 输出开始影响监管证据,文档不足可能拖慢项目或合作方尽调 | 要求提供内部 AI 治理框架、验证标准,以及任何 FDA / EMA 互动备忘录 |
| EU AI Act + GDPR 重叠 | 欧洲联盟 | 已生效;带来健康、安全、权利和数据保护义务 | 中 | 高 | 可把面向欧盟的部署排在后面,并使用经验丰富的法律顾问 | 若 Xaira 缺少可用于欧盟的文档和数据处理姿态,欧洲扩张或受监管合作可能卡住 | 要求提供欧盟就绪度评估、GDPR 法律基础分析和数据处理模板 |
| X-Cell 非商业许可 | 全球外部复用 | 公开发布采用 CC BY-NC-SA 4.0 条款 | 高 | 中 | 单独商业协议可以覆盖公开许可的限制 | 研究采用度会提高,但没有定制条款时,商业嵌入或合作方再分发在法律上可能变得别扭 | 要求提供 X-Cell 商业许可策略,以及下游数据 / 模型访问安排 |
| 公开合规材料缺口 | 外部尽调界面 | 有隐私政策;产品专属合规包未公开 | 高 | 高 | 私有尽调资料室最终可以补齐缺口 | 在解决前,企业买家和后期合作方可能把缺口本身视为红旗 | 要求提供 DPA、安全问卷、GxP 定位,以及任何验证或质量体系材料 |
| 敏感数据 / 受监管用途扩张 | 未来临床和转化工作流 | 当 Xaira 更接近患者关联或提交相关用途时,该风险会变得重要 | 中 | 高 | 可把初始用例限制在研究场景,并让人类保留监督环节 | 如果 Xaira 把模型接到患者或申报场景,却没有同样成熟的治理和文档,风险会升高 | 要求管理层界定产品、数据和 AI 治理升级到申报级的精确触发点 |
严重性和可能性是基于公开材料的分析师判断;私有合规或法律文档可能降低或提高剩余风险敞口。
[CR001, CR002, CR004, CR005, CR008, CR009]7.2 科学、运营与安全风险
Xaira 最难的风险仍是科学转化。Orion、Pisces 和 X-Cell 拼出了一套可信的数据与模型叙事,但外部评论对虚拟细胞模型能在多大程度上泛化到患者结局仍然谨慎。这很关键,因为投资逻辑不只是 Xaira 能生成漂亮的生物表征,而是这些表征能改善靶点选择、疗法设计,并最终提高临床成功率。公开 X-Cell 材料也仍不完整:Xaira 的 Hugging Face 和 GitHub 页面仍显示模型权重和推理代码即将发布。这限制了可复现性,拖慢第三方基准测试,也让外部产品面更像一项推进中的研究发布,而不是成品平台。 运营上,Xaira 的护城河取决于能否把大规模单细胞数据生成变成可重复引擎。Orion 发布把这台引擎与 10x Genomics 的 Chromium 平台相连;即使它不是唯一依赖,10x 也是重要的上游工作流依赖。安全性和可靠性在公开层面同样记录不足。Xaira 的隐私政策称公司使用适当措施,也承认任何传输或存储方式都不可能完全安全,但没有已审阅公开证据显示其具备 SOC 2、ISO 27001、正式正常运行时间承诺、灾备细节或受监管数据控制。NIST 和 CISA 指南已经明确,AI 部署越来越要承担生命周期治理和安全内建预期。因此,残余运营风险不只是宕机或网络事件;更在于 Xaira 可能先迎来买方或监管审查,而其公开控制面还没有成熟到能高效通过尽调。[CR011, CR012, CR013, CR014, CR015, CR016]
| 失效模式 | 可能性 | 严重性 | 缓释成熟度 | 剩余风险敞口 | 未决缺口 |
|---|---|---|---|---|---|
| 虚拟细胞输出未能转化为有治疗价值的患者结局 | 中高 | 严重 | 中 — 团队和数据资产世界级,但公开证据仍处早期 | 平台看起来科学上很亮眼,却没有产出差异化药物成果 | 没有公开披露 Orion / X-Cell 到管线或临床里程碑的转化指标 |
| 大规模数据生成引擎表现不及预期或成为瓶颈 | 中 | 高 | 中 — Xaira 高度重视数据且近期有发布,但工作流复杂度仍高 | 数据新鲜度下降、模型改进放慢,护城河变弱 | 没有公开披露内部数据引擎的吞吐、成本、失败率或可复现性 |
| 10x Chromium 依赖扰动数据集生成或扩展 | 中低 | 中高 | 中 — 依赖关系明确,但没有证据表明它永远是唯一技术路径 | 如果供给、成本或性能变化,Xaira 核心数据管线可能放慢 | 未公开披露应急方案或替代平台策略 |
| 公开交付不完整(权重 / 推理仍即将推出)拖慢可复现性和尽调 | 高 | 中高 | 中低 — 活跃的公开代码仓库和文档有帮助,但交付不完整仍在 | 外部用户无法完整基准测试或部署 Xaira 公开营销的内容 | 未公开承诺外部产品界面完整交付日期 |
| 安全 / 合规成熟度落后于买家预期 | 中高 | 高 | 低 — 有隐私表述,但未看到经审查的公开审计或可用性材料 | 企业或受监管买家可能在安全审查处止步,技术评估还没开始就被挡住 | 未发现公开 SOC 2、ISO 27001、渗透测试、灾备或 SLA 证据 |
各行结合了直接公开观察和推断的失效模式;Xaira 未公开披露事故历史、可用性指标或内部吞吐数据。
[CR011, CR012, CR013, CR014, CR015, CR016]7.3 商业、合作伙伴与依赖风险
Xaira 现在有可见的外部兴趣,但公司仍未跨过公开可读商业证明这条线。Orion 下载量、Hugging Face 互动和具名科学验证者,证明开放科学社区正在关注。它们不证明合同、续约或部署经济性。最可能的变现路径看起来仍是与生物技术或制药买方合作,而不是广泛自助式软件业务。这种模型可以成立,但一旦成立,通常会很快带来收入集中:少数高价值对手方比长尾小用户更重要。 Xaira 现有对外触达面还存在结构性依赖取舍。在 GitHub 和 Hugging Face 上开放发布,有利于认知、技术验证和社区触达,但不能替代 Xaira 自控的企业交付、支持、安全审查或可审计性。同样的发布也制造模仿风险,因为竞争对手和潜在合作方可以在 Xaira 公开证明更优疗法或合作成果之前,先研究其公开数据和模型表面。由此产生的商业风险是不对称的。科学牵引力会让 Xaira 显得更真实,但在它转化为具名合作伙伴、试点或带管线的交易之前,投资者应把客户层视为未验证且很可能集中。这意味着合作伙伴依赖不只是某一个供应商的问题;它关乎一个仍然私有的平台,能否在路上不免费泄露太多价值的情况下,把研究相关性转成少数高价值、成败攸关的商业关系。[CR022, CR023, CR024, CR025, CR026, CR027]
| 依赖 / 风险 | 交易对手 / 细分市场 | 角色 | 集中度 | 失效情景 | 严重性 | 缓释措施 | 剩余风险敞口 |
|---|---|---|---|---|---|---|---|
| 商业证明缺口 | 潜在生物技术 / 制药合作方 | 将验证变现能力和战略相关性 | 高 — 尚无公开付费客户多元化证据 | 科学牵引未能转化为合同或试点 | 高 | 开放科学牵引带来一定漏斗顶部可信度 | 没有公开具名客户、部署或续约证据 |
| 合作集中度 | 少数大型交易对手 | 若变现跑通,可能是早期收入路径 | 高 | 一次谈判流失会实质改变收入前景 | 高 | 大额合作仍可产生有吸引力的经济性 | 收入质量可能仍然波动大、由合作方驱动 |
| 公开分发平台 | GitHub / Hugging Face | 社区触达、开发者参与和发布托管 | 中 | 如果第三方界面变化或限流,外部触达或支持工作流会断裂 | 中 | 公开界面易于采用,且已经活跃 | 这些不能替代面向具体买家的交付、支持或合规控制 |
| 开放发布带来的模仿 | 竞争对手和潜在合作方 | 可以检查公开数据 / 模型界面 | 中高 | 同业从发布中学习的速度快于 Xaira 证明差异化经济性的速度 | 中高 | 研究领导力和内部私有数据仍可能保住优势 | 在变现被证明前,公开发布会侵蚀信息不对称 |
| 买家采购摩擦 | 企业生物技术 / 制药 / 受监管交易对手 | 部署前需要对安全、法律和支持有信心 | 高 | 技术好奇心始终没有变成获批采购流程 | 高 | 领导层信誉可能有助于打开大门 | 若缺少更强合规 / 支持证据,高端买家可能卡在尽调 |
该登记表区分了可见的研究社区牵引和仍未披露的商业证明;缺失买家数据应被视为真实尽调缺口,而不是零风险。
[CR022, CR023, CR024, CR025, CR026, CR027]图中梳理公开材料中最可见的依赖:上游数据生成伙伴和方法、外部发布平台、领导层集中度,以及最终通向合作者、监管机构和治疗项目的路径。
[CR013, CR015, CR027, CR028, CR031, CR032]7.4 人员、治理与融资风险
Xaira 的人才画像既是最强资产之一,也是最清晰的集中风险之一。很少有初创公司一出场就拥有这样深的董事会和领导层:Marc Tessier-Lavigne、David Baker、Bo Wang、Hetu Kamisetty、Debbie Law、Paulo Fontoura,以及董事 Scott Gottlieb,共同给 Xaira 带来少见的科学、技术、临床和监管可信度。但这并不消除集中度。公开材料仍显示,AI、生物学、临床开发和公司建设这几条线,承重过多地压在相对少数的高级领导者身上。公司在 2024 年末和 2025 年仍在补齐高管团队,公开招聘信号也显示 2026 年仍有不少开放职位。 因此,融资风险更适合理解为证明负担风险,而不是短期现金跑道风险。超过 $1 billion 的承诺资本是强缓冲。但 Xaira 要同时跑资本密集型的 AI 研究、湿实验数据生成和疗法开发,而 2026 年生物制药 融资环境仍偏好临床和商业路径更清晰的公司。如果在下一次融资事件变得重要之前,Xaira 不能拿出足够证据证明管线转化、合作方转化或平台成熟度已去风险,市场会要求更硬的证明,不管种子资金池有多大。这里最强的缓释因素是团队和投资方组合的质量。残余风险是,即使输入条件精英化,也未必能压缩打造可重复药物发现和商业化引擎所需的时间。[CR031, CR032, CR033, CR034, CR035, CR036]
| 角色 / 职能 | 依赖或缺口 | 可能性 | 严重性 | 缓释措施 | 尽调路径 |
|---|---|---|---|---|---|
| 顶尖科学和技术领导层 | Marc Tessier-Lavigne、Bo Wang、Hetu Kamisetty、Debbie Law、Paulo Fontoura、David Baker 承载了公司相当大一部分公信力和技术诀窍 | 中 | 高 | 董事会厚度和近期新增高管拓宽了管理梯队 | 要求提供继任计划、留任方案,以及 AI、生物和临床职能之间清晰的汇报线 |
| 组织扩张 | 公司仍在积极招聘,近期也在补齐多个关键 C-suite 角色 | 高 | 中高 | 大额融资让公司有时间谨慎招聘 | 要求按职能提供组织架构图、未填补关键岗位和招聘周期指标 |
| 董事会 / 治理相对于运营证明 | 声望高的董事会能打开大门,但不能替代执行证据 | 中 | 中 | 董事会包含监管和大型药企经验 | 要求提供董事会运行节奏、项目评审流程,以及围绕模型 / 组合取舍的治理安排 |
| 资本强度 | AI 研究 + 湿实验室数据生成 + 疗法开发天然烧钱 | 高 | 高 | 超过 $1B 启动资本是很强的初始缓冲 | 要求按职能披露现金消耗、资金续航情景,以及到达首个决定性价值拐点所需支出 |
| 下一轮证明负担 | 2026 年融资环境对更清晰临床或商业证明仍然挑剔 | 中 | 高 | 强大的投资财团和战略兴趣改善融资通道 | 如果证明点滞后,即便资金充足,公司也可能面临定价压力或战略漂移 |
人员和融资各行只依赖公开领导层、招聘和市场信号;私有现金消耗、留任和投资者权利细节可能实质改变严重性。
[CR031, CR032, CR033, CR034, CR035, CR036]7.5 缓释因素与推翻论点的触发器
Xaira 确实有真实缓释因素。$1 billion 起步资本实质性降低短期融资压力。领导团队和董事会比典型临床前初创公司深得多。Orion 及相关公开发布形成了科学社区验证循环,让公司比纯隐身 techbio 拥有更多外部证明。这些优势意味着,Xaira 不需要立刻做到完美,仍可保持可融资性和战略吸引力。 但残余风险集中在三处:转化为疗法、转化为具商业意义的对手方、成熟为可供企业或受监管场景尽调的平台。这些阈值上,公开证明仍最薄。投资者应把关键监测项设得具体且有时间约束:公开产品面完整交付;具名商业或战略伙伴出现;可信安全 / 合规包公开或私下出具;平台输出开始影响治疗资产;高级领导团队保持稳定。如果这些信号没有改善,Xaira 当前强项会反过来成为压力,因为外界预期已经很高。正确的风险姿态不是认为 Xaira 今天很脆弱,而是承认 Xaira 已经把自己的证明门槛抬高。后续估值工作因此应更多折价于尚未解决的执行和转化风险,而不是泛泛的创业公司稀缺性或融资风险。[CR039, CR040, CR041, CR042]
| 风险 | 可监控触发点 | 阈值 / 事件 | 行动含义 |
|---|---|---|---|
| 监管 / 合规就绪度 | 是否具备 AI 治理、验证、DPA 和质量材料 | 当 Xaira 寻求企业部署或提交相邻用例时,如果这些材料仍缺失 | 在承销更快商业化或临床相邻主张前,要求管理层解释闸门计划 |
| 公开产品完整交付 | 权重、推理代码和可复现文档普遍可用 | 如果到下一个重大尽调周期,X-Cell 仍然只是部分交付 | 折价看待产品成熟度和外部开发者采用主张 |
| 商业合作方转化 | 具名合作、试点或战略买家证据 | 如果持续科学宣传后仍没有可信交易对手证据出现 | 将开放科学牵引视为营销,而非变现证据 |
| 疗法转化 | 有证据显示平台输出影响实际资产、靶点决策或受监管里程碑 | 如果转化仍停留在叙事而非可衡量结果 | 在估值和情景分析中提高执行风险折价 |
| 关键人物稳定性 | 领导层离职或角色不稳定 | 一名或多名核心科学 / 技术 / 临床领导离开 | 立即重新评估运营连续性、招聘难度和知识转移风险 |
| 安全 / 采购成熟度 | 企业尽调包、审计或可靠性承诺 | 如果大型买家接触后,Xaira 仍无法满足基础尽调要求 | 假设企业销售周期更长、转化概率更低 |
这些触发点用于投资者监控,而非概率估计;每一项在实践中被视为推翻投资论点前,都应配合直接管理层尽调。
[CR039, CR040, CR041, CR042]Xaira 主要风险的定性热力图。右上角风险同时具备高可能性和高影响或关键影响;左下角项目是监控风险或二阶暴露。
可能性和影响位置来自分析师基于公开证据的判断。Xaira 未公开披露精算风险数据、事件历史或内部控制指标。
[CR005, CR018, CR020, CR024, CR025, CR032]DAG 将核心风险来源连接到商业化、融资和估值的下游后果。最关键的传导路径集中在治疗转化、商业转化和合规成熟度。
[CR018, CR024, CR029, CR037, CR040, CR041]7.6 展品
08估值
8.1 建议与入场纪律
第一个估值问题不是 Xaira 是否有意思,而是公开证据是否足够为它定价。Xaira 显然有不寻常的起步优势:超过 $1 billion 承诺资本,AI 和生物技术人才高度密集,围绕 Orion 和 X-Cell 的开放科学牵引力也已经可见。但这些优势不等于一个已经可定价的投资案例。公开来源没有披露 Xaira 的投后估值、每股价格、稀释机制或优先权结构。因此,在具体入场价格上给出硬性的「买入」或「回避」判断,会是假精确。 更诚实的建议是继续研究、但对价格敏感。目前,AI 赋能药物发现及相邻临床前疗法公司的公开可比公司,市值大致在 $0.9B-$2.5B 区间。仅凭团队和资本,Xaira 或许应较其中几家公司享有溢价,但公开证据还不足以把它视作已被验证的前沿平台。护栏很简单:如果私募定价接近公开可比公司 集群再加上理性溢价,就继续尽调;如果定价已经隐含大规模商业或疗法证明,公开证据提示应暂停。这不是否定 Xaira 的潜力,而是承认投资者被要求为可能性定价,而不是为已测量的经济性定价。在价格、条款和证明更清楚之前,估值纪律比叙事热情更重要。[CV001, CV002, CV003, CV004, CV005, CV042]
| 维度 | 评估 | 置信度 | 决策含义 |
|---|---|---|---|
| 总体建议 | 继续研究 | 中 | 不要仅凭公开证据为未披露的溢价估值背书;只有价格和条款明确后才继续。 |
| 风险评级 | 高 | 高 | 临床前科学、商业化不透明和融资条款不确定,共同构成高风险画像。 |
| 估值立场 | 对价格敏感;公开证据支撑的参考区间低于前沿 AI 软件叙事 | 中 | 相对公开 techbio 可比公司给溢价可能合理,但公开证据不足以支撑盲目的两位数十亿美元级溢价。 |
| 入场纪律 | 仅在接近可比公司集群加溢价的定价时推进 | 中 | 如果私有定价隐含的证明点公开来源看不到,就等待;如果条款更接近证据支撑区间,则继续尽调。 |
| 对公开证据的信心 | 科学中等,经济性和条款较弱 | 中 | 足以设定护栏,不足以为实际轮次定价。 |
该摘要明确对价格敏感,因为 Xaira 的实际融资条款不公开;拿到投资条款清单后,应重新推演所有决策含义。
[CV001, CV002, CV004, CV005, CV038, CV039]流程图把品类增长、Xaira 优势、缺失的定价和证明数据,以及最终「继续研究」建议连接起来。
[CV001, CV002, CV004, CV006, CV008, CV043]8.2 投资论点与反论点
牛市情景从真正稀缺的输入条件开始。Xaira 结合了很大的起步资本基础,一支围绕 AI、生物学和临床领导力搭建的团队,以及贯穿模型开发、数据生成和疗法开发的一体化架构。Orion 和 X-Cell 也让公司比典型隐身 techbio 初创公司拥有更多公开科学证明。在许多公司仍要求投资者相信路演材料的市场里,Xaira 至少已经展示了真实数据和模型资产。终端市场背景也有利:AI 药物发现仍是增长中的类别,更广泛的生物制药行业仍需要提高生产率。 反论点是,市场最终为证明付费,而不是为输入付费。开放科学牵引力不等于具名商业牵引力。公开来源仍看不到定价、客户、合作经济性,也看不到资产级别转化为患者结局。行业本身仍拥挤,差异化很难。Biomed Nexus 很好地抓住了核心问题:许多 AI 药物发现公司仍是尚未产生收入的平台,真正的验证周期在临床。Xaira 因此或许应较普通临床前公司享有有意义的溢价,但溢价必须受缺失项约束。投资问题不是 Xaira 是否令人印象深刻,而是现有公开记录是否足以支撑新资金可能被要求支付的溢价。现在,答案只有部分肯定。[CV006, CV007, CV008, CV009, CV010, CV011]
| 论据 | 支撑 | 什么会改变观点 |
|---|---|---|
| 论点:Xaira 作为 AI 药物发现公司,起步投入异常强 | >$1B 资本、精英团队密度、一体化 AI / 数据 / 疗法架构,以及公开 Orion / X-Cell 成果 | 如果这些投入没有开始转化为交易对手、资产或可衡量验证,则下调判断 |
| 论点:开放科学发布比纯隐身模式更能建立可信度 | Orion 下载、讨论和公开技术界面让 Xaira 比黑箱创业公司更容易尽调 | 如果科学关注转化为具名战略或商业关系,则上调判断 |
| 论点:市场背景支持溢价 techbio 估值 | AI 药物发现仍是增长品类,biopharma 仍需要提高生产率 | 只有 Xaira 展示真实平台复利、而非品类热度时,溢价才会扩大 |
| 反论点:没有公开价格或条款结构可供承销 | 留存来源中没有投后估值、股价、优先权结构或稀释细节 | 用投资条款清单、股权结构表和清算瀑布解决 |
| 反论点:商业证明仍缺失 | 没有公开具名付费客户、定价或部署指标 | 具名合作方、定价或带里程碑的合作证据出现后上调判断 |
| 反论点:转化仍未解决 | 公开证据仍未达到患者结局或资产级证明,外部评论也保持谨慎 | 出现与 Xaira 平台相连的可衡量资产选择、靶点验证或受监管里程碑证明后上调判断 |
该表把公司质量和可投资性分开。Xaira 可以在战略上很亮眼,但如果价格不对,支撑证据仍不足。
[CV006, CV007, CV008, CV009, CV010, CV011]8.3 可比公司组合与估值语境
Xaira 最好的公开锚点不是纯软件 AI 实验室,而是试图把计算驱动药物发现或精准疗法变现的上市和私营公司。Recursion 是最接近的全栈上市参照,因为它结合了平台、管线和合作;但它的市值仍约 $1.73B。Relay 显示,一个资本充足、尚未产生收入的疗法平台,在公开市场大约可以拿到 $2.46B。Schrödinger 则说明,即使成熟的计算平台拥有合作和自有药物项目,也可能在 $1B 以下交易。Absci 显示,拥有内部和合作项目的生成式 AI 生物制剂平台,也可能仍停留在 sub-$1B 区间。 把 Xaira 大幅推高到这些公开标尺之上,主要理由是私募市场有时会为可选性付费,尤其是在人才、资本和战略稀缺性异常突出时。Isomorphic 的 $600M 外部融资,是近期最清晰的信号,说明私募资本仍在给前沿 AI 药物设计叙事定价。但这个信号也不完整,因为估值没有披露,而且 Isomorphic 受益于 DeepMind 和 Alphabet 血统,不能直接迁移到 Xaira 身上。因此,可比组合同时做两件事:它用公开市场对更成熟证明的定价防止高价买入,也保留空间——如果私下尽调发现强于公开来源的商业或疗法牵引力,Xaira 可以享有溢价。这就是为什么 Xaira 应被估作高溢价 techbio,而不是不受约束的前沿 AI 软件故事。[CV013, CV014, CV015, CV016, CV017, CV018]
| 可比公司 | 指标 | 估值 / 状态 | 相关性 | 局限 |
|---|---|---|---|---|
| Recursion Pharmaceuticals | 2026 年 5 月市值;公开 AI 原生平台 + 管线 + 合作 | $1.73B 市值;10-K 显示 $753.9M 现金且无产品收入 | 最接近的公开全栈 AI 生物技术参照,覆盖平台 + 管线 + 合作方经济性 | 比 Xaira 更成熟且已经上市;仍不是干净的私募轮定价可比 |
| Relay Therapeutics | 2026 年 5 月市值;尚未有收入的公开疗法平台 | $2.46B 市值;10-K 显示 $554.5M 现金,资金续航到 2029 年 | 显示资本充足、尚未有收入的疗法平台在公开市场可获得的估值 | 并非 AI 优先,因此低估前沿 AI 叙事溢价 |
| Schrödinger | 2026 年 5 月市值;计算平台 + 疗法模式 | $0.95B 市值;公开软件 + 药物发现混合体 | 适合作为平台变现和混合商业模式参照 | 经济性不同,因为 Schrödinger 已经变现软件,运营历史也长得多 |
| Absci | 2026 年 5 月市值;生成式 AI 生物制品平台 | $0.90B 市值;10-K 显示 $2.8M 2025 年收入和 $115.2M 净亏损 | 比多数通用 AI 可比公司更接近的生物制品 / AI 平台同业 | 资本基础更小,管线成熟度也不同于 Xaira |
| Isomorphic Labs | 2025 年私募融资信号 | $600M 外部融资轮;估值未披露 | 证明私募资本仍愿为前沿 AI 药物设计可选性付费 | 估值未披露,DeepMind / Alphabet 背景让它更像愿景型参照,而不是干净可比 |
该表混合了公开市值和私募融资信号,因为 Xaira 本身是私有公司且没有公开定价。公开可比公司锚定下行;私有可比信号提示溢价潜力。
[CV013, CV014, CV015, CV016, CV017, CV018]条形图对比当前公开 AI-biotech 市值和 Xaira 在不同证明状态下的参考值。
[CV001, CV013, CV015, CV016, CV017, CV021]8.4 牛市 / 基准 / 熊市情景区间
由于实际入场价格未公开,最干净的估值表达方式不是 IRR,而是有证据支撑的参考区间。熊市情景中,Xaira 保持科学可信度,但仍无法展示具名合作、清晰商业转化或资产级转化。在那个世界里,估值会回落到公开可比公司集群:约 $1.5B-$2.5B。基准情景中,Xaira 把科学可信度转化为至少一个重要合作或明确内部资产证明点,并保住其资本基础和团队密度优势。这支撑更接近 $3B-$5B 的区间。牛市情景中,Xaira 足够快地拿出多个证明点,使私募投资者即使面对更低的公开市场标尺,仍继续支付前沿可选性溢价。估值可到约 $6B-$9B。 推动估值在这些区间之间移动的,不是抽象的市场情绪本身,而是具体证明的到来。从熊到基准至少需要一个有意义的转化信号——真实对手方、合作关系或可衡量的资产里程碑。从基准到牛市需要重复性:不止一个证明点,证据显示平台确实能复利,以及私募市场继续偏好前沿 AI 生物学。估值若远高于牛市区间,就需要强到足以说明现有公开可比组合已不是合适决策工具的非公开证据。这种证据可能存在于私密数据室,但这里不可见。因此,投资者应把这些区间用作价格纪律护栏,而不是声称 Xaira 的天花板被永久封顶。[CV021, CV022, CV023, CV024, CV025, CV026]
| 情景 | 假设 | 估值 / 回报逻辑 | 关键风险 | 概率信号 |
|---|---|---|---|---|
| 熊市 | 科学兴趣延续,但没有具名合作或清晰资产转化出现;定价仍不透明 | $1.5B-$2.5B 参考价值,接近当前公开 AI-techbio 集群;没有干净路径证明大幅溢价合理 | 开放科学牵引没有转化;公开证据仍然输入重、输出轻 | 如果 12-18 个月过去仍没有具名交易对手或可衡量的平台到资产证明,最可能落入该情景 |
| 基准 | 出现一项严肃合作、战略交易对手或可信内部资产证明;团队和资本溢价保持完好 | $3B-$5B 参考价值;相对公开可比公司给资本、人才密度和首个转化信号溢价 | 证明可能仍窄或不可重复;价格仍可能跑在证据前面 | 如果 Xaira 形成一个决定性证明点,但尚未形成模式,最可能落入该情景 |
| 牛市 | 多个证明点出现:合作牵引、可见资产转化,以及私有市场对前沿 AI 生物学的持续兴趣 | $6B-$9B 参考价值;需要稀缺性溢价延续,并有可信运营证明支撑 | 转化或迁移稍有滑坡,溢价会迅速塌缩 | 需要可重复性,而不只是一条公告或一次技术发布 |
区间为 $B 股权参考价值,意在作为公开证据护栏。由于实际 Xaira 入场价未披露,它们不是回报测算。
[CV021, CV022, CV023, CV024, CV025, CV026]区间图展示 Xaira 在悲观、基准和乐观情景下的低 / 基准 / 高参考值,并叠加当前公开可比公司区间。
[CV021, CV022, CV023, CV024, CV025, CV026]8.5 退出准备度、尽调问题与最终结论
按公开证据看,Xaira 尚未达到 IPO 就绪。私有股没有公开定价历史,没有披露财务报表,没有可见商业指标,也没有成熟合规包。这不说明公司弱,只说明公开记录不完整。因此,较近的价值实现路径更可能是合作、合作方支持的资产,或战略收购逻辑,而不是近期独立上市。大型 制药公司、战略 techbio 公司,或主要 AI 与数据平台拥有者,是最自然的对手方类型,因为 Xaira 的差异化叙事是模型、数据生成和疗法雄心的组合,而不是单一软件产品。 这直接引出最终尽调问题。投资者需要投资条款清单和股权结构表,而不只是故事;需要现金消耗和基于里程碑的预算逻辑,而不只是 $1B 这个头条数字;需要合作管线数据、定价逻辑,以及平台正在产出资产级决策或吸引外部伙伴的证据。投资者也需要后续买方或监管者会索要的合规和安全材料。因此,公开证据质量是不对称的:团队和科学强,经济性和条款弱。最终结论:继续研究。只有当定价接近公开 可比公司集群加合理溢价,且私下尽调补上最大证据缺口时,才应继续。如果定价假设了尚不可见的前沿级别证明,就应放弃,并在证明跟上后再看。[CV029, CV030, CV031, CV032, CV033, CV034]
| 触发项 | 阈值 | 对投资论点的传导 | 行动含义 |
|---|---|---|---|
| 缺乏证据支撑的激进定价 | 私募轮对 Xaira 的定价远高于「可比公司集群 + 溢价」逻辑,且没有私有证据补齐关键缺口 | 投资论证变成叙事套利,而非证据支撑的投资判断 | 条款或证据改善前放弃 |
| 未出现具名合作或商业交易方 | 12-18 个月内仍只有科学宣传,未出现清晰的战略或商业转化 | 削弱开放科学通往商业质地的变现桥梁 | 转向悲观区间,并收紧估值上限 |
| 无资产层面转化证据 | 平台与靶点、分子或监管里程碑之间仍没有可衡量连接 | 削弱平台能复利生成疗法价值的判断 | 大幅折减平台溢价 |
| 安全 / 合规材料包仍缺失 | 大型买家开始接触,但 Xaira 拿不出可信的尽调材料 | 采购摩擦升级为结构性商业化障碍 | 准备度改善前推迟投资 |
| 核心领导层不稳定 | 核心科学、技术或临床负责人离职 | 削弱投资人今天愿意支付的主要溢价 | 从可比公司底部重新核保,而不是按溢价情景 |
| 资本消耗快于证据生成 | 预算消耗上升,但合作伙伴或资产证据没有同步出现 | 融资规模从资产变成预警信号 | 重新评估下行稀释和下一轮融资时间风险 |
这些是可监控的投资论点失效标准,不是概率估计;目的在于公开条款缺位时守住价格纪律。
[CV031, CV032, CV033, CV034, CV039, CV040]| 主题 | 缺失证据 | 重要性 | 负责人 / 尽调路径 |
|---|---|---|---|
| 价格与条款清单 | 投后估值、每股价格、证券类型、优先权条款和投资人权利 | 没有这些,公开证据无法落到可投或不可投的价格上 | 董事会 / CFO / 法务尽调 |
| 股权结构表与稀释瀑布 | 完全稀释后持股、期权池、清算优先权、侧信和任何结构化融资条款 | 决定同一个名义估值是否对应可接受的普通股经济性 | 财务 + 法务资料室 |
| 现金消耗与里程碑预算 | 按职能拆分的现金跑道、>$1B 融资计划用途,以及与支出挂钩的里程碑计划 | 只有在下一次融资节点变得关键前买到决定性证据,资本规模才是优势 | CFO / FP&A / 运营审查 |
| 合作与商业管线 | 具名交易方、讨论阶段、定价逻辑和预期经济性 | 这是从科学证据走向商业证据最干净的桥 | BD / CEO / 管线审查 |
| 平台到资产的转化 | 将 Orion、Pisces、X-Cell 或内部模型与靶点或分子决策相连的案例 | 转化证据是估值高于公开可比公司的最重要依据 | CSO / CTO / 投资组合委员会尽调 |
| 安全与合规准备度 | DPA、审计材料、模型治理、质量体系和客户尽调材料 | 未来买家、合作伙伴和监管方在规模化部署或用于申报之前都会要求这些材料 | 安全 / 法务 / 质量尽调 |
这些问题聚焦具体信息:它们能把 Xaira 从有意思的公开叙事推进到可定价的私募机会。
[CV030, CV032, CV033, CV034, CV035, CV036]面向 IC 的评分卡,从市场、团队、产品证明、商业证明、经济性可见度、风险、估值支撑、融资韧性和证据质量等维度评价 Xaira。
[CV006, CV008, CV009, CV010, CV021, CV037]8.6 展品
免责声明
本报告是基于公开证据的尽调快照,不构成投资建议。重要的财务、法律、技术和合同事实仍未公开;作出任何投资决定前,应直接向管理层和一手文件核验。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | Xaira Therapeutics was incorporated in 2023, operated in stealth, and launched publicly on April 23, 2024. | 高 | SO011, SO020 |
| CO002 | Xaira was jointly incubated by ARCH Venture Partners and Foresite Labs before launch. | 高 | SO011, SO022 |
| CO003 | Xaira’s official address is 700 Gateway Blvd, 4th Floor, South San Francisco, California. | 高 | SO001, SO010 |
| CO004 | Xaira describes itself as an integrated biotechnology or AI life sciences company focused on end-to-end drug discovery and development. | 高 | SO001, SO002 |
| CO005 | Official materials define Xaira’s operating model around three pillars: advanced AI research, expansive data generation, and therapeutic product development. | 高 | SO002, SO011 |
| CO006 | Xaira says its models are intended to support target identification, therapeutic design, and patient or disease-state selection across the drug development process. | 高 | SO001, SO016 |
| CO007 | David Baker is a Xaira co-founder and scientific advisor who directs the Institute for Protein Design at the University of Washington and won the 2024 Nobel Prize in Chemistry. | 高 | SO008, SO012 |
| CO008 | Marc Tessier-Lavigne is a Xaira co-founder, chairman, and CEO who previously served as Genentech chief scientific officer and as president of both Rockefeller University and Stanford University. | 高 | SO004, SO011 |
| CO009 | Hetu Kamisetty is a Xaira co-founder and CTO whose background includes Meta and postdoctoral work in David Baker’s lab. | 高 | SO005, SO013 |
| CO010 | Launch coverage characterizes Robert Nelsen and Vik Bajaj as the venture co-founding sponsors who helped assemble Xaira. | 高 | SO020, SO021 |
| CO011 | Xaira recruited researchers associated with RFdiffusion and RFantibody from David Baker’s lab into the company. | 高 | SO011, SO018 |
| CO012 | Launch materials said Xaira integrated functional genomics capabilities spun out from Illumina and a proteomics group from Interline Therapeutics. | 高 | SO011, SO020 |
| CO013 | Debbie Law joined Xaira as chief scientific officer in October 2024 after senior research roles at Bristol Myers Squibb, Merck, Jounce, and Ablynx. | 高 | SO007, SO012 |
| CO014 | Paulo Fontoura joined Xaira as chief medical officer effective early 2025 after a long Roche career spanning translational medicine and clinical development. | 中 | SO013 |
| CO015 | Bo Wang joined Xaira in April 2025 as SVP and head of biomedical AI after academic leadership roles at the University of Toronto, University Health Network, and the Vector Institute. | 高 | SO014, SO023 |
| CO016 | Jeff Jonker joined Xaira as president and COO in July 2025 to help scale operations and business development. | 高 | SO006, SO015 |
| CO017 | Rachel Lane joined Xaira in March 2026 as SVP of business development and operations to help drive partnerships and operational scale. | 高 | SO009, SO016 |
| CO018 | By 2026 Xaira publicly disclosed South San Francisco, Seattle, and London as its office or innovation-center footprint. | 高 | SO010, SO016, SO028 |
| CO019 | GeekWire reported in August 2024 that Xaira had about 80 employees, with roughly 15 in Seattle and a handful in London while most staff remained in the Bay Area. | 中 | SO018 |
| CO020 | Endpoints reported at launch that Xaira had about 50 employees across Seattle and California. | 中 | SO020 |
