OpenAI
使命治理的前沿 AI 平台与应用公司
OpenAI 的体量已成气候且业务仍在高速增长,但 $852B 的入局估值依然偏高,主因是收入持续性与自由现金流尚未完全证明。
覆盖范围与披露说明
本次更新将估值、企业级规模、产品广度以及风险证据推进至 2026 年 9 月的报告基准日,同时保留公司受使命治理的核心框架。
封面要素
公司概况
OpenAI 是一家受基金会控制的前沿 AI 公司,业务横跨个人对话、企业工作流、开发者工具以及安全与访问控制产品。公司拥有极高的大众触达度与强劲的收入增长势头,但底层经济模型仍高度依赖资本投入,治理架构也很特殊;尽调既要盯紧使命控制权,也要看清商业变现质量。
- 成立时间
- 2015-12-08
- 创始人
- Sam Altman, Greg Brockman
- 创立地点
- San Francisco, California, USA
- 总部
- San Francisco, California
- 产品
- ChatGPT、Codex、API 接入、企业工作流,以及围绕前沿模型的网络安全与访问控制。
- 客户
- 个人消费者、开发者、企业客户,以及强监管行业的垂直采购方。
- 商业模式
- 订阅制、按用量计费的 API 收入、企业级合同,以及垂直行业与工作流打包方案。
- 阶段
- late-stage private
- 融资情况
- 2026 年 3 月以 $852B 估值完成 $122B 融资,并扩大了 $4.7B 的循环信贷额度;公开的首次公开募股(IPO)讨论十分活跃。
执行摘要
主要优势
- 消费者、企业与开发者全渠道产品触达极强。
- 收入体量已足够撑起二级市场叙事。
- 使命导向的治理架构与战略资本获取能力仍属罕见优势。
主要风险
- 算力强度与烧钱速度对照当前毛利率依然过高。
- 治理、网络安全与系统可靠性已成为核心尽调风险。
- 具体的融资结构与自由现金流跑通路径仍不透明。
未决问题
- 经审计的分部收入、毛利率、烧钱速度与现金储备均未公开。
- 2026 年具体的融资优先股堆叠、清算权益、对赌棘轮及二级市场老股交易机制均未公开。
- 基础设施合作伙伴之间的承诺支出、预付款及照付不议调度表均未披露。
- 客户留存与持续性指标(包括 NRR、流失率、合同周期与大客户集中度)依然缺失。
目录
01公司概况
1.1 身份、使命与组织架构
OpenAI 将自身定位为 AI 研发与部署公司,其使命是确保通用人工智能造福全人类。在风投支持的初创企业中,这种组织架构极为罕见:非营利实体 OpenAI Foundation 控制着 OpenAI Group PBC;设立营利实体旨在推进使命,同时撬动规模扩张与资本募集。该双重架构是本报告后续所有尽调问题的核心锚点。这种独特性足以重塑采购方、合作伙伴或投资方对控制权、知情权及使命偏离风险的评估逻辑。[CO001, CO002, CO003, CO004, CO005]
1.2 领导层、治理架构与业务规模
董事会与高管层权力依然高度集中。Sam Altman 仍是核心的对外公众高管;基金会任命的董事会成员由少数资深运营者、投资人及安全与治理专家组成。公司公开可见的规模十分庞大:OpenAI 称其周活跃用户数超过 10 亿、企业客户超过 100 万家、ChatGPT 工作场所席位数突破 700 万。如此体量足以支撑赛道龙头逻辑,但也意味着哪怕轻微的执行失误,如今都会演变成企业级和大众级的重大事件。公开披露的投资者阵容同样十分广泛,涵盖 Microsoft、SoftBank、Amazon 以及 NVIDIA,这既让公司资金实力异常雄厚,也使其极度依赖少数战略合作关系。公开披露的投资者阵容同样十分广泛,涵盖 Microsoft、SoftBank、Amazon 以及 NVIDIA,这既让公司资金实力异常雄厚,也使其极度依赖少数战略合作关系。这也意味着外部尽调必须聚焦集中度风险,毕竟少数几个人依然在战略、融资与信任维度上承担着过大的责任。[CO006, CO008, CO009, CO010, CO013]
| 指标 | 数值/状态 | 日期 | 置信度 | 信息缺口 |
|---|---|---|---|---|
| 成立年份 | 2015 | 2015 | 高 | |
| 总部所在地 | 加利福尼亚州旧金山 | 2026-09 | 中 | |
| 运营架构 | OpenAI Foundation 控制 OpenAI Group PBC | 高 | ||