| CO021 | Xaira launched with more than $1 billion of committed capital from ARCH, Foresite, and a syndicate of named investors. | 高 | SO011, SO017, SO020 |
| CO022 | Bob Nelsen said ARCH alone planned to contribute more than $200 million and described Xaira’s committed capital as hard money that could grow beyond the initial figure. | 中 | SO021 |
| CO023 | Named backers at launch included F-Prime, NEA, Sequoia, Lux, Lightspeed, Menlo, Two Sigma Ventures, PICI, Byers Capital, Rsquared, and SV Angel. | 高 | SO011, SO022 |
| CO024 | Xaira announced in December 2024 that it would move its headquarters to BioMed Realty’s Gateway of Pacific III campus in South San Francisco. | 中 | SO013 |
| CO025 | A local-development report said Xaira planned to occupy 73,075 square feet of new San Francisco space around July 1, 2025. | 中 | SO024 |
| CO026 | Xaira’s publicly disclosed board includes Scott Gottlieb, Alex Gorsky, Carolyn Bertozzi, Stephen Knight, Mathai Mammen, Robert Nelsen, Richard Scheller, Bryan White, and Marc Tessier-Lavigne. | 高 | SO003, SO011 |
| CO027 | Xaira’s scientific advisory bench includes David Baker, Regina Barzilay, Anima Anandkumar, Chris Garcia, Rod MacKinnon, Sarah Teichmann, Jonathan Weissman, Tim Behrens, Richard Heyman, George Kadifa, and Olivia Zetter. | 中 | SO003 |
| CO028 | Xaira’s core strategic pitch is that AI can shorten the path from lab insight to clinical candidates for previously difficult or undruggable targets. | 高 | SO019, SO029 |
| CO029 | GeekWire reported that Xaira was founded to build on Institute for Protein Design tools such as RFdiffusion and ProteinMPNN and extend them into therapeutics. | 中 | SO018 |
| CO030 | Endpoints reported that Xaira’s initial therapeutic focus was antibody drugs, even though management believed the platform could extend to other modalities over time. | 中 | SO020 |
| CO031 | Launch reporting said Xaira declined to disclose when it expected to have its first drug in human trials. | 中 | SO017 |
| CO032 | By March 2026 Xaira executives were still describing the company as actively building a pipeline rather than presenting a named clinical-stage asset. | 中 | SO029 |
| CO033 | The Stanford review that preceded Marc Tessier-Lavigne’s 2023 resignation found important flaws and shortcomings in papers from his lab and faulted him for not decisively correcting the scientific record, while not accusing him of personal fraud. | 高 | SO024, SO025 |
| CO034 | Retraction Watch reported that two Science papers bearing Tessier-Lavigne’s name were retracted after an institutional investigation found manipulated data by others in his lab. | 中 | SO025 |
| CO035 | TechCrunch said some observers viewed Tessier-Lavigne’s appointment as Xaira CEO as unexpected because he had resigned from Stanford only months earlier. | 中 | SO017 |
| CO036 | Endpoints’ launch profile preserved outside skepticism that de novo antibody generation was mature enough to make medicines even as Xaira argued the technology was ready. | 中 | SO020 |
| CO037 | The combination of more than $1 billion of committed capital, marquee founders, and a high-profile board made Xaira one of the best-backed AI drug discovery startups to emerge in 2024. | 高 | SO011, SO020, SO022 |
| CO038 | Reviewed public sources disclose committed capital but do not disclose a contemporaneous private valuation for Xaira as of 2026-05-12. | 高 | SO011, SO017, SO020, SO022 |
| CO039 | Reviewed public sources do not disclose revenue, customer count, cash on hand, or a named first clinical candidate for Xaira as of 2026-05-12. | 高 | SO001, SO017, SO029 |
| CO040 | Rachel Lane’s appointment materials described Xaira as both a platform and a pipeline company spanning target identification, drug design, and patient stratification. | 中 | SO016 |
| CO041 | Xaira publicly unveiled X-Atlas/Orion in June 2025 as a genome-wide Perturb-seq dataset profiling more than 8 million single cells. | 中 | SO027 |
| CO042 | Xaira launched X-Cell in March 2026 as a 4.9-billion-parameter virtual cell model trained on the 25.6 million-cell X-Atlas/Pisces dataset. | 高 | SO028, SO029 |
| CO043 | Xaira said it would make a subset of the Pisces dataset and the X-Cell model available to the scientific community. | 中 | SO028 |
| CO044 | Fierce Biotech reported that Xaira’s disclosed therapeutic focus in 2026 centered on inflammatory and immunological diseases and antibody therapeutics. | 中 | SO029 |
| CO045 | GeekWire quoted Xaira scientists describing the company’s ambition as “conquering undruggable targets.” | 中 | SO018 |
| CO046 | Launch coverage compared Xaira with earlier AI-enabled drug discovery companies and highlighted that the broader field remained in its early innings despite large funding rounds. | 高 | SO017, SO020 |
| CM001 | Xaira publicly positions itself as an integrated AI life sciences company that combines AI research, data generation, and therapeutic product development rather than as a single-point software vendor. | 高 | SM001, SM002, SM026 |
| CM002 | Official Xaira materials say its AI capabilities are meant to span biological discovery, molecule design, and clinical development. | 高 | SM002, SM026 |
| CM003 | Xaira says its data platform spans molecular-to-human-scale data and is designed to make biology more computable. | 中 | SM002 |
| CM004 | Independent 2026 reporting says Xaira is actively building an inflammatory and immunological pipeline, initially working on antibody therapeutics, while treating the AI platform as the engine that comes first. | 高 | SM004, SM005 |
| CM005 | Xaira says X-Cell is intended for target identification, mechanism-of-action work, matching targets to patients, and toxicity prediction. | 中 | SM003 |
| CM006 | Mordor Intelligence estimates the AI drug discovery market at $3.25 billion in 2026, growing to $10.29 billion by 2031 at a 25.94% CAGR. | 中 | SM009 |
| CM007 | Worldmetrics compiles alternative AI drug discovery estimates, including $2.3 billion in 2023 to $6.2 billion in 2028 at a 21.9% CAGR and $1.5 billion in 2020 to $10.9 billion in 2030 at a 24.8% CAGR. | 低 | SM010 |
| CM008 | McKinsey cites a broader pharma AI market projection of more than $4 billion in 2025 to $25.7 billion by 2030, which is directionally useful but not equivalent to AI drug discovery alone. | 中 | SM011 |
| CM009 | Mordor says pharmaceutical and biotechnological companies represented 67.43% of AI drug discovery spend in 2025, while academic and research institutes are the fastest-growing end-user segment. | 中 | SM009 |
| CM010 | Mordor says target identification and validation held 28.43% of AI drug discovery spend in 2025, while de novo design is the fastest-growing application at a 28.54% CAGR. | 中 | SM009 |
| CM011 | Precedence Research sizes the inflammatory disease market at $133.50 billion in 2026 and $241.34 billion by 2035, implying a 6.80% CAGR. | 中 | SM019 |
| CM012 | Precedence says biologics account for 45% of the inflammatory disease market by drug class. | 中 | SM019 |
| CM013 | Global Market Insights sizes the anti-inflammatory drugs market at $141.3 billion in 2026 and $293.4 billion by 2035 at an 8.5% CAGR. | 中 | SM021 |
| CM014 | Global Market Insights says anti-inflammatory biologics held a 75.5% share in 2025. | 中 | SM021 |
| CM015 | Fortune Business Insights sizes the immunology market at $123.05 billion in 2026 and $228.18 billion by 2034 at an 8.02% CAGR. | 中 | SM020 |
| CM016 | Fortune says monoclonal antibodies are expected to hold 65.02% of the immunology market in 2026 and hospital pharmacies 48.16% of distribution. | 中 | SM020 |
| CM017 | Coherent Market Insights sizes the immunology market at $122.16 billion in 2026 and $280.35 billion by 2033 at a 12.6% CAGR. | 低 | SM022 |
| CM018 | Coherent Market Insights sizes the broader antibodies market at about $323.0 billion in 2026 and $764.7 billion by 2033, which is much larger than Xaira’s disclosed focus because it spans disease areas beyond immunology and inflammation. | 低 | SM024 |
| CM019 | Precedence sizes the antibody production market at $31.71 billion in 2026 and $93.76 billion by 2035 at a 12.83% CAGR. | 中 | SM023 |
| CM020 | Precedence says pharmaceutical and biotechnology companies account for more than 56% of antibody production end-use demand. | 中 | SM023 |
| CM021 | Xaira’s practical commercial opportunity spans two different revenue pools: near-term platform or partnering spend and long-term therapeutic revenue in inflammatory and immunology markets. | 中 | SM002, SM004, SM009, SM019, SM020 |
| CM022 | Large pharma and large biotech R&D or BD organizations are the primary near-term budget owners for Xaira-like platform deals, while Xaira itself is the primary internal buyer for its owned-asset path. | 中 | SM004, SM009, SM023 |
| CM023 | Academic and translational institutes matter more as validation collaborators and secondary users than as core economic buyers for Xaira. | 中 | SM007, SM009, SM023 |
| CM024 | On the owned-drug path, clinicians and patients are users, but insurers, government programs, and hospital-controlled channels are the economic payers. | 中 | SM020, SM022 |
| CM025 | North America carries roughly 40% to 55% share across multiple immunology and inflammatory-market estimates, making US reimbursement and launch dynamics central to Xaira’s commercialization case. | 中 | SM019, SM020, SM022 |
| CM026 | McKinsey says pharma has not yet seen substantially shorter development timelines or better preclinical or clinical success rates despite rapidly rising AI investment. | 中 | SM011 |
| CM027 | McKinsey argues that successful AI deployment in pharma requires redesigning workflows rather than bolting AI onto legacy processes. | 中 | SM011 |
| CM028 | McKinsey also highlights data quality, integrated tech stacks, and cross-functional talent as prerequisites for scaling AI beyond pilots. | 高 | SM011, SM012 |
| CM029 | Deloitte’s 2025 lab survey found 53% of respondents reported increased throughput, 45% reduced human error, 30% greater cost efficiencies, and 27% faster therapy discovery from lab modernization. | 中 | SM014 |
| CM030 | The same Deloitte survey found only 11% of labs are fully predictive today, while 59% plan to continue investing in lab modernization over the next two to three years. | 中 | SM014 |
| CM031 | Deloitte’s earlier survey found more than 60% of life sciences companies spent over $20 million on AI initiatives and identified business-case selection, data, and integration as top challenges. | 中 | SM013 |
| CM032 | McKinsey’s R&D productivity work says the industry barely recouped the full value of capital over the past decade and remains burdened by rising costs and declining success probabilities. | 高 | SM012, SM017 |
| CM033 | Accenture estimates that bringing a new treatment to market costs roughly $2.6 billion to $6.7 billion and argues digital and data-led R&D could save $1.2 billion to $1.7 billion per successful medicine. | 中 | SM016 |
| CM034 | The ACS Omega review says traditional drug discovery often takes more than a decade, costs above $2 billion, and only about 10% of clinical candidates reach approval; high-throughput screening hit rates are around 2.5%. | 中 | SM018 |
| CM035 | GEN’s X-Cell coverage says target identification to approval takes about thirteen years on average and roughly 90% of molecules fail in the clinic. | 中 | SM008 |
| CM036 | Mordor says R&D productivity declined 40% from 2010 to 2024 and notes examples where predictive algorithms shortened lead-optimization cycles from 18 months to 6 months. | 中 | SM009 |
| CM037 | McKinsey reports that some organizations have accelerated preclinical candidate nomination to first-subject-in from 21–26 months to 12–15 months and moved molecules to IND nine months faster through learning loops. | 中 | SM012 |
| CM038 | Mordor identifies explainability, talent scarcity, data fragmentation, and IP or liability uncertainty as structural restraints on AI drug discovery adoption. | 中 | SM009 |
| CM039 | Mordor says emerging 2025 regulatory guidance requires AI model lineage and decision-boundary documentation, adding compliance overhead and slowing validation. | 中 | SM009 |
| CM040 | IQVIA’s 2026 report says development timelines worsened in 2025, inter-trial intervals increased by three months, and immunology remained a long-term growth area even amid short-term volatility. | 中 | SM015 |
| CM041 | IQVIA also reports 79 novel active substances launched globally in 2025 and says AI increasingly enabled R&D, suggesting progress but not yet system-wide proof of productivity transformation. | 中 | SM015 |
| CM042 | X-Cell was trained on 25.6 million perturbed single-cell transcriptomes across seven biological contexts, at 4.9 billion parameters, with a roadmap into primary cells, iPSC-derived cell types, organoids, and in vivo perturbations. | 高 | SM003, SM025 |
| CM043 | Fierce reports that Xaira is using perturbation data to search for previously unknown inflammatory and immunology targets, including in T-cell activation. | 中 | SM004 |
| CM044 | Drug Discovery & Development says Bo Wang frames the product as a virtual cell built around an AI prediction-validation loop in which wet-lab data improves the model. | 中 | SM007 |
| CM045 | Endpoints reports that Xaira initially focused on antibodies and believes AI could deliver two- to three-fold improvements in speed and success if deployed across discovery, molecule design, and clinical trials. | 中 | SM005 |
| CM046 | GeekWire notes that biologics accounted for roughly one-third of drug approvals in 2022, supporting the view that Xaira’s protein and antibody emphasis targets a large established modality. | 中 | SM006 |
| CM047 | The market estimates conflict because some sources measure end-market drug revenue, others discovery-platform spend, and others support ecosystems such as antibody production. | 中 | SM009, SM019, SM020, SM021, SM023, SM024 |
| CM048 | No independent public source reviewed here isolates an exact market size for the specific intersection of virtual-cell models, antibody design, and inflammatory-disease therapeutics that Xaira appears to be pursuing. | 高 | SM004, SM009, SM019, SM020, SM023 |
| CM049 | The most defensible way to size Xaira is with multiple lenses rather than one TAM: small but fast AI-platform spend, mid-sized antibody/discovery ecosystem spend, and very large downstream therapeutic value pools. | 中 | SM009, SM019, SM020, SM023 |
| CM050 | High treatment costs, adverse-effect risk, and biosimilar or reimbursement pressure mean that only part of the large immunology and inflammatory end-market is realistically available for premium pricing by a new entrant. | 中 | SM020, SM021, SM022 |
| CP001 | Xaira's relevant competitive set includes direct AI-first therapeutics platforms, adjacent biologics-design specialists, substitute software/tooling vendors, and large-pharma internal build because Xaira publicly frames itself as end-to-end AI research plus data generation plus therapeutics. | 高 | SP001, SP002, SP003 |
| CP002 | Xaira's currently disclosed wedge is inflammatory and immunological antibody therapeutics powered by causal cell-biology data rather than a generic horizontal AI software product. | 高 | SP001, SP002, SP003 |
| CP003 | The closest direct peers for Xaira are Generate, Isomorphic, insitro, Recursion/Exscientia, and Absci because each publicly combines differentiated AI with proprietary data, experimental systems, or internal / partnered asset creation. | 中 | SP004, SP008, SP012, SP016, SP025, SP026 |
| CP004 | Chai Discovery and Nabla Bio are adjacent rather than perfect full-stack matches, but they crowd the same antibody and protein-design budget that Xaira's early biologics wedge will likely target. | 中 | SP003, SP021, SP022, SP023, SP024 |
| CP005 | Schrödinger and large-pharma internal build are substitute paths rather than direct replicas of Xaira because they let buyers solve parts of the same discovery job through software, partnered discovery, and internal data/chemistry stacks. | 中 | SP013, SP017, SP027, SP028, SP029 |
| CP006 | Generate says it has generated, built, and tested 42,000 proteins and operates more than 140,000 square feet of space, underscoring unusually large wet-lab scale for an AI-biologics peer. | 中 | SP004 |
| CP007 | Generate's Novartis collaboration includes $65 million upfront, more than $1 billion in milestones, and tiered royalties up to low double-digits. | 高 | SP006, SP007 |
| CP008 | Fierce reports that Generate also has an Amgen collaboration worth up to $1.9 billion, raised a $273 million Series C after a $370 million Series B, and has two clinical candidates. | 中 | SP007 |
| CP009 | Isomorphic publicly positions itself as an autonomous AI drug-design company building on and beyond AlphaFold. | 高 | SP008, SP009 |
| CP010 | Isomorphic's IsoDDE article claims strong gains versus AlphaFold 3 on difficult protein-ligand systems and antibody-antigen interfaces, supporting technical credibility in both small molecules and biologics. | 中 | SP009 |
| CP011 | Independent press coverage says Isomorphic's Lilly and Novartis partnerships are worth nearly $3 billion combined, with $45 million and $37.5 million upfronts respectively. | 中 | SP010, SP011 |
| CP012 | insitro says it integrates human clinical data with cellular data and, across 2025–2026 company materials, points to more than $700 million to roughly $800 million of capital raised plus meaningful collaboration revenue. | 高 | SP012, SP013, SP014 |
| CP013 | insitro's expanded BMS collaboration triggered a $10 million milestone payment tied to the nomination of two additional ALS targets. | 中 | SP014 |
| CP014 | insitro's Lilly collaborations show a less traditional packaging model in which insitro can retain global program rights while Lilly contributes technology, receives milestones, or earns royalties. | 高 | SP013, SP015 |
| CP015 | Recursion says it has aggregated more than 50 petabytes of biological and chemical data and uses that stack to support both internal programs and major partnerships. | 高 | SP016, SP017 |
| CP016 | Recursion's partner page says the Sanofi collaboration started with a $100 million upfront payment and can yield up to $5.2 billion plus royalties, while Bayer can reach up to $1.5 billion plus royalties. | 中 | SP017 |
| CP017 | Independent coverage of the Recursion–Exscientia merger says the transaction valued Exscientia at about $688 million, set 74/26 post-close ownership, and combined roughly $850 million of cash. | 中 | SP018, SP030 |
| CP018 | Recursion's 2025 pipeline cuts are material adverse evidence because they followed the merger and reflect both execution prioritization and investor concern about burn. | 高 | SP019, SP020 |
| CP019 | Chai Discovery publicly markets de novo antibody design against challenging targets with atomic precision, but the reviewed public sources do not show a full Xaira-like end-to-end commercial footprint. | 中 | SP021 |
| CP020 | Nabla combines de novo design with human-relevant testing and has public evidence of both financing and repeat pharma collaborations, including Takeda. | 高 | SP022, SP023, SP024 |
| CP021 | Absci says it operates a 77,000+ square-foot wet lab, screens antibody variants at more than 4,000x traditional throughput, and can cycle from data to validated designs in about six weeks. | 高 | SP025, SP026 |
| CP022 | Schrödinger is better viewed as a substitute computational design platform than as a direct Xaira-style causal-biology peer because its public story centers on software, simulation, and partnered molecular discovery across many modalities. | 高 | SP027, SP028, SP029 |
| CP023 | Schrödinger publicly discloses Lilly collaboration economics up to $425 million plus low single- to low double-digit royalties in immunology. | 中 | SP028 |
| CP024 | Large pharma buyers are also platform builders: Lilly's TuneLab appears inside Schrödinger's LiveDesign ecosystem while Novartis, Sanofi, Bayer, and others are simultaneously running their own AI-enabled discovery agendas with external partners. | 中 | SP011, SP013, SP017, SP029 |
| CP025 | Public monetization across reviewed peers is dominated by bespoke collaborations with milestones, research funding, equity, and royalties rather than transparent per-seat SaaS pricing. | 高 | SP006, SP011, SP013, SP014, SP017, SP023, SP028 |
| CP026 | No reviewed private-company peer publishes transparent list pricing for AI drug discovery access; Chai, Nabla, Absci, and Xaira all describe capabilities without price sheets. | 中 | SP001, SP021, SP022, SP025, SP026 |
| CP027 | The biologics and antibody-design wedge around Xaira is crowded because Generate, Absci, Chai, and Nabla all explicitly position AI around proteins, antibodies, or biologics discovery. | 高 | SP003, SP005, SP021, SP024, SP025, SP026 |
| CP028 | Xaira's most plausible differentiation versus biologics-design peers is its emphasis on causal perturbation data and virtual-cell modeling, not antibody design alone. | 中 | SP001, SP002, SP003 |
| CP029 | Isomorphic and Schrödinger provide stronger public evidence of frontier model or computational-design tooling than of a broad Xaira-like wet-lab causal-biology system. | 中 | SP009, SP027, SP028 |
| CP030 | Recursion and insitro have much stronger public proof of repeat big-pharma go-to-market than Xaira currently does. | 中 | SP003, SP014, SP017, SP019 |
| CP031 | Big pharma appears comfortable multi-homing across AI-discovery vendors: Lilly works with Isomorphic, insitro, and Schrödinger; Novartis with Isomorphic and Generate; Sanofi and Bayer with Recursion. | 高 | SP006, SP011, SP013, SP017, SP028 |
| CP032 | In AI biopharma, switching costs are more likely to come from proprietary data, embedded workflows, and asset-rights structures than from simple user-interface lock-in. | 中 | SP001, SP014, SP017, SP028 |
| CP033 | Public partner disclosures often omit exclusivity, exact target counts, or downstream profit splits, which limits apples-to-apples comparison of moat strength. | 中 | SP010, SP014, SP017, SP028 |
| CP034 | Recursion's merger plus subsequent pipeline cuts are disconfirming evidence that more data, more capital, and more programs do not automatically translate into durable execution or pricing power. | 高 | SP018, SP019, SP020, SP030 |
| CP035 | Generate's and Nabla's disclosed biologics deals show that AI-biologics platforms can win very large back-end economics even before broad late-stage clinical validation is public. | 中 | SP006, SP007, SP023, SP024 |
| CP036 | Isomorphic's nearly $3 billion of reported Lilly and Novartis collaborations show that frontier model credibility alone can support billion-dollar back-end economics. | 中 | SP010, SP011 |
| CP037 | insitro's Lilly structure suggests some platforms can negotiate biotech-favorable terms, including retained global rights, rather than classic full handoff discovery deals. | 高 | SP013, SP015 |
| CP038 | Trust and regulatory posture are becoming competitive variables: Generate explicitly discusses responsible AI stewardship and Isomorphic emphasizes benchmark-heavy technical validation rather than marketing alone. | 中 | SP005, SP009 |
| CP039 | Across the reviewed peer set, public disclosures emphasize discovery capability and partnership optionality much more than marketed products or late-stage clinical proof. | 中 | SP007, SP011, SP019, SP025, SP028 |
| CP040 | The reviewed sources still do not provide enough public evidence to benchmark Xaira's actual pricing power or customer traction against top peers. | 中 | SP001, SP002, SP003 |
| CP041 | No reviewed public source disclosed named Xaira platform partnership economics or a named clinical-stage Xaira program. | 中 | SP001, SP002, SP003 |
| CP042 | For valuation, the most relevant competitive benchmark is likely partner economics and data-moat credibility rather than generic software multiples alone. | 中 | SP003, SP006, SP017, SP028 |
| CP043 | Xaira's X-Cell disclosure gives the company a credible scale claim before commercial proof because it cites 25.6 million perturbed single-cell transcriptomes and a 4.9-billion-parameter model. | 中 | SP002 |
| CP044 | Xaira's central moat question is whether causal-cell biology can turn into partner economics or internal assets before buyers settle on other platforms and multi-homing habits. | 中 | SP002, SP011, SP017, SP028 |
| CP045 | The competitive evidence therefore sets a high bar for Xaira: scientific novelty is visible, but commercial benchmarkability remains to be proven publicly. | 中 | SP002, SP003, SP030 |
| CI001 | Xaira launched in 2024 with more than $1 billion of committed capital. | 高 | SI001, SI007 |
| CI002 | Xaira's official spending agenda spans AI research, expansive data generation, and therapeutic product development rather than a narrow software SKU. | 高 | SI001, SI002 |
| CI003 | No reviewed public source disclosed Xaira revenue, collaboration revenue, or a named external platform customer. | 中 | SI001, SI002, SI003, SI004 |
| CI004 | The most plausible near-term Xaira monetization paths are collaboration revenue, milestones, royalties, and asset deals rather than direct product sales. | 中 | SI002, SI019, SI020, SI021, SI022, SI023 |
| CI005 | Any meaningful internal product revenue would be long-duration because no reviewed public source disclosed a clinical-stage Xaira asset. | 中 | SI003, SI004 |
| CI006 | Xaira's revenue quality cannot currently be treated as recurring or diversified because public monetization evidence is absent. | 中 | SI003, SI004 |
| CI007 | Independent reporting moved Xaira's public staffing proxy from about 50 employees at launch to roughly 80 employees in 2024, with most in the Bay Area and 15 in Seattle. | 中 | SI005, SI007 |
| CI008 | An independent development tracker reported that Xaira planned a 73,075 square foot San Francisco buildout in 2025. | 低 | SI008 |
| CI009 | Xaira's official X-Cell roadmap expands data generation beyond current perturbation datasets into primary cells, organoids, and in vivo perturbations. | 中 | SI003 |
| CI010 | Drug Discovery Trends frames Xaira's operating requirements around talent, compute, and data, enabled by very large funding. | 中 | SI006 |
| CI011 | Fierce quotes Xaira leadership that the company's integrated R&D plan will take multiple years and perhaps a billion dollars or more. | 中 | SI004 |
| CI012 | Taken together, public Xaira sources imply a cost structure dominated by talent, compute, wet-lab experimentation, and therapeutic development rather than by sales and marketing. | 中 | SI002, SI003, SI004, SI005, SI006 |
| CI013 | Recursion's full-year 2025 results were $74.7 million of revenue, $475.3 million of R&D expense, $753.9 million of cash, and runway into early 2028. | 中 | SI009 |
| CI014 | Recursion reported 2025 cash operating expense of about $399.2 million and expects 2026 cash operating expense below $390 million. | 中 | SI009 |
| CI015 | Schrödinger's full-year 2025 results were $255.9 million of revenue, $199.5 million of software revenue, $56.4 million of drug-discovery revenue, $309.5 million of operating expenses, and $402.3 million of cash. | 中 | SI012 |
| CI016 | Schrödinger's 74% software gross margin and hosted-software transition show why a software-plus-discovery model has very different economics from Xaira's current public profile. | 中 | SI012, SI026 |
| CI017 | Relay's Q1 2025 disclosure showed roughly $710.3 million of cash, $7.7 million of revenue, $73.8 million of R&D expense, and runway into 2029 after cost reductions. | 中 | SI015 |
| CI018 | Absci's Q3 2025 disclosure showed $152.5 million of cash, $0.4 million of revenue, $19.2 million of R&D expense, and runway into the first half of 2028. | 中 | SI016 |
| CI019 | Public AI-biotech comparables place annual cash consumption anywhere from roughly $100 million at smaller scale to almost $400 million at larger full-stack clinical scale. | 中 | SI009, SI012, SI015, SI016 |
| CI020 | Across peer platforms, pricing is bespoke and milestone-heavy rather than list-priced. | 高 | SI019, SI020, SI021, SI022, SI023 |
| CI021 | Financially, Xaira looks closer to a collaboration-driven techbio platform than to a recurring-revenue software company. | 中 | SI002, SI012, SI019, SI020, SI021, SI022, SI023 |
| CI022 | Xaira's current cash on hand is not publicly disclosed. | 中 | SI001, SI003, SI004 |
| CI023 | Because current cash is private, any runway analysis for Xaira has to start from launch capital and peer burn analogs rather than audited balances. | 中 | SI001, SI009, SI012, SI015, SI016 |
| CI024 | A low-confidence Xaira burn proxy of roughly $120 million to $260 million annually is reasonable given the public scale signals and peer range. | 低 | SI004, SI005, SI008, SI009, SI012, SI015, SI016 |
| CI025 | Under that proxy, Xaira likely still has multi-year runway, but not indefinite runway, especially if internal clinical development scales before monetization. | 低 | SI001, SI004, SI009, SI012, SI015, SI016 |
| CI026 | Publicly stated uses of Xaira's capital include model development, data generation, and therapeutic product development across multiple programs and modalities. | 高 | SI001, SI003 |
| CI027 | The likely next-round or major strategic trigger for Xaira is proof of differentiated platform output or internal asset progress, not near-term product revenue. | 中 | SI004, SI018, SI020 |
| CI028 | Debt, project-finance obligations, and fixed lease or compute commitments are not publicly disclosed for Xaira. | 中 | SI001, SI025 |
| CI029 | J.P. Morgan describes the 2026 biopharma capital environment as selective, with licensing and M&A carrying much of the financing load and deal structures remaining milestone-heavy. | 中 | SI018 |
| CI030 | SVB says healthcare fundraising dollars are down and 2025 is on track for the sector's worst fundraising year in more than a decade. | 中 | SI017 |
| CI031 | This external financing backdrop raises the bar for any future Xaira financing despite the large launch round. | 中 | SI001, SI017, SI018 |
| CI032 | Generate, Isomorphic, insitro, Recursion, and Nabla show that differentiated AI platforms can monetize through upfronts, milestones, and royalties well before broad profitability. | 中 | SI019, SI020, SI021, SI022, SI023 |
| CI033 | Unlike Schrödinger, Xaira has no public software ACV, retention, or hosted-revenue metrics. | 中 | SI012, SI026, SI003 |
| CI034 | Unlike Recursion, Xaira has no public collaboration revenue or cash operating expense metrics. | 中 | SI009, SI003 |
| CI035 | Unlike Relay and Absci, Xaira has no public quarterly cash, R&D, or net-loss disclosure. | 中 | SI015, SI016, SI003 |
| CI036 | Any Xaira unit-economics model built today is mostly input-driven rather than financial-statement-driven. | 中 | SI005, SI009, SI012, SI015, SI016 |
| CI037 | The main financial diligence blockers are current cash, actual burn, partner economics, and program-level spend. | 中 | SI003, SI009, SI012, SI017, SI018 |
| CI038 | Xaira is pre-revenue, capital-intensive, and probably still well funded, but too opaque for bottom-up underwriting. | 中 | SI001, SI003, SI004, SI009, SI012 |
| CI039 | Xaira should enter valuation as an option on platform monetization and internal asset creation rather than as a company with proven revenue quality. | 中 | SI003, SI020, SI021, SI022, SI023 |
| CI040 | If Xaira eventually monetizes via collaborations, gross margins could be attractive, but the margin path remains unproven because no realized revenue mix is public. | 低 | SI012, SI013, SI020, SI021, SI022 |
| CI041 | Public-company filing surfaces such as Schrödinger's SEC filings page exist for peers, while Xaira's private status removes that level of financial transparency. | 中 | SI001, SI014 |
| CI042 | Xaira's official work-with-us page indicates the company is still in hiring and build mode rather than harvesting mode. | 低 | SI025 |
| CE001 | Xaira publicly defines itself as an integrated platform spanning advanced AI research, expansive data generation, and therapeutic product development. | 高 | SE001, SE002 |
| CE002 | In workflow terms, Xaira is better understood as an internal drug-discovery operating system than as a publicly commercialized software SKU. | 中 | SE001, SE002, SE019, SE020 |
| CE003 | X-Atlas/Orion introduced FiCS Perturb-seq and an 8 million-cell public atlas targeting all human protein-coding genes, with deep sequencing above 16,000 UMIs per cell. | 高 | SE003, SE004, SE017 |
| CE004 | FiCS Perturb-seq is presented as a scalable, reproducible perturbation platform with high sensitivity, low batch effects, and a workflow leveraging 10x Chromium. | 高 | SE003, SE004, SE017 |
| CE005 | Xaira framed Orion as a public community contribution under non-commercial terms rather than as a private-only dataset. | 中 | SE003, SE017 |
| CE006 | X-Atlas/Pisces expands the causal-data layer to 25.6 million perturbed single-cell transcriptomes across seven CRISPRi screens and 16 biological contexts. | 高 | SE005, SE011, SE014, SE018, SE025 |
| CE007 | X-Cell's public model family runs up to 4.9 billion parameters, while the documented public mini variant is 55M parameters. | 高 | SE005, SE011, SE012, SE014, SE018, SE025 |
| CE008 | X-Cell is publicly described as a set-level diffusion transformer that iteratively refines predictions across four diffusion steps. | 高 | SE011, SE012, SE014, SE018, SE025 |
| CE009 | X-Cell integrates multi-modal biological priors through cross-attention, including ESM-2, STRING, GenePT, DepMap, JUMP-Cell Painting, and gene-level embeddings tied to the model stack. | 高 | SE011, SE012, SE014, SE018, SE025 |
| CE010 | Public X-Cell docs expose a planned API surface built around AnnData / .h5ad control-cell inputs and model.predict() calls. | 中 | SE010, SE013, SE014 |
| CE011 | X-Cell Mini is documented as a 12-layer, 8-head model with four cross-attention layers and a minimum 8 GB GPU footprint. | 中 | SE012 |
| CE012 | The quickstart says X-Cell expects log1p CP10k expression inputs and zero-imputes genes outside its vocabulary. | 中 | SE013 |
| CE013 | Xaira's public roadmap calls for extending the atlas from current perturbation datasets into primary cells, iPSC-derived cell types, organoids, and in vivo perturbations. | 高 | SE005, SE018, SE025 |