| 最新公开估值(十亿美元) | 852 | 2026-03 | 高 | |
| 企业客户数 | 1,000,000 | 2026-09 | 高 | |
| 周活跃用户数 | 1,000,000,000 | 2026-09 | 高 | |
| ChatGPT 工作场所席位数 | 7,000,000 | 2026-09 | 高 | |
| 月度收入运行率(十亿美元) | 2 | 2026-09 | 高 | |
| 员工总数 | 2026-09 | 低 | 当前披露材料中未对公开员工总数进行重新核对勾稽。 | |
| 累计融资金额 | 2026-09 | 低 | 公开信息源仅支持主要融资轮次,未能提供完全对账核准的累计总额。 |
当前规模数据源自公司公开说法或第三方报告;缺乏证据支撑的指标保持为 null,不作主观推测。
[CO003, CO005, CO007, CO008, CO009, CO010, CO011, CO014]| 人员 | 职务 | 背景 | 创始人契合度或职能分工 | 关键人依赖度 |
|---|---|---|---|---|
| Sam Altman | CEO 兼董事会成员 | 联合创始人,也是公司核心的对外代表 | 资本筹集、产品方向与利益相关方信任 | 高 |
| Greg Brockman | 总裁 / 产品负责人 | 联合创始人,主导技术演进与产品延续性 | 技术执行与模型部署 | 高 |
| Bret Taylor | 董事会主席 | 独立治理负责人 | 董事会议事流程与外界信任度 | 中 |
| Fidji Simo | 应用业务 CEO | 具备大规模消费者业务与平台生态运营经验 | 商业化落地与产品化 | 中 |
| Sarah Friar | CFO | 拥有资本市场与企业财务运营背景 | 融资纪律与 IPO 筹备就绪度 | 中 |
领导层名单特意选取代表性样本,并非完整组织架构图。
[CO006, CO007]1.3 资本基础与里程碑时间线
OpenAI 最新披露的资本运作是 2026 年 3 月以 $852 billion 估值完成的 $122 billion 融资。公司还表示当前月收入已达 $2 billion。因此,公开的里程碑记录已不再是一段创业叙事,而是一部由持续融资、产品发布、治理调整与服务故障构成的演进纪实。尽调面临的核心悬念不是公司是否有体量,而是这一体量究竟已有多少具备持久性,又有多少仍依赖于高度集中的资本与基础设施支撑。当前的财务全貌因而融合了公开估值、战略伙伴敞口以及股权结构表中悬而未决的经济利益。正是由于这种结合,即便未公开全部股权细节,投资者图谱也格外关键。当前的财务全貌因而融合了公开估值、战略伙伴敞口以及股权结构表中悬而未决的经济利益。正是由于这种结合,即便未公开全部股权细节,投资者图谱也格外关键。相关纪实如今不仅包含产品上线,还包括资本重组、融资进展以及服务中断,这些都在影响公众叙事。这是理解股权稀释、控制权分配以及该估值实际含金量的恰当切入点。[CO007, CO011, CO012, CO014]
| 利益相关方 | 角色 | 控制权或经济重要性 | 尽调关注点 |
|---|---|---|---|
| OpenAI Foundation | 控股非营利实体 | 控制集团并持有完整治理权 | 确认董事会保留事项与使命保护机制。 |
| Microsoft | 战略合作伙伴 | 算力与分发领域的实质性合作伙伴 | 厘清当前的排他性条款与商业经济效益。 |
| SoftBank / 2026 财团 | 主要出资方 | 领投最新一轮私募融资 | 确认后续追加投资权与任何老股交易机制。 |
| NVIDIA | 算力与资本合作伙伴 | 战略供应商兼投资方 | 将硬件供应协议与单纯的财务投资敞口区分开来。 |
| 企业客户 | 收入基本盘 | 已大规模采购席位与 API 调用额度 | 持续跟踪客户留存率、席位扩容及工作流嵌入深度。 |
股权结构表与抽屉协议未公开披露,因此本表属于尽调图谱,而非完整的股权明细表。
[CO005, CO007, CO008, CO009, CO010]OpenAI 的公开轨迹从非营利机构起步,演进为使命受控的公共利益公司(PBC),进而完成 $852B 巨额融资;在此期间,业务中断事故与企业端增长始终并存。
[CO003, CO004, CO005, CO007, CO008, CO009, CO010, CO012]1.4 附图与附表
02市场分析