| CE014 | Public Xaira materials and related reporting frame X-Cell as useful for target identification, mechanism-of-action work, patient stratification, and toxicity prediction. | 中 | SE005, SE018, SE020, SE025 |
| CE015 | Xaira now has a visible public developer surface spanning GitHub, raw docs, Hugging Face model and dataset cards, and a documented package/API plan. | 高 | SE009, SE010, SE011, SE012, SE013, SE014, SE015, SE016 |
| CE016 | The public X-Cell release is still partial because the repo, model card, docs, and Hugging Face page all say model weights and inference code are coming soon. | 高 | SE009, SE010, SE011, SE014, SE015 |
| CE017 | The Pisces dataset release is also partial: the dataset card says uploads are coming soon and the dataset viewer is unavailable. | 中 | SE016 |
| CE018 | Official and independent X-Cell materials say only a subset of Pisces and X-Cell is being made available to the scientific community. | 中 | SE005, SE025 |
| CE019 | Xaira's operating model is an AI-to-wet-lab validation loop in which predictions guide experiments and experimental output improves future models. | 高 | SE001, SE020, SE021 |
| CE020 | GeekWire reports that Xaira's Seattle team uses high-throughput lab systems to test designed proteins and feed those data back into the models quickly. | 中 | SE021 |
| CE021 | Xaira's molecule-design layer is rooted in Institute for Protein Design work such as RFdiffusion and ProteinMPNN, which GeekWire says the company was founded to build on and extend. | 中 | SE021, SE023 |
| CE022 | Relative to the virtual-cell stack, Xaira's molecule-design and antibody-design layer is strategically important but less publicly specified. | 中 | SE001, SE019, SE020, SE021 |
| CE023 | Fierce reports that Xaira is working on antibody therapeutics and that management sees the AI platform as the precursor to the pipeline it generates. | 中 | SE019, SE025 |
| CE024 | Drug Discovery Trends reports that Xaira wants to connect sequence-model work with expression-model work and explicitly mentions protein-design and antibody-design collaboration with David Baker's team. | 中 | SE020 |
| CE025 | No reviewed public source names a Xaira-originated antibody asset, public protein-design product, or clinically staged program directly generated by the disclosed stack. | 中 | SE001, SE006, SE019, SE020, SE021 |
| CE026 | Xaira's privacy policy, effective January 1, 2025, covers analytics, cookies, fraud protection, and technical/organizational/administrative safeguards for company services. | 中 | SE007 |
| CE027 | Xaira's careers page includes a job-scam alert warning against unofficial platforms and payment requests, showing at least one public security-awareness control. | 中 | SE008 |
| CE028 | No reviewed public source disclosed SOC 2, ISO 27001, HIPAA, GxP, or 21 CFR Part 11 claims for X-Cell, X-Atlas, or their public release surfaces. | 中 | SE006, SE007, SE011, SE014, SE015 |
| CE029 | The clearest public product-governance statement is that X-Cell is intended for research use in computational biology and genomics. | 高 | SE011, SE015 |
| CE030 | No reviewed public source disclosed hosted inference endpoints, uptime SLAs, enterprise support terms, or named production deployments for external users of X-Cell. | 中 | SE006, SE009, SE010, SE011, SE014, SE015, SE016 |
| CE031 | The most visible outside 'users' today are researchers inspecting partial open releases rather than named enterprise customers buying a finished software product. | 中 | SE005, SE006, SE015, SE016, SE017, SE018 |
| CE032 | Xaira's stack depends on 10x-linked perturbation workflows, large-scale wet-lab operations, curated biological priors, GPU compute, and high-throughput validation. | 中 | SE003, SE004, SE012, SE020, SE021 |
| CE033 | Xaira's public differentiation appears to rest more on owning an interventional data-plus-validation loop than on exposing a fully productized external model offering today. | 中 | SE001, SE005, SE019, SE020, SE021, SE025 |
| CE034 | The path from Orion to Pisces shows Xaira broadening from the initial public 8M-cell atlas to a larger, more context-diverse 25.6M-cell training corpus. | 高 | SE003, SE005, SE017, SE018, SE025 |
| CE035 | Xaira's docs and cards show a genuine developer surface, but one that is still aspirational until the key runnable artifacts actually ship. | 中 | SE009, SE010, SE011, SE014, SE015, SE016 |
| CE036 | Because the public package is incomplete, outsiders still cannot reproduce real runtime behavior, benchmark support burden, or audit deployment quality end-to-end. | 中 | SE009, SE011, SE014, SE015, SE016 |
| CE037 | Xaira's official news timeline shows a platform progression from company launch in 2024 to public data release in 2025 and public model release in 2026. | 高 | SE002, SE003, SE005, SE006 |
| CE038 | A March 2026 Business Times preview said Xaira was hiring for 25 positions, consistent with a platform still in buildout mode. | 低 | SE022 |
| CE039 | The Pisces dataset card shows modest but nonzero public traction, including 80 downloads last month and six likes at the time of access. | 低 | SE016 |
| CE040 | The public release package emphasizes open-science and research orientation through non-commercial licensing more than commercial API monetization. | 中 | SE010, SE011, SE014, SE015, SE016 |
| CE041 | Nature and GeekWire evidence suggests Xaira's protein-design layer draws on serious frontier science, but the company has not publicly specified how much of that layer is already industrialized inside its own stack. | 中 | SE021, SE023, SE024 |
| CE042 | Xaira's public roadmap and product story are materially ahead of its public proof package: the evidence is strongest at the data and model layer, weaker at the downstream therapeutic-output and external-product layers. | 中 | SE005, SE019, SE020, SE021, SE025 |
| CU001 | Xaira's visible customer universe splits into internal platform users, open-science researchers, prospective commercial collaborators, and future large-pharma buyers. | 高 | SU001, SU005, SU006, SU007, SU011 |
| CU002 | No reviewed public source disclosed a named paying customer, external platform contract, or recurring-revenue account for Xaira. | 中 | SU001, SU002, SU006, SU007 |
| CU003 | The open scientific community is the clearest externally visible user segment today. | 高 | SU005, SU011, SU012, SU013, SU015 |
| CU004 | Xaira explicitly signals willingness to work with commercial entities interested in collaborating, but no named collaborator-customer is public. | 中 | SU007, SU011 |
| CU005 | Internal Xaira teams are likely still the dominant power users of the platform because public external adoption proof remains limited while platform buildout continues. | 中 | SU001, SU004, SU022, SU024 |
| CU006 | Payer logic differs by segment: academic/open-source use is non-commercial, internal use is Xaira-funded, and future commercial use would likely sit inside biotech or pharma R&D budgets. | 中 | SU001, SU007, SU011 |
| CU007 | R&D World reported that Orion had already been downloaded more than 16,451 times within two weeks of release. | 中 | SU011 |
| CU008 | The Hugging Face Orion discussions page showed 22 likes and two community discussions by the run date. | 中 | SU012 |
| CU009 | An external user named zboldyga asked Xaira for sgRNA count data, and Xaira's Ann Huang replied with exact Figshare filenames, proving a real outside usage-and-support interaction. | 中 | SU012, SU013 |
| CU010 | Hugging Face's parquet-converter made Orion queryable through standard data tools such as DuckDB, Pandas, and Polars, reducing friction for external users. | 中 | SU014 |
| CU011 | Pisces also shows early public interest, with 80 downloads in the last month and six likes on its Hugging Face card. | 中 | SU018 |
| CU012 | Orion is currently the stronger adoption surface than X-Cell because it has download counts, community questions, and external validation, while the model release is newer and less complete. | 中 | SU011, SU012, SU018, SU019 |
| CU013 | Emma Lundberg publicly described X-Atlas/Orion as a significant contribution to the scientific community and robust virtual-cell modeling. | 中 | SU009, SU016 |
| CU014 | Hani Goodarzi said Orion provides substantial resources for training foundation models across the community. | 中 | SU009, SU016 |
| CU015 | External reception is not uniformly bullish: GEN quoted Noetik CEO Ron Alfa arguing that patient-outcome prediction is still a step away from current virtual-cell progress. | 中 | SU010 |
| CU016 | Named external proof today is scientific-validation proof and open-user proof, not commercial case-study proof. | 高 | SU009, SU011, SU012, SU013 |
| CU017 | The open scientific community is the only segment with quantified public adoption evidence today. | 中 | SU011, SU012, SU018 |
| CU018 | Public X-Cell adoption proof is capped by partial shipment: only a subset of the model and Pisces dataset is publicly available. | 中 | SU006, SU017, SU019, SU023 |
| CU019 | No public source disclosed NRR, GRR, churn, renewal rates, contract duration, or retention cohorts for Xaira. | 中 | SU001, SU002, SU006, SU007 |
| CU020 | Downloads, likes, and community questions do not prove repeat usage, customer satisfaction, or revenue durability. | 中 | SU011, SU012, SU013, SU018 |
| CU021 | The only publicly visible repeat-usage proxy is ongoing community interaction months after Orion's release, not a formal renewal metric. | 中 | SU012, SU013, SU018 |
| CU022 | Xaira's reply to zboldyga suggests some external-support responsiveness, but one answered discussion does not imply a mature customer-success motion. | 中 | SU013 |
| CU023 | Because Xaira's public release is open-science and non-commercial in orientation, classical SaaS-style satisfaction and retention metrics are absent by design at this stage. | 中 | SU011, SU018, SU019, SU021 |
| CU024 | The most plausible expansion path is open-science usage into citations, benchmarking, and collaboration inquiries, then into future commercial or pharma deals. | 中 | SU011, SU007, SU008, SU017, SU025 |
| CU025 | RDWorld quotes Xaira saying the dataset is free for academics while the company is happy to work with commercial entities interested in collaborating, implying a two-track adoption model. | 中 | SU011 |
| CU026 | If Xaira monetizes successfully, revenue concentration risk is likely to be high because monetization appears more likely to come from a few large collaborations than from a broad self-serve base. | 中 | SU002, SU007, SU025 |
| CU027 | Commercial procurement friction is elevated because the public package lacks complete shipment, visible enterprise support terms, and public compliance credentials. | 中 | SU006, SU019, SU020, SU021 |
| CU028 | Hiring and buildout signals suggest Xaira is still scaling internal platform capacity rather than operating a mature customer-service organization. | 中 | SU004, SU022, SU024 |
| CU029 | Geographic segmentation of external adoption is largely unknown; the visible evidence is internet-native community use rather than a named institutional customer roster by geography. | 中 | SU011, SU012, SU013, SU017 |
| CU030 | No public evidence shows channel partners, resellers, or marketplaces driving paid customer acquisition for Xaira. | 中 | SU001, SU003, SU006, SU007 |
| CU031 | Xaira is more legible as a future partnership-led business than as a current many-customer software platform. | 中 | SU001, SU006, SU007, SU011 |
| CU032 | Public customer proof is stronger for scientific relevance than for commercial monetization or durability. | 高 | SU009, SU011, SU012, SU013, SU016 |
| CU033 | Press amplification and open-source artifacts do not establish production deployment or long-term revenue quality. | 中 | SU006, SU017, SU019, SU020 |
| CU034 | Any retention cohort or repeat-usage analysis today must be treated as a proxy scenario rather than as company-reported fact. | 中 | SU011, SU012, SU018, SU019 |
| CU035 | Xaira's strongest present-day 'customers' are researchers and evaluators of Orion, Pisces, and X-Cell, not disclosed enterprise buyers. | 中 | SU011, SU012, SU013, SU018, SU019 |
| CU036 | Potential future commercial buyers are likely to be biotech and pharma discovery leaders, translational teams, and computational biology groups rather than general-purpose software buyers. | 中 | SU001, SU006, SU007, SU008 |
| CU037 | Academic and open-source usage can create strategic value even with little immediate revenue by improving benchmarking, citations, and collaborator discovery. | 中 | SU011, SU012, SU013, SU015, SU025 |
| CU038 | The customer evidence is fresh: Orion community engagement was visible by late 2025 and remained public into the 2026 run date. | 中 | SU012, SU013 |
| CU039 | X-Cell customer-like adoption proof lags Orion because the model release is newer and the public package is less complete. | 中 | SU018, SU019, SU020, SU023 |
| CU040 | The customer conclusion for later chapters is that Xaira has real early adoption proof in the open-science community but no publicly legible commercial customer base, retention proof, or revenue diversification yet. | 中 | SU001, SU006, SU011, SU012, SU013, SU019 |
| CR001 | FDA and EMA now treat AI in the drug lifecycle as a risk-managed discipline requiring context-of-use, documentation, and lifecycle oversight. | 高 | SR003, SR004, SR005, SR006 |
| CR002 | The EU AI Act and GDPR create a European layer of obligations around AI systems, safety, rights, and personal-data processing. | 高 | SR001, SR002, SR006 |
| CR003 | EMA explicitly says AI in the medicinal product lifecycle introduces new risks that must be mitigated to protect patient safety and the integrity of clinical evidence. | 中 | SR006 |
| CR004 | FDA's 2025 draft guidance asks sponsors to use a risk-based credibility assessment framework when AI supports regulatory decision-making for drugs and biologics. | 高 | SR003, SR004, SR005 |
| CR005 | If Xaira begins using AI outputs in submission-relevant evidence or regulated workflows, early regulatory interaction is likely expected rather than optional best practice. | 高 | SR003, SR004, SR005, SR006 |
| CR006 | Reviewed public Xaira materials do not disclose product-specific GxP documentation, AI-governance packages, or regulatory-readiness artifacts. | 中 | SR012, SR013, SR016, SR018, SR031, SR032 |
| CR007 | Xaira's public compliance surface is currently anchored by a website privacy policy rather than a product-grade diligence package. | 中 | SR012 |
| CR008 | X-Cell's public release is governed by a CC BY-NC-SA 4.0 license that prohibits commercial use by third parties absent separate rights. | 高 | SR009, SR010, SR030, SR031 |
| CR009 | That non-commercial license improves research dissemination but creates legal friction for commercial embedding, redistribution, or partner reuse. | 中 | SR009, SR010, SR018, SR030 |
| CR010 | Xaira's public legal and compliance packaging is currently sufficient for website and research-distribution contexts but not enough to prove readiness for regulated or enterprise deployment. | 中 | SR004, SR005, SR006, SR007, SR008, SR012 |
| CR011 | Xaira's operating model explicitly combines AI research, expansive data generation, and therapeutic product development in a single loop. | 高 | SR013, SR016, SR021 |
| CR012 | Because the company is trying to move from models to biology to patients, execution failure in any one layer can slow the whole system. | 中 | SR013, SR016, SR021, SR023 |
| CR013 | Xaira's Orion data-generation stack is explicitly tied to 10x Genomics' Chromium platform, making 10x a meaningful workflow dependency. | 高 | SR011, SR017 |
| CR014 | That 10x dependency matters because Xaira's data volume and reproducibility are central to its claimed moat, even if 10x is not the only technology in the stack. | 中 | SR011, SR013, SR017 |
| CR015 | Public X-Cell materials still said model weights and inference code were coming soon by the run date. | 高 | SR030, SR031, SR032 |
| CR016 | Partial shipment limits third-party reproducibility, benchmarking, and buyer diligence. | 中 | SR029, SR030, SR031, SR032 |
| CR017 | Independent commentary in GEN says predicting patient outcomes is still a step away even if virtual-cell models are scientifically valuable. | 中 | SR025 |
| CR018 | The highest scientific risk is translation from perturbation-scale biological prediction into therapeutically useful outcomes. | 中 | SR013, SR023, SR025 |
| CR019 | Xaira's public security disclosure mainly says it uses appropriate measures and that no storage or transmission method is completely secure. | 中 | SR012 |
| CR020 | No reviewed public source disclosed SOC 2, ISO 27001, penetration testing, uptime commitments, disaster recovery detail, or regulated-data certifications for Xaira's platform. | 中 | SR012, SR013, SR031, SR032 |
| CR021 | NIST and CISA both frame AI deployment as a lifecycle security and risk-management problem, increasing the bar Xaira will eventually need to clear with sophisticated buyers or regulated uses. | 高 | SR007, SR008 |
| CR022 | Xaira has meaningful open-science traction but no publicly disclosed paying customers, recurring-revenue accounts, or commercial deployments. | 中 | SR018, SR022, SR026, SR027, SR030 |
| CR023 | Orion downloads, likes, and community discussions prove external interest, but they do not prove contracts or durable deployment. | 高 | SR024, SR026, SR027, SR028 |
| CR024 | The most plausible monetization path still looks like collaborations with biotech or pharma counterparties rather than a broad self-serve software model. | 中 | SR013, SR022, SR023, SR026 |
| CR025 | If monetization is collaboration-led, early revenue concentration is likely to be high because only a small number of counterparties would matter. | 中 | SR022, SR023, SR026, SR033 |
| CR026 | Public releases create imitation and information-leakage risk because competitors can study Xaira's data and model surface before Xaira has publicly proven superior economics. | 中 | SR018, SR024, SR029, SR030, SR031 |
| CR027 | Xaira's visible external distribution currently depends on third-party surfaces such as Hugging Face and GitHub. | 高 | SR027, SR029, SR030, SR031, SR032 |
| CR028 | Those public platforms help community reach but are not substitutes for Xaira-controlled enterprise delivery, support, or auditability. | 中 | SR012, SR027, SR030, SR031, SR032 |
| CR029 | The absence of public enterprise security or compliance materials increases procurement friction for any large biotech or pharma buyer. | 中 | SR007, SR008, SR012, SR020 |
| CR030 | Commercial conversion risk remains unresolved because the public record shows curiosity and validation, not funnel, renewal, or deployment metrics. | 中 | SR024, SR026, SR027, SR028, SR029 |
| CR031 | Xaira has an unusually deep leadership and board bench for a company at this stage, including former FDA Commissioner Scott Gottlieb on the board. | 高 | SR014, SR016 |
| CR032 | That depth does not eliminate concentration risk because a relatively small set of leaders still carries outsized scientific, technical, and clinical credibility. | 中 | SR014, SR019, SR020, SR035 |
| CR033 | Xaira was still building out its senior team through late 2024 and early 2025, which shows the organization remains in active construction mode. | 高 | SR019, SR020 |
| CR034 | Public hiring signals, including a broad jobs page and 25 reported open positions in March 2026, imply meaningful scale-up needs remain. | 高 | SR015, SR034 |
| CR035 | The more-than-$1B launch financing materially reduces immediate insolvency risk. | 高 | SR016, SR020 |
| CR036 | A combined AI-research, wet-lab-data, and therapeutics-development model is still likely capital intensive even with a very large starting raise. | 中 | SR015, SR016, SR023, SR033 |
| CR037 | If Xaira fails to translate scientific credibility into partner, pipeline, or maturity proof before the next financing inflection, dilution or valuation pressure could rise despite its large initial raise. | 中 | SR016, SR026, SR033 |
| CR038 | Xaira's financing risk is therefore less about short-run runway and more about how much proof the market will demand before rewarding the next step up in valuation. | 中 | SR016, SR033 |
| CR039 | Xaira's strongest mitigants are the scale of its starting capital, the quality of its leadership and board, and the scientific-community validation created by Orion and related releases. | 高 | SR014, SR016, SR024, SR026 |
| CR040 | Residual exposure is highest around therapeutic translation, commercial conversion, and security/compliance maturity. | 中 | SR012, SR018, SR025, SR026, SR033 |
| CR041 | The most important thesis-break triggers are failure to ship complete public product surfaces, failure to show any commercial partner proof, inability to produce credible compliance materials, or loss of key leaders. | 中 | SR015, SR019, SR030, SR031, SR033 |
| CR042 | Overall, Xaira's risk profile is execution-heavy: it has unusually strong inputs, but public evidence still stops well short of proving repeatable output in therapeutics, contracts, or compliant deployment. | 中 | SR013, SR016, SR025, SR026, SR033 |
| CV001 | As of May 2026, the closest public AI-techbio comparables trade in roughly a $0.9B-$2.46B market-cap range. | 中 | SV005, SV006, SV007, SV008 |
| CV002 | Xaira publicly disclosed more than $1B of committed capital, but no public source disclosed post-money valuation, share price, dilution, or liquidation preferences. | 高 | SV011, SV033 |
| CV003 | The public record shows unusually strong inputs and real scientific traction, but still little measurable commercial or clinical output to price confidently. | 高 | SV013, SV014, SV017, SV018, SV019, SV020, SV029 |
| CV004 | Because price and terms are undisclosed, the correct public-evidence stance is research-more and price-sensitive rather than a clean invest decision. | 中 | SV005, SV006, SV011, SV017, SV029 |
| CV005 | Public evidence supports some premium to public techbio comps, but not blind acceptance of a frontier-AI-style valuation. | 中 | SV005, SV006, SV007, SV008, SV009, SV011 |
| CV006 | The bull thesis begins with unusual inputs: >$1B capital, elite team density, and an integrated AI/data/therapeutics architecture. | 高 | SV011, SV012, SV022, SV033 |
| CV007 | Orion and X-Cell give Xaira more public scientific proof than a typical stealth techbio startup has. | 高 | SV013, SV014, SV017, SV018, SV019, SV020 |
| CV008 | The market backdrop is real but not automatically monetizable: AI drug discovery is growing, broader R&D needs productivity gains, and the sector is still crowded and proof-hungry. | 高 | SV030, SV031, SV032 |
| CV009 | Open-science traction strengthens Xaira's top-of-funnel credibility but does not yet constitute monetized adoption. | 高 | SV017, SV018, SV019, SV020 |
| CV010 | No retained public source discloses named paying customers, pricing, or deployment metrics for Xaira. | 中 | SV015, SV017, SV018, SV019, SV020 |
| CV011 | Public evidence still stops short of proving translation from Xaira's platform into patient outcomes or asset-level value creation. | 中 | SV016, SV028, SV029 |
| CV012 | The core anti-thesis is that investors may be asked to price aspiration and team quality more than measurable economics. | 中 | SV005, SV010, SV011, SV017, SV029 |
| CV013 | Recursion is the most relevant public full-stack AI-biotech comparable and still trades around $1.73B market cap. | 高 | SV001, SV005, SV026 |
| CV014 | Recursion's more mature platform, partnerships, and pipeline still have not translated into a simple premium multiple, which is a cautionary signal for Xaira. | 中 | SV001, SV026, SV027 |
| CV015 | Relay shows that a well-capitalized pre-revenue therapeutic platform can still be valued around $2.46B in public markets. | 高 | SV004, SV008 |
| CV016 | Schrödinger shows that even a computational-platform-plus-therapeutics model with real commercialization can trade near $0.95B market cap. | 中 | SV002, SV006, SV024 |
| CV017 | Absci shows that a generative-AI biologics platform can still trade around $0.90B market cap despite platform ambition and partnered programs. | 中 | SV003, SV007, SV025 |
| CV018 | Isomorphic's $600M external round proves private appetite for frontier AI drug design, but without disclosed valuation it is a premium signal rather than a pricing anchor. | 中 | SV009 |
| CV019 | The comp set anchors Xaira's downside and reference value far below frontier-AI software narratives. | 中 | SV005, SV006, SV007, SV008, SV009 |
| CV020 | Xaira may still deserve a premium to Absci or Schrödinger because it starts with more capital and a denser elite-team narrative. | 中 | SV006, SV007, SV009, SV011, SV022, SV033 |
| CV021 | Bear case: if Xaira shows no named collaboration or asset proof, value drifts toward roughly $1.5B-$2.5B. | 中 | SV005, SV006, SV007, SV008, SV017, SV018, SV029 |
| CV022 | Base case: if Xaira shows one serious counterparty or a clear internal asset proof point, value moves toward roughly $3B-$5B. | 中 | SV009, SV011, SV013, SV014, SV017, SV022 |
| CV023 | Bull case: if Xaira shows multiple proof points and private-market appetite holds, value can reach roughly $6B-$9B. | 中 | SV009, SV011, SV012, SV013, SV014, SV022 |
| CV024 | Because the actual entry price is not public, valuation work should be expressed as reference ranges rather than return math to a specific entry. | 中 | SV002, SV010, SV011 |
| CV025 | A $1.5B-$2.5B bear range is consistent with public-comp territory when proof is limited. | 中 | SV005, SV006, SV007, SV008 |
| CV026 | A $3B-$5B base range assumes Xaira earns a premium for capital, team, and first proof points. | 中 | SV005, SV006, SV009, SV011, SV022 |
| CV027 | A $6B-$9B bull range requires meaningful proof plus continued willingness by private investors to pay for frontier AI biology optionality. | 中 | SV009, SV011, SV012, SV014, SV022 |
| CV028 | Any valuation materially above the bull range would require non-public evidence around contracts, pipeline, or terms that is not visible in retained public sources. | 中 | SV011, SV017, SV018, SV019, SV020, SV028, SV029 |
| CV029 | Xaira is not IPO-ready on public evidence. | 中 | SV002, SV011, SV021, SV023, SV028 |
| CV030 | Nearer-term value realization is more likely through collaborations, partner-backed assets, or strategic M&A than a clean near-term IPO. | 中 | SV009, SV011, SV014, SV015, SV024 |
| CV031 | The most logical counterparties are large pharma, strategic techbio, or AI platform owners seeking integrated biology/data/model capabilities. | 中 | SV009, SV011, SV012, SV014, SV015 |
| CV032 | Final diligence should prioritize price and term sheet, cap table/preferences, burn, collaboration pipeline, translation metrics, and compliance/security artifacts. | 高 | SV010, SV021, SV028, SV029 |
| CV033 | Unknown cap-table and preference terms prevent rigorous common-equity return modeling. | 中 | SV002, SV011, SV033 |
| CV034 | Unknown burn allocation and milestone budgeting prevent high confidence that >$1B is enough for the proof timetable implied by the story. | 中 | SV010, SV016, SV023, SV033 |
| CV035 | Unknown partner pipeline, pricing, and contract structures prevent revenue underwriting. | 中 | SV015, SV017, SV018, SV019, SV020 |
| CV036 | Unknown platform-to-asset translation metrics prevent valuation based on therapeutics productivity instead of narrative. | 中 | SV013, SV014, SV016, SV029 |
| CV037 | Evidence quality is strong on team and science, but weak on economics and terms. | 中 | SV011, SV017, SV021, SV022, SV029 |
| CV038 | Recommendation confidence is medium because public comps and risk evidence set guardrails but do not set the actual round price. | 中 | SV005, SV006, SV007, SV008, SV010, SV028, SV029 |
| CV039 | Risk rating should remain high because Xaira combines platform complexity, preclinical science, pricing opacity, and commercialization uncertainty. | 中 | SV011, SV017, SV021, SV028, SV029 |
| CV040 | Upgrade triggers are named counterparty proof, measurable asset translation, clearer compliance readiness, and rational price disclosure. | 中 | SV017, SV021, SV028, SV029, SV033 |
| CV041 | Downgrade triggers are aggressive pricing without proof, continued absence of counterparties, or evidence that open-science attention is not becoming strategic leverage. | 中 | SV005, SV006, SV017, SV018, SV019, SV020, SV029 |
| CV042 | Xaira deserves more credit than a typical early public techbio because of its financing scale and team density. | 高 | SV009, SV011, SV022, SV033 |
| CV043 | The public record still supports research-more rather than invest because the market is being asked to price potential, not measurable economics. | 中 | SV001, SV005, SV010, SV017, SV029 |
| CV044 | Final verdict: continue diligence only if pricing lands near the public-comp cluster plus a rational premium; otherwise pass until proof catches up. | 中 | SV005, SV006, SV007, SV008, SV011, SV017, SV029 |
| 编号 | 出版方 | 标题 | 引文 |
|---|---|---|---|
| SO001 | Xaira Therapeutics | Xaira Therapeutics homepage | We are pioneering the transformative artificial intelligence that will help discover and develop the next generation of life-changing medicines. |
| SO002 | Xaira Therapeutics | Our Approach | Xaira Therapeutics | Xaira has three core elements: advanced AI research, expansive data generation and robust therapeutic product development. |
| SO003 | Xaira Therapeutics | Our Team | Xaira Therapeutics | Leadership ... Board of Directors ... Scientific Advisory Board. |
| SO004 | Xaira Therapeutics | Marc Tessier-Lavigne Bio | Xaira Therapeutics | Marc Tessier-Lavigne is co-founder, Chairman & CEO of Xaira Therapeutics. |
| SO005 | Xaira Therapeutics | Hetu Kamichetty Bio | Xaira Therapeutics | Hetu Kamichetty is a co-founder and CTO of Xaira and has played a pivotal role in scaling the firm since its inception in 2023. |
| SO006 | Xaira Therapeutics | Jeff Jonker Bio | Xaira Therapeutics | Jeff serves as the President & Chief Operating Officer at Xaira. |
| SO007 | Xaira Therapeutics | Debbie Law Bio | Xaira Therapeutics | Debbie Law currently serves as CSO of Xaira. |
| SO008 | Xaira Therapeutics | David Baker Bio | Xaira Therapeutics | David Baker, a co-founder of Xaira Therapeutics ... is a recipient of numerous awards, including the 2024 Nobel Prize in Chemistry. |
| SO009 | Xaira Therapeutics | Rachel Lane Bio | Xaira Therapeutics | Rachel Lane PhD is the Senior Vice President of Business Development and Operations. |
| SO010 | Xaira Therapeutics | Work With Us | Xaira Therapeutics | Xaira has offices in South San Francisco, Seattle and London. |
| SO011 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Launches to Deliver Transformative Medicines by Advancing and Harnessing AI for Drug Discovery and Development | Xaira launched with more than $1 billion of committed capital from lead investors ARCH Venture Partners and Foresite Capital. |
| SO012 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Appoints Dr. Debbie Law as Chief Scientific Officer and Julia Tran as Chief People Officer | Since our launch in April, we have made important progress towards our ambitious goals. |
| SO013 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Announces the Appointment of Dr. Paulo Fontoura as Chief Medical Officer and Dr. Hetu Kamisetty as Chief Technology Officer | Additionally, the company will be moving its headquarters to the Gateway of Pacific III campus, a BioMed Realty building, in South San Francisco. |
| SO014 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Announces the Appointment of Bo Wang as SVP and Head of Biomedical AI | Dr. Wang will lead the company’s efforts to develop AI-driven models to help elucidate the molecular basis of poorly treated diseases and to match novel treatments to patients most likely to respond. |
| SO015 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Announces the Appointment of Jeff Jonker as President and Chief Operating Officer | Jeff Jonker ... will help scale the organization and integrate cutting-edge machine learning with therapeutic development. |
| SO016 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Announces the Appointment of Rachel Lane, Ph.D., as Senior Vice President, Business Development and Operations | Rachel Lane ... will oversee the business development strategy ... and drive partnerships to integrate cutting-edge machine learning with therapeutic development. |
| SO017 | TechCrunch | Xaira, an AI drug discovery startup, launches with a massive $1B, says it’s ready to start developing drugs | The company declined to say when it expects to have its first drug available for human trials. |
| SO018 | GeekWire | Inside the Seattle labs of Xaira, the AI-powered startup launched with $1B from investors | Most of Xaira’s 80 employees work from its headquarters in the Bay Area, with a handful in London and 15 people in Seattle. |
| SO019 | Goldman Sachs | How AI is Driving Drug Discovery: Xaira Therapeutics’ Marc Tessier Lavigne | Xaira Therapeutics is harnessing AI to fundamentally change how we discover and develop medicines, shortening the path from lab to clinic for previously "un-druggable" targets. |
| SO020 | Endpoints News | Exclusive: In $1B+ bet on AI, biopharma heavyweights back new startup to upend drug R&D | The company, which has about 50 employees today at sites in Seattle and California, was co-founded by two of biotech’s biggest venture capitalists, Bob Nelsen of ARCH Venture Partners and Vik Bajaj at Foresite Labs. |
| SO021 | Endpoints News | Why VC legend Bob Nelsen is making the biggest initial bet of his 37-year career on Xaira | ARCH will contribute over $200 million, Nelsen said. |
| SO022 | pharmaphorum | Enter Xaira, with $1bn for its AI in drug discovery platform | The San Francisco-based company has emerged ... with more than $1 billion in funding. |
| SO023 | Drug Discovery & Development | How scGPT pioneer Bo Wang, Ph.D. and Xaira’s $1B+ war chest aim to build a virtual cell | My answer was simple: I want to build the first virtual cell in the world. |
| SO024 | KQED / Associated Press | Stanford University President to Resign After Concerns About His Research | The review ... did find that Tessier-Lavigne did not work hard enough to get some of the problematic papers retracted. |