2.1 市场边界与替代品
所涉市场绝非一个单薄的 TAM 标题数字。衡量 OpenAI 的实际业务边界,关键在于能够替代搜索、软件工具、分析以及部分服务人力的各项开支——横跨个人助理、企业工作流自动化与开发者/API 基础设施。公司自身披露的资料显示,该市场由高频重复工作流驱动,而非偶发查询工具;替代品涵盖内部自研、其他模型厂商以及现有企业级软件体系。该边界还囊括了各类相邻工具,原因在于采购方在同一个决策会议上,会将 OpenAI 与传统软件、自研方案以及其他 AI 平台放在一起比对。正因如此,这一品类正在不断突破旧有的聊天机器人标签。[CM001, CM007, CM012, CM013]
2.2 规模测算视角与受证据约束的普及节奏
OpenAI 当前的商业化底盘已经足够大,足以支持基于公司内部数据自上而下测算市场规模。企业报告与融资公告披露,公司拥有超过 100 万家企业客户、700 万个工作场所席位、逾 9 亿周活跃用户,以及每月 $2 billion 的收入运行水平。相比单一的 TAM 数字,更具实操价值的尽调视角是多维观察框架:个人端规模、企业级席位与开发者/API 调用量,共同指向一个商业化变现面积仍在扩大的品类。因此,更准确的理解方式是分层普及体系,而非固守单一市场规模数字。这种视角更契合一家能在获客后不断加深使用黏性的公司,同时也解释了为何收入——而非仅仅是使用量——才是反映市场体量最清晰的现实指标。[CM002, CM003, CM004, CM005, CM006, CM008]
| 测算视角 | 指标 | 数值 | 解读 | 置信度 |
|---|---|---|---|---|
| 个人端覆盖 | 周活跃用户数 | 1.0B+ | 大众市场转化漏斗与渗透深度 | 高 |
| 企业端渗透 | 企业客户数 | 1.0M+ | 组织内部的商业化渗透 | 高 |
| 职场覆盖度 | ChatGPT 工作场所席位数 | 7.0M+ | 席位扩张与高频复用基本盘 | 高 |
| 收入现状 | 月度收入 | $2B | 当前已变现的需求盘子 | 高 |
| 工作流使用强度 | 推理 Token 消耗增速 | 320x YoY | 业务使用不仅在拓宽,更在不断加深 | 高 |
| 使用活跃度 | 企业级周消息量 | 8x YoY | 企业端参与度仍在加速提升 | 高 |
上述指标仅为观察视角,并非正式的 TAM/SAM/SOM 测算模型。
[CM002, CM003, CM004, CM005, CM006]OpenAI 的市场从消费者触达延伸至企业席位及开发者 / API 调用,各层级均体现出更深的商业变现与工作流嵌入。
各层宽度仅表示层级,不代表数值比例。
本视角基于当前公开披露数据梳理,并非正式的 TAM 模型。
[CM001, CM013, CM014]2.3 采购方、驱动因素与制约条件
最早的采购方并非零散的个人消费者,而是拥有明确预算负责人、深受工作流痛点困扰、且具备足够业务量来承担集成成本的在职员工与组织。OpenAI 自己的企业报告强调了生产力提升、使用频率增加,以及技术、医疗健康、制造业、金融和专业服务等领域更高的渗透强度。关键制约因素在于治理、集成与信任。随着这些瓶颈逐步缓解,市场正从前期的摸索尝试扩展至嵌入式工作流基础设施。采购方看重的是节约的时间、治理保障,以及 AI 能否深度嵌入内部可复用的工作流,而非仅当作新鲜玩具。换言之,买单方买的是工作流成果,而非模型演示。[CM008, CM009, CM010, CM011, CM012]
| 细分市场/品类 | 包含的支出 | 排除的支出 | 采购方/付款方 | 战略关联度 |
|---|---|---|---|---|
| 个人助手 | 订阅费、付费升级、广告变现 | 通用搜索与基础即时通讯 | 个人 / 家庭 | 流量漏斗顶层触达与品牌漏斗 |
| 企业工作流 AI | 席位费、内部副驾驶(Copilot)、智能体工作流工具 | 未接入 AI 的传统通用办公软件 | 企业 IT / 业务条线负责人 | 核心商业化变现路径 |
| 开发者/API 基础设施 | 按 Token 计费的模型调用及内嵌 AI 功能 | 非 AI 托管软件与传统基础设施 | 工程 / 产品 / 平台团队 | 经由开发者生态实现分发 |
| 内部自研与开源替代方案 | 私有化部署、定制模型、自研智能体 | 已计入别处的公有 SaaS AI 支出 | 工程 / 平台团队 | 主要替代方案与价格施压来源 |
本表旨在界定市场边界,并非给出精确的自上而下 TAM 数字。
[CM001, CM012, CM013]| 客群 | 采购决策方 | 实际使用者 | 付款方 | 工作流 | 预算归属方 | 采纳触发点 |
|---|---|---|---|---|---|---|
| 个人消费者 | 个人 | 个人 | 个人 | 通用日常助手 | 家庭 / 个人 | 低门槛尝试与习惯养成 |
| 中小企业 / 中端市场 | 运营 / 业务条线主管 | 团队成员 | 企业 | 文本起草、客户支持、分析 | 部门 / 运营预算 | 提高生产力或节省时间 |
| 大型企业 | IT / 业务条线负责人 | 企业员工 | 企业 | 知识工作、编程、数据分析 | IT / 职能预算 | 系统集成与治理审批通过 |
| 开发者 / 平台 | 技术工程主管 | 工程师 | 工程研发部门 | API 集成与智能体工作流 | 平台工程 | 对自主可控、迭代速度与差异化的诉求 |
使用者与买单人往往并非同一方,采购决策者与系统所有者分离的问题因此格外关键。
[CM007, CM008, CM009, CM010, CM012]2.4 附图与附表
03竞争格局
3.1 竞争格局与替代方案
OpenAI 的对手早已不仅是其他模型厂商。公司现有产品矩阵已延伸至代码仓库工作流、编程智能体及内部工作流自动化,直接对上了 GitHub 这类操作界面、企业自研方案,以及由企业团队拼装的模块化工具链。核心威胁并非某一个具体对手,而是客户可能在多家模型厂商之间多栖采购(multi-home),同时牢牢把控自己的工作流界面。这导致了这样一种市场格局:客户跨供应商采购模型,却把工作流界面留在别处。正因如此,分发能力与客户留存率成了远比单次模型基准跑分发布更为关键的命题。[CP001, CP003, CP007, CP010]
3.2 捆绑优势与分发能力
Codex 扭转了竞争叙事,因为它并非通用的聊天机器人。它是一款能够并行处理任务、融入 CLI 与 ChatGPT 工作流的工作流智能体,且紧贴代码审查、代码重构与拉取请求(PR)环节。这让分发渠道成了护城河的第一关键变量:OpenAI 必须卡位工程师现有的工作场景,尤其是 GitHub、终端以及企业自动化界面。实际操作中,开发者面对同一款模型可以选择不同的操作界面,这正是竞争格局的关键变数。因此,公司的护城河不仅关乎智力水平,更是一场关于触达点卡位的较量。[CP002, CP004, CP005, CP006, CP008]
3.3 护城河持久度与被替代风险
护城河因而正从单一的模型能力转向工作流集成、治理体系与产品广度。OpenAI 越是深度捆绑对话、智能体、API 调用以及感知代码仓库的工具链,就越能降低客户单独替换某一层组件的可能。反向风险则是同质化:若竞争对手追平模型层能力并保持工作流层开放,OpenAI 或将面临价格战压力,而非通吃全局的赢家通吃局面。只有当工作流层保持高黏性、且客户不再把 OpenAI 视为可随时替换的模型 API 时,护城河才守得住。一旦调用迁移至周边操作界面,OpenAI 虽仍能吸引眼球,却将丧失部分定价权。[CP005, CP007, CP009, CP010]
| 价格 / 计量单位 / 合同模式 | OpenAI | 竞争对手类型 | 未知要素 | 业务启示 |
|---|---|---|---|---|
| 席位费加用量计费 | 是 | 企业级套件 | 客户真实折扣情况未知 | 产品捆绑可降低切换摩擦 |
| API Token 计费 | 是 | 模型厂商 / 开源 API | 净单位经济效益不透明 | 相比宣传的模型测试跑分,性价比更为关键 |
| CLI / 智能体任务计费 | 是 | 工作流智能体 | 用量上限与速率限制不透明 | 智能体交互入口演变为用量消耗型业务 |
竞争对手的公开定价跨越不同合同与计费模式,无法直接等价对比。
[CP002, CP005, CP010]3.4 附图与附表
04财务情况
4.1 收入规模与利润率全貌
公开数据呈现的景象令人瞩目:OpenAI 称其单月收入达 $2 billion,而 Sacra 估计其年化收入接近 $40 billion,2026 年第二季度收入为 $6.7 billion。同一家第三方机构的分析指出,企业端业务如今贡献了超过一半的收入,且毛利率在 33% 左右,这使公司看起来更像一个体量庞大、仍在扩张的基础设施平台,而非成熟的软件企业。这一业务结构意义重大,因为公司同时在消费端行为与企业工作流深度两个维度推进商业化变现,而非依赖单一渠道。这是一个强劲的规模扩张故事,但尚不具备成熟软件巨头的成本结构。[CI002, CI003, CI004, CI008]