| SO025 | Retraction Watch | Stanford president retracts two Science papers following investigation | Marc Tessier-Lavigne ... is retracting two papers from Science following an institutional investigation that found data manipulation in multiple figures. |
| SO026 | Nature | Designed endocytosis-inducing proteins degrade targets and amplify signals | Designed endocytosis-inducing proteins degrade targets and amplify signals. |
| SO027 | Business Wire / Xaira Therapeutics | X-Atlas/Orion: Xaira Therapeutics Unveils Largest Publicly Available Genome-Wide Perturb-seq Dataset to Power Next-Generation AI for Biology | X-Atlas/Orion ... profiles over 8 million single cells. |
| SO028 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Launches X-Cell, Its First Virtual Cell Model, Trained on the Largest-Ever Genome-Wide Perturbation Dataset, X-Atlas/Pisces | X-Cell ... reaches 4.9 billion parameters, the largest causal perturbation model built to date. |
| SO029 | Fierce Biotech | Xaira exec divulges R&D focus, how $1B fundraise fuels AI-driven hunt for what the industry is hungriest for | We are actively working on building a pipeline. Part of what makes us a next-gen biopharma company is that the AI platform came first and then the pipeline that it generates will come second. |
| SM001 | Xaira Therapeutics | Xaira Therapeutics homepage | We are pioneering the transformative artificial intelligence that will help discover and develop the next generation of life-changing medicines. |
| SM002 | Xaira Therapeutics | Our Approach | Xaira Therapeutics | Xaira has three core elements: advanced AI research, expansive data generation and robust therapeutic product development. |
| SM003 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Launches X-Cell, Its First Virtual Cell Model, Trained on the Largest-Ever Genome-Wide Perturbation Dataset, X-Atlas/Pisces | X-Cell should become increasingly useful for multiple purposes in drug discovery, including target identification, mechanism of action identification, matching targets to patients, and toxicity predictions. |
| SM004 | Fierce Biotech | Xaira exec divulges R&D focus, how AI company is chasing what the industry is hungriest for | We are actively working on building a pipeline ... the AI platform came first and then the pipeline that it generates will come second. |
| SM005 | Endpoints News | In biggest-ever bet on using AI to design drugs, biotech heavyweights launch Xaira with $1B in backing | AI is going to transform every step of the drug discovery process ... you could get two-, three-fold improvements in speed and success rates. |
| SM006 | GeekWire | Inside the Seattle labs of Xaira, the AI-powered startup launched with $1B from investors | Biologics like protein-based therapeutics accounted for a third of drug approvals in 2022. |
| SM007 | Drug Discovery & Development | How scGPT pioneer Bo Wang, Ph.D. and Xaira’s $1B+ war chest aim to build a virtual cell | We want to design a wet lab to help not just test hypotheses, but to generate informative data that improves the model performance. |
| SM008 | GEN Edge | Xaira’s First Virtual Cell Model Is Largest to Date Toward Complex Biology | Target identification to drug approval takes an average of thirteen years while 90% of molecules fail at the clinic. |
| SM009 | Mordor Intelligence | Artificial Intelligence In Drug Discovery Market Size and Share | The Artificial Intelligence In Drug Discovery Market size was valued at USD 2.58 billion in 2025 and is estimated to grow from USD 3.25 billion in 2026 to reach USD 10.29 billion by 2031. |
| SM010 | Worldmetrics | AI Drug Discovery Statistics | 2026 Sourced Report | AI in drug discovery market expected to grow from $2.3 billion in 2023 to $6.2 billion by 2028 at a CAGR of 21.9%. |
| SM011 | McKinsey & Company | How pharma is rewriting the AI playbook: Perspectives from industry leaders | In the pharma industry alone, the AI market is projected to grow from more than $4 billion this year to a whopping $25.7 billion by 2030. Amid this surge, medicine makers have yet to see substantially shorter development timelines or improvements in preclinical or clinical success rates. |
| SM012 | McKinsey & Company | Making more medicines that matter | We have observed, for instance, companies creating such learning loops when moving molecules from lead identification to investigational-new-drug submission nine months faster. |
| SM013 | Deloitte Insights | Scaling up AI across the life sciences value chain | More than 60% of life sciences companies spent over US$20 million on AI initiatives in 2019. |
| SM014 | Deloitte Insights | Modernizing biopharma R&D labs is important for improving research productivity and ensuring the sustainable replenishment of drug pipelines | 53% of respondents reported increased laboratory throughput, while 45% saw a reduction in human error, 30% achieved greater cost efficiencies, and 27% noted faster therapy discovery. |
| SM015 | IQVIA Institute | Global R&D Trends 2026 | Although end-to-end clinical development timelines have increased ... artificial intelligence increasingly enabled R&D, manifesting in increased success rates among AI-driven programs. |
| SM016 | Accenture | From billions to millions: transforming pharma R&D productivity and costs | Depending on the therapeutic area, treatment modality and disease complexity, the cost of bringing a new treatment to market is between $2.6B and $6.7B. |
| SM017 | L.E.K. Consulting | Redefining Biopharma R&D Productivity: New Insights and Strategies | R&D productivity stands as one of the most critical issues for biopharma executives, as it directly addresses the ability to transform pipeline investments into tangible revenue streams. |
| SM018 | ACS Omega | AI-Driven Drug Discovery: A Comprehensive Review | The traditional drug discovery process is complex, costly, and time-consuming, often spanning over a decade ... only approximately 10% of drugs that enter clinical trials ultimately achieve regulatory approval. |
| SM019 | Precedence Research | Inflammatory Disease Market Size, Share and Trends 2026 to 2035 | The global inflammatory disease market ... is predicted to increase from USD 133.50 billion in 2026 to approximately USD 241.34 billion by 2035. |
| SM020 | Fortune Business Insights | Immunology Market | The market is projected to grow from USD 123.05 billion in 2026 to USD 228.18 billion by 2034 ... the monoclonal antibody (mAb) segment is projected to dominate the market with a share of 65.02% in 2026. |
| SM021 | Global Market Insights | Anti-inflammatory Drugs Market Size | The global anti-inflammatory drugs market was valued at USD 132.1 billion in 2025. The market is expected to grow from USD 141.3 billion in 2026 to USD 293.4 billion in 2035. |
| SM022 | Coherent Market Insights | Immunology Market Size, Share, Trends & Forecast, 2026–2033 | Global immunology market is estimated to be valued at USD 122.16 Bn in 2026 and is expected to reach USD 280.35 Bn by 2033. |
| SM023 | Precedence Research | Antibody Production Market Size, Share, and Trends 2026 to 2035 | The global antibody production market size ... is projected to be worth USD 31.71 billion by 2026 ... By End-use, the pharmaceutical and biotechnology companies segment captured more than 56% of revenue share in 2025. |
| SM024 | Coherent Market Insights | Antibodies Market Size, Share, Trends & Forecast, 2026–2033 | Antibodies Market is estimated to be valued at USD 3,23,043.7 Mn in 2026 and is expected to reach USD 7,64,714.8 Mn in 2033. |
| SM025 | Business Wire / Xaira Therapeutics | X-Atlas/Orion: Xaira Therapeutics Unveils Largest Publicly Available Genome-Wide Perturb-seq Dataset to Power Next-Generation AI for Biology | X-Atlas/Orion ... profiles over 8 million single cells. |
| SM026 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Launches to Deliver Transformative Medicines by Advancing and Harnessing AI for Drug Discovery and Development | Xaira launched with more than $1 billion of committed capital ... to bring together leading talent across machine learning, data generation, and integrated drug discovery and development. |
| SP001 | Xaira Therapeutics | Our Approach | Xaira Therapeutics | Xaira has three core elements: advanced AI research, expansive data generation and robust therapeutic product development. |
| SP002 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Launches X-Cell, Its First Virtual Cell Model, Trained on the Largest-Ever Genome-Wide Perturbation Dataset, X-Atlas/Pisces | X-Cell is trained on X-Atlas/Pisces ... 25.6 million perturbed single-cell transcriptomes ... The model reaches 4.9 billion parameters. |
| SP003 | Fierce Biotech | Xaira exec divulges R&D focus, how AI company is chasing what the industry is hungriest for | We are actively working on building a pipeline ... the AI platform came first and then the pipeline that it generates will come second ... we are working on building antibody therapeutics. |
| SP004 | Generate:Biomedicines | Generate:Biomedicines homepage | 42,000 proteins generated, built, and tested ... 140k+ square feet of space in our new Boynton Yards and Andover locations. |
| SP005 | Generate:Biomedicines | Generative Biology | Generate:Biomedicines | We generate custom protein therapeutics—from short peptides to complex antibodies, enzymes, gene therapies, and yet-to-be-described protein compositions. |
| SP006 | Generate:Biomedicines | Generate:Biomedicines announces multi-target collaboration with Novartis | Generate will receive a total upfront payment of $65 million in cash from Novartis ... and is also eligible to receive more than $1 billion in performance-based milestone payments, in addition to tiered royalties up to low double-digits. |
| SP007 | Fierce Biotech | Novartis inks $1B biobucks deal with Flagship's Generate:Biomedicines | Amgen inked an agreement worth up to $1.9 billion biobucks ... Generate ... currently has two candidates in the clinic. |
| SP008 | Isomorphic Labs | Isomorphic Labs homepage | Isomorphic Labs is here to advance human health by building on and beyond the Nobel-winning AlphaFold system. |
| SP009 | Isomorphic Labs | The Isomorphic Labs Drug Design Engine unlocks a new frontier | IsoDDE more than doubles the accuracy of AlphaFold 3 ... and outperforms AlphaFold 3 by 2.3x ... on a challenging, novel antibody-antigen test set. |
| SP010 | Isomorphic Labs | Isomorphic Labs announces Novartis collaboration expansion | IsoLabs and Novartis will expand the scope of the initial collaboration, adding up to three additional research programs on the same financial terms as the original agreement. |
| SP011 | pharmaphorum | Isomorphic signs Lilly, Novartis $3bn AI drug hunt | Alphabet's artificial intelligence start-up Isomorphic Labs ... announcing its first pharma partnerships with Eli Lilly and Novartis, worth almost $3 billion ... $37.5 million upfront from Novartis ... The Lilly deal ... includes $45 million upfront and up to $1.7 billion at the back end. |
| SP012 | insitro | insitro homepage | insitro's ML-driven platform integrates in vitro cellular data produced in our labs with human clinical data to help redefine disease. |
| SP013 | insitro | insitro partners with Lilly to build first-in-kind machine learning models to advance small molecule drug discovery | This collaboration expands the relationship between insitro and Lilly, announced in 2024 ... With more than $700 million in capital raised to date, insitro is building a pipeline through platform. |
| SP014 | Business Wire / insitro via Financial Times Markets | insitro and Bristol Myers Squibb expand ALS collaboration | The companies will leverage multiple therapeutic modalities ... insitro received a $10 million milestone payment ... Backed by ~$800M in capital ... including ~$150M in revenue from collaborations with BMS, Lilly, and Gilead. |
| SP015 | BioPharma Dive | The latest deal in AI drug discovery is a twist on the usual big pharma-startup collaboration model, with Insitro licensing technology and Lilly eligible for royalties | The alliance allows Insitro to retain full global rights to all of its research programs, while Lilly will be eligible for payments for reaching certain milestones ... and may also receive royalties. |
| SP016 | Recursion | Recursion homepage | Over the last decade, we have generated and aggregated one of the largest fit-for-purpose proprietary biological and chemical datasets in the world — >50 petabytes. |
| SP017 | Recursion | Partners | Recursion | As part of this agreement, we received an upfront cash payment of $100 million, with the potential to receive up to $5.2 billion in total aggregate milestone payments plus tiered royalties ... Bayer ... up to $1.5 billion plus royalties. |
| SP018 | pharmaphorum | AI biotechs Exscientia and Recursion agree $688m merger | Recursion Pharma has agreed to join with Exscientia in an all-stock transaction valued at $688 million ... Exscientia shareholders ... will end up owning around 26% of the combined company. |
| SP019 | BioPharma Dive | Recursion cuts pipeline programs after earnings report | AI drug discovery specialist Recursion Pharmaceuticals is shelving three of its most advanced drug prospects ... The cuts will help extend its financial runway into the middle of 2027. |
| SP020 | GEN Edge | Recursion halts four pipeline programs, sharpening cancer, rare disease focus | The company said it will end efforts to develop three clinical programs and one preclinical program ... Investors ... appeared less optimistic, as Recursion's shares fell nearly 17% Monday. |
| SP021 | Chai Discovery | Chai Discovery homepage | Drug-like antibody design against challenging targets with atomic precision ... With Chai-2, we're moving de novo antibody design past binding. |
| SP022 | Nabla Bio | Nabla Bio homepage | We combine de novo drug design with large-scale, human-relevant testing ... By building and owning the data, AI, and integrated dry/wet-lab systems as one engine. |
| SP023 | Business Wire / Nabla Bio | Nabla Bio signs second Takeda collaboration to advance AI-driven design of protein therapeutics | Nabla Bio will receive double-digit millions in upfront and research cost payments and is eligible to receive success-based payments that may exceed $1 billion in total. |
| SP024 | Business Wire / Nabla Bio | Nabla Bio secures $26M Series A financing and collaborations with AstraZeneca, Bristol Myers Squibb and Takeda | Nabla Bio ... announced the close of a $26 million Series A financing ... and strategic collaborations ... worth more than $550 million in upfront and milestone payments, plus royalties. |
| SP025 | Absci | Absci Pipeline | We're advancing a robust pipeline of internal and partnered programs designed with generative AI ... ABS-201 ... potential Best-in-Class therapeutic in 24 months. |
| SP026 | Absci | Absci Technology | We've built a 77,000+ sq ft wet lab ... Our ACE Assay then screens millions of antibody sequence variants with billions of parameters at >4,000x throughput ... new therapeutic designs in as little as six weeks. |
| SP027 | Schrödinger | Drug Discovery | Schrödinger | Maximize your creativity with the industry-leading computational platform for molecular design, discovery, and collaboration. |
| SP028 | Schrödinger | Therapeutic Pipeline | Schrödinger | Under the terms of the agreement, Schrödinger received an upfront payment and is eligible to receive up to $425 million in discovery, development, and commercial milestone payments, as well as low single- to low double-digit royalties on net sales. |
| SP029 | Schrödinger IR | Schrödinger provides update on progress across the business and outlines 2026 strategic priorities | The Lilly TuneLab platform will be integrated into LiveDesign ... Schrödinger has approximately 800 employees operating from 15 locations globally. |
| SP030 | Pharmaceutical Technology | Recursion has agreed to merge with Exscientia | Recursion and Exscientia shareholders will hold 74% and 26% of the new company respectively ... collectively hold $850m in cash and cash equivalents ... projected to yield annual synergies of $100m. |
| SI001 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Launches to Deliver Transformative Medicines by Advancing and Harnessing AI for Drug Discovery and Development | Xaira launched with more than $1 billion of committed capital ... Xaira brings together three core elements: advanced machine learning research, expansive data generation, and robust therapeutic product development. |
| SI002 | Xaira Therapeutics | Our Approach | Xaira Therapeutics | Xaira has three core elements: advanced AI research, expansive data generation and robust therapeutic product development. |
| SI003 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Launches X-Cell, Its First Virtual Cell Model, Trained on the Largest-Ever Genome-Wide Perturbation Dataset, X-Atlas/Pisces | The company's roadmap calls for continued expansion of X-Atlas into primary cells, iPSC-derived cell types, organoids, and in vivo perturbations. |
| SI004 | Fierce Biotech | Xaira exec divulges R&D focus, how AI company is chasing what the industry is hungriest for | Our plan is to build a completely integrated R&D platform ... That's going to take multiple years and it's going to take a billion dollars—maybe more. |
| SI005 | GeekWire | Inside the Seattle labs of Xaira, the AI-powered startup launched with $1B from investors | Most of Xaira's 80 employees work from its headquarters in the Bay Area, with a handful in London and 15 people in Seattle. |
| SI006 | Drug Discovery & Development | How scGPT pioneer Bo Wang, Ph.D. and Xaira’s $1B+ war chest aim to build a ‘virtual cell’ | Right now, it is well received that there are three key pillars for any AI success nowadays: One is talent, two is compute, three is data ... Xaira hits the three buckets altogether. |
| SI007 | Endpoints News | In biggest-ever bet on using AI to design drugs, biotech heavyweights launch Xaira with $1B in backing | The company, which has about 50 employees today at sites in Seattle and California ... has more than a billion dollars. |
| SI008 | Intelligence360 | Xaira Therapeutics to expand into 73,075 square feet of space in San Francisco, California | Xaira Therapeutics plans to build out 73,075 square feet of new space in San Francisco. |
| SI009 | Nasdaq / Recursion | Recursion Reports Fourth Quarter and Full Year 2025 Financial Results and Provides Business Update | Cash, cash equivalents and restricted cash were $753.9 million ... Total revenue ... was $74.7 million ... Research and development expenses ... were $475.3 million ... runway extends into early 2028. |
| SI010 | Recursion | Partners | Recursion | As part of this agreement, we received an upfront cash payment of $100 million, with the potential to receive up to $5.2 billion in total aggregate milestone payments plus tiered royalties. |
| SI011 | BioPharma Dive | Recursion pipeline cuts after earnings report | The cuts, which involve three of Recursion's most advanced drug programs, are expected to help extend the company's cash runway into the middle of 2027. |
| SI012 | Business Wire / Schrödinger | Schrödinger Reports Fourth Quarter and Full-Year 2025 Financial Results | Total revenue was $255.9 million ... Operating expenses were $309.5 million ... cash, cash equivalents, restricted cash and marketable securities of approximately $402.3 million. |
| SI013 | Schrödinger | Therapeutic Pipeline | Schrödinger | Schrödinger's therapeutics group is working on a number of collaborative drug discovery programs. We are eligible to receive milestones ... and royalties on sales for certain approved products. |
| SI014 | Schrödinger IR | SEC Filings | Schrödinger | |
| SI015 | Nasdaq / Relay Therapeutics | Relay Therapeutics extends cash runway into 2029 amid clinical trial advancements | Approximately $710 million in cash, cash equivalents and investments at end of Q1 2025 ... Revenue was $7.7 million ... R&D Expenses were $73.8 million ... Net loss was $77.1 million. |
| SI016 | StockTitan / Absci | Absci Reports Business Updates and Third Quarter 2025 Financial and Operating Results | Cash, cash equivalents, and marketable securities as of September 30, 2025 were $152.5 million ... Revenue was $0.4 million ... Research and development expenses were $19.2 million ... into the first half of 2028. |
| SI017 | Silicon Valley Bank | Healthcare Industry Trends 2025 Mid-Year Report | The healthcare innovation economy is on track for its worst fundraising year in more than a decade. |
| SI018 | J.P. Morgan | Q1 2026 Biopharma Licensing and Venture Report | Biopharma capital markets opened 2026 with selective momentum ... licensing and M&A remained strong ... Deal structures remained milestone-heavy, while upfront economics stayed strong for the most competitive assets. |
| SI019 | Generate:Biomedicines | Generate:Biomedicines announces multi-target collaboration with Novartis | Generate will receive a total upfront payment of $65 million in cash from Novartis ... and is eligible to receive more than $1 billion in performance-based milestone payments. |
| SI020 | pharmaphorum | Isomorphic signs Lilly, Novartis $3bn AI drug hunt | Isomorphic Labs ... announcing its first pharma partnerships with Eli Lilly and Novartis, worth almost $3 billion ... $45 million upfront ... $37.5 million upfront. |
| SI021 | insitro | insitro partners with Lilly to build first-in-kind machine learning models to advance small molecule drug discovery | With more than $700 million in capital raised to date ... This collaboration expands the relationship between insitro and Lilly, announced in 2024. |
| SI022 | Business Wire / insitro via Financial Times Markets | insitro and Bristol Myers Squibb expand ALS collaboration | Backed by ~$800M in capital ... including ~$150M in revenue from collaborations with BMS, Lilly, and Gilead. |
| SI023 | Business Wire / Nabla Bio | Nabla Bio signs second Takeda collaboration to advance AI-driven design of protein therapeutics | Nabla Bio will receive double-digit millions in upfront and research cost payments and is eligible to receive success-based payments that may exceed $1 billion in total. |
| SI024 | Absci | Absci Technology | We've built a 77,000+ sq ft wet lab ... Our ACE Assay then screens millions of antibody sequence variants ... at >4,000x throughput. |
| SI025 | Xaira Therapeutics | Work With Us | Xaira Therapeutics | We are seeking extraordinary scientists, engineers, operators, and everyone in between. |
| SI026 | Schrödinger IR | Schrödinger provides update on progress across the business and outlines 2026 strategic priorities | Schrödinger has approximately 800 employees operating from 15 locations globally. |
| SE001 | Xaira Therapeutics | Our Approach | Xaira Therapeutics | Xaira has three core elements: advanced AI research, expansive data generation and robust therapeutic product development. |
| SE002 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Launches to Deliver Transformative Medicines by Advancing and Harnessing AI for Drug Discovery and Development | Xaira brings together three core elements: advanced machine learning research, expansive data generation, and robust therapeutic product development. |
| SE003 | Business Wire / Xaira Therapeutics | X-Atlas/Orion: Xaira Therapeutics Unveils Largest Publicly Available Genome-Wide Perturb-seq Dataset to Power Next-Generation AI for Biology | X-Atlas/Orion is the largest publicly available Perturb-seq atlas ... comprises 8 million cells ... FiCS Perturb-seq platform ... leverages the Chromium platform from 10x Genomics. |
| SE004 | bioRxiv | X-Atlas/Orion: Genome-wide Perturb-seq Datasets via a Scalable Fix-Cryopreserve Platform for Training Dose-Dependent Biological Foundation Models | FiCS Perturb-seq exhibits high sensitivity and low batch effects ... X-Atlas/Orion ... comprises eight million cells deeply sequenced to over 16,000 UMIs per cell. |
| SE005 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Launches X-Cell, Its First Virtual Cell Model, Trained on the Largest-Ever Genome-Wide Perturbation Dataset, X-Atlas/Pisces | X-Cell is trained on X-Atlas/Pisces ... 25.6 million perturbed single-cell transcriptomes ... The model reaches 4.9 billion parameters ... roadmap calls for continued expansion of X-Atlas into primary cells, iPSC-derived cell types, organoids, and in vivo perturbations. |
| SE006 | Xaira Therapeutics | News & Content | Xaira Therapeutics | Read news & views about Xaira and our science ... X-Cell ... March 17, 2026 ... X-Atlas/Orion ... June 17, 2025. |
| SE007 | Xaira Therapeutics | Privacy Policy | Xaira Therapeutics | We seek to protect your Personal Data from unauthorized access, use and disclosure using appropriate physical, technical, organizational and administrative security measures. |
| SE008 | Xaira Therapeutics | Work With Us | Xaira Therapeutics | We are seeking extraordinary scientists, engineers, operators, and everyone in between ... Xaira never uses Google Chat for recruitment communications. |
| SE009 | GitHub | GitHub - Xaira-Therapeutics/X-Cell | Status: Model weights and inference code coming soon. The Python API, model weights, and tutorials are under active development. |
| SE010 | GitHub / Xaira Therapeutics | X-Cell README | pip install xcell ... X-Cell Mini ... 55M ... X-Atlas/Pisces is available at Xaira-Therapeutics/X-Atlas-Pisces. |
| SE011 | GitHub / Xaira Therapeutics | X-Cell MODEL_CARD.md | X-Cell is a set-level diffusion transformer ... intended for research use in computational biology and genomics ... model weights and inference code coming soon. |
| SE012 | GitHub / Xaira Therapeutics | X-Cell docs/model.md | X-Cell Mini ... 55M ... Layers 12 ... Attention heads 8 ... Cross-attn layers 4 ... Min GPU VRAM 8 GB (1 GPU). |
| SE013 | GitHub / Xaira Therapeutics | X-Cell docs/quickstart.md | adata should contain log-normalized (log1p CP10k) expression values ... genes not in the vocabulary are zero-imputed. |
| SE014 | GitHub / Xaira Therapeutics | X-Cell docs index | Model weights and inference code are coming soon ... X-Cell achieves Pearson Δ of 0.51 on held-out iPSC perturbations ... over 5× higher than the next-best method. |
| SE015 | Hugging Face | Xaira-Therapeutics/X-Cell · Hugging Face | Status: Model weights and inference code coming soon ... Full documentation: xaira-therapeutics.github.io/X-Cell. |
| SE016 | Hugging Face | Xaira-Therapeutics/X-Atlas-Pisces · Datasets at Hugging Face | Dataset viewer is not available ... (Coming Soon) The following data will be uploaded to this dataset ... Downloads last month 80 ... like 6. |
| SE017 | GEN Edge | Xaira Therapeutics Releases Largest Perturb-Seq Dataset to Power the Virtual Cell | X-Atlas/Orion is comprised of eight million cells ... By releasing X-Atlas/Orion's methods, Xaira aims to allow more labs to generate Perturb-seq data at large, high-quality, and standardized scale. |
| SE018 | GEN Edge | Xaira’s First Virtual Cell Model Is Largest To-Date, Toward Complex Biology | X-Cell ... sizes up to 4.9 billion parameters ... the first scaling law demonstrator in the virtual cell domain ... integrates biological prior knowledge through a cross-attention mechanism. |
| SE019 | Fierce Biotech | Xaira exec divulges R&D focus, how AI company is chasing what the industry is hungriest for | Part of what makes us a next-gen biopharma company is that the AI platform came first and then the pipeline that it generates will come second ... we are working on building antibody therapeutics. |
| SE020 | Drug Discovery and Development | How scGPT pioneer Bo Wang, Ph.D. and Xaira’s $1B+ war chest aim to build a ‘virtual cell’ | AI provides prediction, the wet lab provides validation, and this validation further improves the AI predictions ... we are very excited to work with his team on enhancing some of the AI models for protein design, antibody design. |
| SE021 | GeekWire | Inside the Seattle labs of Xaira, the AI-powered startup launched with $1B from investors | Xaira was founded with the aim of building on IPD models like RFDiffusion and ProteinMPNN ... laboratory tests assess how well the proteins stick ... The data are quickly fed back into the protein models. |
| SE022 | San Francisco Business Times | Xaira Therapeutics raised nearly $1 billion. Here's its next act | The company is hiring for 25 positions. |
| SE023 | Nature | RFdiffusion: De novo protein design using diffusion models | We construct a RF-based diffusion model, RFdiffusion ... broadly applicable for protein design. |
| SE024 | Nature | Designed endocytosis-inducing proteins degrade targets and amplify signals | We reasoned that de novo protein design could enable the creation of bio-orthogonal endocytosis-inducing proteins ... customized for the target receptor. |
| SE025 | BioPharmaTrend | Xaira Therapeutics launches X-Cell, its first virtual cell model | Xaira's roadmap calls for expanding X-Atlas into primary cells, iPSC-derived cell types, organoids, and in vivo perturbations ... a subset of the Pisces dataset and X-Cell model will be made available to the scientific community. |
| SU001 | Xaira Therapeutics | Our Approach | Xaira Therapeutics | Xaira has three core elements: advanced AI research, expansive data generation and robust therapeutic product development. |
| SU002 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Launches to Deliver Transformative Medicines by Advancing and Harnessing AI for Drug Discovery and Development | Xaira brings together three core elements: advanced machine learning research, expansive data generation, and robust therapeutic product development. |
| SU003 | Xaira Therapeutics | News & Content | Xaira Therapeutics | X-Atlas/Orion ... June 17, 2025 ... X-Cell ... March 17, 2026. |
| SU004 | Xaira Therapeutics | Work With Us | Xaira Therapeutics | We are seeking extraordinary scientists, engineers, operators, and everyone in between. |
| SU005 | Business Wire / Xaira Therapeutics | X-Atlas/Orion: Xaira Therapeutics Unveils Largest Publicly Available Genome-Wide Perturb-seq Dataset to Power Next-Generation AI for Biology | X-Atlas/Orion is now publicly available here ... This industrialized platform and the Orion dataset will empower scientists to build more predictive models of complex biology. |
| SU006 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Launches X-Cell, Its First Virtual Cell Model, Trained on the Largest-Ever Genome-Wide Perturbation Dataset, X-Atlas/Pisces | Xaira is making a subset of the Pisces dataset and X-Cell model available to the scientific community. |
| SU007 | Fierce Biotech | Xaira exec divulges R&D focus, how AI company is chasing what the industry is hungriest for | Part of what makes us a next-gen biopharma company is that the AI platform came first and then the pipeline that it generates will come second. |
| SU008 | Drug Discovery and Development | How scGPT pioneer Bo Wang, Ph.D. and Xaira’s $1B+ war chest aim to build a ‘virtual cell’ | Instead of doing expensive wet lab experiments, you can prompt the virtual cell ... Xaira can provide the necessary resources. |
| SU009 | GEN Edge | Xaira Therapeutics Releases Largest Perturb-Seq Dataset to Power the Virtual Cell | 'To build a robust model of any system, perturbation data is critical, and the released X-Atlas/Orion dataset marks a significant contribution to the scientific community,' said Emma Lundberg. |
| SU010 | GEN Edge | Xaira’s First Virtual Cell Model Is Largest To-Date, Toward Complex Biology | While virtual cell models that generalize to new contexts provide a valued advance toward understanding biology, predicting patient outcomes is still a step away. |
| SU011 | R&D World | How Xaira aims to fuel biology’s ‘ImageNet moment’ with a 521-GB open-source dataset for training biological foundation models | The resource has already been downloaded more than 16,451 times at the time of writing, just two weeks after its release ... Xaira is happy to work with any commercial entity who might be interested in collaborating with us. |
| SU012 | Hugging Face | Xaira-Therapeutics/X-Atlas-Orion · Discussions | like 22 ... Community 2 ... sgRNA counts? ... Conversion to Parquet. |
| SU013 | Hugging Face | Xaira-Therapeutics/X-Atlas-Orion · sgRNA counts? | Very interesting work! I'm curious if you also plan to include the sgRNA count data? ... [ann-huang]: we've uploaded the sgRNA count data to figshare ... please check it out there. |
| SU014 | Hugging Face | Xaira-Therapeutics/X-Atlas-Orion · [bot] Conversion to Parquet | The Parquet version of the dataset is available for you to use ... you can use HF Datasets, ClickHouse, DuckDB, Pandas, PostgreSQL, or Polars. |
| SU015 | BioPharmaTrend | Xaira Publishes Largest Public Perturb-seq Atlas to Advance Virtual Cell Modeling | X-Atlas/Orion is now publicly available here ... Xaira's team indicates the dataset could contribute to the training of virtual cell models. |
| SU016 | TMCnet / Business Wire | Xaira Therapeutics Unveils Largest Publicly Available Genome-Wide Perturb-seq Dataset | This industrialized platform and the Orion dataset will empower scientists to build more predictive models of complex biology. |
| SU017 | talk.bio | Xaira Therapeutics Launches X-Cell | A subset of the Pisces dataset and X-Cell model is being made available to the scientific community. |
| SU018 | Hugging Face | Xaira-Therapeutics/X-Atlas-Pisces · Datasets at Hugging Face | (Coming Soon) The following data will be uploaded to this dataset ... Downloads last month 80 ... like 6. |
| SU019 | Hugging Face | Xaira-Therapeutics/X-Cell · Hugging Face | Status: Model weights and inference code coming soon ... intended for research use in computational biology and genomics. |
| SU020 | GitHub | GitHub - Xaira-Therapeutics/X-Cell | Status: Model weights and inference code coming soon. The Python API, model weights, and tutorials are under active development. |
| SU021 | Xaira Therapeutics | Privacy Policy | Xaira Therapeutics | We seek to protect your Personal Data from unauthorized access, use and disclosure using appropriate physical, technical, organizational and administrative security measures. |
| SU022 | San Francisco Business Times | Xaira Therapeutics raised nearly $1 billion. Here's its next act | The company is hiring for 25 positions. |
| SU023 | BioPharmaTrend | Xaira Therapeutics Launches X-Cell, Its First Virtual Cell Model | A subset of the Pisces dataset and X-Cell model will be made available to the scientific community. |
| SU024 | GeekWire | Inside the Seattle labs of Xaira, the AI-powered startup launched with $1B from investors | From molecules, to biology and patients, we are building models and collecting data ... we think of the data, models, and iteration across the entire spectrum. |
| SU025 | GitHub / Xaira Therapeutics | X-Cell docs index | See Quick Start for full examples ... If you use X-Cell or X-Atlas/Pisces in your research, please cite. |
| SU026 | Life Science Washington | Xaira Therapeutics Releases Largest Perturb-Seq Dataset to Power the Virtual Cell | Xaira Therapeutics ... has made a major scientific contribution in its first year. The company released X-Atlas/Orion, the largest publicly available Perturb-seq dataset. |
| SR001 | EUR-Lex | Regulation (EU) 2024/1689 — Artificial Intelligence Act | The purpose of this Regulation is to improve the functioning of the internal market by laying down a uniform legal framework ... while ensuring a high level of protection of health, safety, fundamental rights. |
| SR002 | EUR-Lex | Regulation (EU) 2016/679 — General Data Protection Regulation | Rapid technological developments and globalisation have brought new challenges for the protection of personal data. |
| SR003 | FDA | Artificial Intelligence for Drug Development | AI will undoubtedly play a critical role in the drug development life cycle and CDER plans to continue developing and adopting a risk-based regulatory framework that promotes innovation and protects patient safety. |
| SR004 | FDA | Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products | This guidance provides a risk-based credibility assessment framework that may be used for establishing and evaluating the credibility of an AI model for a particular context of use. |
| SR005 | FDA / EMA | Guiding Principles of Good AI Practice in Drug Development | These 10 guiding principles are intended to lay the foundation for developing good practice that addresses the unique nature of these technologies. |
| SR006 | European Medicines Agency | Reflection paper on the use of Artificial Intelligence (AI) in the medicinal product lifecycle | As these models often contain exceptionally large numbers of trainable parameters arranged in non-transparent model architectures, new risks are introduced that need to be mitigated to ensure the safety of patients and integrity of clinical study results. |
| SR007 | NIST | AI Risk Management Framework | The NIST AI Risk Management Framework is intended for voluntary use and to improve the ability to incorporate trustworthiness considerations into the design, development, use, and evaluation of AI products, services, and systems. |
| SR008 | CISA | Artificial Intelligence | The playbook guides AI providers, developers, and adopters on voluntarily sharing AI-related cybersecurity information with CISA and partners. |
| SR009 | Xaira Therapeutics / GitHub | X-Cell LICENSE | This work is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. |
| SR010 | Creative Commons | Attribution-NonCommercial-ShareAlike 4.0 International | NonCommercial — You may not use the material for commercial purposes. |
| SR011 | 10x Genomics | Chromium Single Cell | Directly link CRISPR guide RNAs to the resulting perturbed phenotypes. |
| SR012 | Xaira Therapeutics | Privacy Policy | Although we work to protect the security of your account and other data that we hold in our records, please be aware that no method of transmitting data over the internet or storing data is completely secure. |
| SR013 | Xaira Therapeutics | Our Approach | Xaira has three core elements: advanced AI research, expansive data generation and robust therapeutic product development. |
| SR014 | Xaira Therapeutics | Our Team | Xaira Leadership ... Marc Tessier-Lavigne ... Paulo Fontoura ... Hetu Kamichetty ... Debbie Law ... Bo Wang ... Scott Gottlieb ... Former FDA Commissioner. |
| SR015 | Xaira Therapeutics | Work With Us | We are seeking extraordinary scientists, engineers, operators, and everyone in between. |
| SR016 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Launches to Deliver Transformative Medicines by Advancing and Harnessing AI for Drug Discovery and Development | Xaira launched with more than $1 billion of committed capital ... Xaira brings together three core elements: advanced machine learning research, expansive data generation, and robust therapeutic product development. |
| SR017 | Business Wire / Xaira Therapeutics | X-Atlas/Orion: Xaira Therapeutics Unveils Largest Publicly Available Genome-Wide Perturb-seq Dataset to Power Next-Generation AI for Biology | Xaira's FiCS Perturb-seq platform, which leverages the Chromium platform from 10x Genomics, delivers the sensitivity, scalability and reproducibility essential for generating high-quality perturbational data. |
| SR018 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Launches X-Cell, Its First Virtual Cell Model, Trained on the Largest-Ever Genome-Wide Perturbation Dataset, X-Atlas/Pisces | Xaira is making a subset of the Pisces dataset and X-Cell model available to the scientific community. |
| SR019 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Announces the Appointment of Dr. Paulo Fontoura as Chief Medical Officer and Dr. Hetu Kamisetty as Chief Technology Officer | These two additions further build out the C-suite for Xaira ... Its new facilities in the Bay Area's biotech hub will support Xaira's continued growth. |
| SR020 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Appoints Dr. Debbie Law as Chief Scientific Officer and Julia Tran as Chief People Officer | Since launch, Xaira has been building AI research capabilities spanning fundamental computational methods development and their application to biological discovery, the design of drug-like matter, and clinical development. |
| SR021 | GeekWire | Inside the Seattle labs of Xaira, the AI-powered startup launched with $1B from investors | From molecules, to biology and patients, we are building models and collecting data ... we think of the data, models, and iteration across the entire spectrum. |
| SR022 | Fierce Biotech | Xaira exec divulges R&D focus, how AI company is chasing what the industry is hungriest for | Part of what makes us a next-gen biopharma company is that the AI platform came first and then the pipeline that it generates will come second. |
| SR023 | Drug Discovery and Development | How scGPT pioneer Bo Wang, Ph.D. and Xaira's $1B+ war chest aim to build a virtual cell | Instead of doing expensive wet lab experiments, you can prompt the virtual cell ... Xaira can provide the necessary resources. |
| SR024 | GEN Edge | Xaira Therapeutics Releases Largest Perturb-Seq Dataset to Power the Virtual Cell | The released X-Atlas/Orion dataset marks a significant contribution to the scientific community. |
| SR025 | GEN Edge | Xaira's First Virtual Cell Model Is Largest To-Date, Toward Complex Biology | While virtual cell models that generalize to new contexts provide a valued advance toward understanding biology, predicting patient outcomes is still a step away. |
| SR026 | R&D World | How Xaira aims to fuel biology's ImageNet moment with a 521-GB open-source dataset for training biological foundation models | The resource has already been downloaded more than 16,451 times ... Xaira is happy to work with any commercial entity who might be interested in collaborating with us. |
| SR027 | Hugging Face | Xaira-Therapeutics/X-Atlas-Orion · Discussions | like 22 ... Community 2 ... sgRNA counts? ... Conversion to Parquet. |
| SR028 | Hugging Face | Xaira-Therapeutics/X-Atlas-Orion · sgRNA counts? | Very interesting work! I'm curious if you also plan to include the sgRNA count data? ... we've uploaded the sgRNA count data to figshare. |
| SR029 | Hugging Face | Xaira-Therapeutics/X-Atlas-Pisces | (Coming Soon) The following data will be uploaded to this dataset ... Downloads last month 80 ... like 6. |
| SR030 | Hugging Face | Xaira-Therapeutics/X-Cell | Status: Model weights and inference code coming soon ... intended for research use in computational biology and genomics. |
| SR031 | GitHub | GitHub - Xaira-Therapeutics/X-Cell | Status: Model weights and inference code coming soon. The Python API, model weights, and tutorials are under active development. |
| SR032 | GitHub / Xaira Therapeutics | X-Cell docs index | See Quick Start for full examples ... If you use X-Cell or X-Atlas/Pisces in your research, please cite. |
| SR033 | J.P. Morgan | Q1 2026 Biopharma Licensing and Venture Report | Biopharma financing and transaction activity in Q1 2026 continued to reflect a selective capital environment, with investors and acquirers concentrating around later-stage assets, differentiated science and programs with clearer clinical and commercial pathways. |
| SR034 | San Francisco Business Times | Xaira Therapeutics raised nearly $1 billion. Here's its next act | The company is hiring for 25 positions. |
| SR035 | Xaira Therapeutics | David Baker Bio | He has published over 640 research papers, co-founded 21 companies, and been awarded more than 100 patents. |
| SV001 | SEC / Recursion Pharmaceuticals | Recursion Pharmaceuticals 2025 Form 10-K | We do not have any products approved for commercial sale and have not generated any revenues from product sales. Cash, cash equivalents and restricted cash totaled $753.9 million as of December 31, 2025. |
| SV002 | SEC / Schrödinger | Schrödinger 2025 Form 10-K | As of June 30, 2025 ... the aggregate market value of the voting and non-voting common equity held by non-affiliates of the registrant was $1,124,429,157. |
| SV003 | SEC / Absci | Absci 2025 Form 10-K | Revenue was $2.8 million for the year ended December 31, 2025 ... We incurred a net loss of $115.2 million for the year ended December 31, 2025. |
| SV004 | SEC / Relay Therapeutics | Relay Therapeutics 2025 Form 10-K | We had cash, cash equivalents, and investments of $554.5 million as of December 31, 2025 ... We believe our existing cash ... will enable us to fund our operating expenses and capital expenditure requirements into 2029. |
| SV005 | CompaniesMarketCap | Recursion Pharmaceuticals market cap | As of May 2026 Recursion Pharmaceuticals has a market cap of $1.73 Billion USD. |
| SV006 | CompaniesMarketCap | Schrödinger market cap | As of May 2026 Schrödinger has a market cap of $0.95 Billion USD. |
| SV007 | CompaniesMarketCap | Absci market cap | As of May 2026 Absci has a market cap of $0.90 Billion USD. |
| SV008 | CompaniesMarketCap | Relay Therapeutics market cap | As of May 2026 Relay Therapeutics has a market cap of $2.46 Billion USD. |
| SV009 | Isomorphic Labs | Isomorphic Labs announces $600m external investment round | Isomorphic Labs announces it has raised $600 Million in its first external funding round. |
| SV010 | J.P. Morgan | Q1 2026 Biopharma Licensing and Venture Report | Biopharma financing and transaction activity in Q1 2026 continued to reflect a selective capital environment, with investors and acquirers concentrating around later-stage assets, differentiated science and programs with clearer clinical and commercial pathways. |
| SV011 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Launches to Deliver Transformative Medicines by Advancing and Harnessing AI for Drug Discovery and Development | Xaira launched with more than $1 billion of committed capital ... Xaira brings together three core elements: advanced machine learning research, expansive data generation, and robust therapeutic product development. |
| SV012 | Xaira Therapeutics | Our Approach | Xaira has three core elements: advanced AI research, expansive data generation and robust therapeutic product development. |
| SV013 | Business Wire / Xaira Therapeutics | X-Atlas/Orion: Xaira Therapeutics Unveils Largest Publicly Available Genome-Wide Perturb-seq Dataset to Power Next-Generation AI for Biology | This industrialized platform and the Orion dataset will empower scientists to build more predictive models of complex biology. |
| SV014 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Launches X-Cell, Its First Virtual Cell Model | Xaira is making a subset of the Pisces dataset and X-Cell model available to the scientific community. |
| SV015 | Fierce Biotech | Xaira exec divulges R&D focus, how AI company is chasing what the industry is hungriest for | Part of what makes us a next-gen biopharma company is that the AI platform came first and then the pipeline that it generates will come second. |
| SV016 | Drug Discovery and Development | How scGPT pioneer Bo Wang, Ph.D. and Xaira's $1B+ war chest aim to build a virtual cell | Instead of doing expensive wet lab experiments, you can prompt the virtual cell ... Xaira can provide the necessary resources. |
| SV017 | R&D World | How Xaira aims to fuel biology's ImageNet moment with a 521-GB open-source dataset for training biological foundation models | The resource has already been downloaded more than 16,451 times ... Xaira is happy to work with any commercial entity who might be interested in collaborating with us. |
| SV018 | Hugging Face | Xaira-Therapeutics/X-Atlas-Orion · Discussions | like 22 ... Community 2 ... sgRNA counts? ... Conversion to Parquet. |
| SV019 | Hugging Face | Xaira-Therapeutics/X-Cell | Status: Model weights and inference code coming soon ... intended for research use in computational biology and genomics. |
| SV020 | GitHub | GitHub - Xaira-Therapeutics/X-Cell | Status: Model weights and inference code coming soon. The Python API, model weights, and tutorials are under active development. |
| SV021 | Xaira Therapeutics | Privacy Policy | Although we work to protect the security of your account and other data that we hold in our records, please be aware that no method of transmitting data over the internet or storing data is completely secure. |
| SV022 | Xaira Therapeutics | Our Team | Xaira Leadership ... Marc Tessier-Lavigne ... Paulo Fontoura ... Debbie Law ... Bo Wang ... Scott Gottlieb ... Former FDA Commissioner. |
| SV023 | Xaira Therapeutics | Work With Us | We are seeking extraordinary scientists, engineers, operators, and everyone in between. |
| SV024 | Schrödinger | Schrödinger Provides Update on Progress Across the Business and Outlines 2026 Strategic Priorities | We are entering 2026 with a clear mandate: to further strengthen our position as the essential design engine for the industry. |
| SV025 | StockTitan Argus / Absci | Absci Reports Business Updates and Third Quarter 2025 Financial and Operating Results | Cash balance of $152.5M as of Sept 30, 2025 ... revenue $0.4M ... market cap to $508.38M at that time. |
| SV026 | MarketBeat | Recursion Pharmaceuticals details AI-driven drug pipeline, Sanofi/Roche milestones, runway to 2028 | Taylor also noted that Recursion has brought in over $500 million from partners ... and said the company ended the year with $754 million in cash, which he said provides runway into early 2028. |
| SV027 | BioPharma Dive | Recursion shelves three drug programs to cut costs after Exscientia merger | The company hasn't yet fulfilled its promise ... pipeline cuts were inevitable given the company's unsustainable cash burn. |
| SV028 | European Medicines Agency | Reflection paper on the use of Artificial Intelligence (AI) in the medicinal product lifecycle | New risks are introduced that need to be mitigated to ensure the safety of patients and integrity of clinical study results. |
| SV029 | GEN Edge | Xaira's First Virtual Cell Model Is Largest To-Date, Toward Complex Biology | While virtual cell models that generalize to new contexts provide a valued advance toward understanding biology, predicting patient outcomes is still a step away. |
| SV030 | Mordor Intelligence | Artificial Intelligence in Drug Discovery Market | The Artificial Intelligence In Drug Discovery Market size was valued at USD 2.58 billion in 2025 and is estimated to grow from USD 3.25 billion in 2026 to reach USD 10.29 billion by 2031. |
| SV031 | IQVIA Institute | Global R&D Trends 2026 | Growing scientific complexity, longer development timelines, and persistent regional disparities ... have put pressure on productivity. |
| SV032 | BioMed Nexus | 25 AI Drug Discovery Companies Actually Delivering Clinical Candidates | Most of them are pre-revenue platform companies with no clinical assets. Some are rebranding basic computational chemistry as AI to attract funding. |
| SV033 | Business Wire / Xaira Therapeutics | Xaira Therapeutics Appoints Dr. Debbie Law as Chief Scientific Officer and Julia Tran as Chief People Officer | Xaira launched with more than $1 billion of committed capital ... Since launch, Xaira has been building AI research capabilities spanning fundamental computational methods development and their application to biological discovery, the design of drug-like matter, and clinical development. |