| 指标 | 数值 | 日期 | 置信度 | 缺口 |
|---|---|---|---|---|
| 最新公开估值 | $852B | 2026-03 | 高 | |
| 年化收入预测 | $40B | 2026-07 | 中 | |
| 季度收入(2026 年 Q2) | $6.7B | 2026-06 | 中 | |
| 毛利率 | 33% | 2026-07 | 中 | |
| 月度收入运行率 | $2B | 2026-09 | 高 | |
| 现金烧钱速度(2026 年) | ~$27B | 2026-09 | 中 | |
| 现金烧钱速度(2027 年) | ~$63B | 2026-09 | 中 | |
| 循环信贷额度 | $4.7B | 2026-03 | 高 |
上述数值兼有公司官方说法与第三方估算;毛利率与烧钱速度并非经审计的公开财务数据。
[CI001, CI002, CI004, CI005, CI006, CI008]OpenAI 的收入规模固然庞大,但烧钱区间同样处于高位,公司仍需证明自身具备持久的经营杠杆。
| 节点 / 连接 | 节点 / 起点 | 终点 | 说明 |
|---|---|---|---|
| 节点 1 | 收入 [rev] | 年化 ~$40B / 月均 $2B | |
| 节点 2 | 毛利率 [gm] | 预计 ~33% | |
| 节点 3 | 烧钱速度 [burn] | 2026 年 ~$27B / 2027 年 ~$63B | |
| 节点 4 | 流动性 [liq] | $4.7B 循环信贷 + $122B 融资轮 | |
| 节点 5 | IPO 叙事 [ipo] | 公开上市路径 | |
| 节点 6 | 自由现金流 [fcf] | 至今仍未经证实 | |
| 连接 1 | 收入 [rev] | 毛利率 [gm] | |
| 连接 2 | 毛利率 [gm] | 烧钱速度 [burn] | |
| 连接 3 | 烧钱速度 [burn] | 流动性 [liq] | |
| 连接 4 | 流动性 [liq] | IPO 叙事 [ipo] | |
| 连接 5 | IPO 叙事 [ipo] | 自由现金流 [fcf] |
收入与烧钱数据均采用当前公开估计值及公司公开说法,非经审计报表口径。
[CI001, CI002, CI005, CI006, CI012, CI013]4.2 烧钱速度、流动性与融资组合
关键症结在于烧钱速度。公开分析预计 2026 年现金消耗约 $27 billion,2027 年消耗约 $63 billion,也就是说,该业务仍需巨额持续资本注入才能跟上自身的需求曲线。OpenAI 拥有的 $4.7 billion 循环信贷额度虽有助益,但并未改变公司高度依赖外部融资的根本事实。2026 年 3 月的融资轮次还将股权结构表扩展至具备算力协同效应的战略投资方,进一步坐实了“金融运作与基础设施建设密不可分”的观点。正因如此,流动性纪律和战略投资方的重要性绝不亚于显眼的估值数字。融资组合的设计因而与经营计划同等关键。[CI005, CI006, CI007, CI010]
4.3 估值逻辑推导
在 $852 billion 的高估值下,市场必须相信收入增速、产品广度以及平台锁定效应足以跑赢高昂的算力成本。公开数据虽能撑起高溢价的成长叙事,却尚未展示出清晰的利润率逻辑。这也是为何即便最新收入体量有所放大,估值立场依旧维持“偏高”:公司增速或许快到足以支撑其当前规模,但仍需证明这种规模能够沉淀为持久的自由现金流。鉴于公开市场可比标的稀缺,最有效的估值推导依据仍是烧钱纪律、毛利演进轨迹以及 IPO 叙事的可信度。当前的估值倍数依然建立在这样一个假设之上:即便盈利拐点尚不清晰,市场也愿意提前为未来的选择权支付高昂溢价。[CI001, CI009, CI011, CI010]
4.4 附图与附表
05产品与技术
5.1 现有产品矩阵
OpenAI 现有的商业化版图横跨对话、编程、企业工作流、垂直行业方案以及安全与访问控制。这种广度举足轻重,意味着公司不再是纯粹的模型提供商,而是演变为一个在模型层之上包裹了工作流、治理体系与部署界面的产品平台。技术栈的延展度如今已足够支持采购方将 OpenAI 视作完整平台,而非单一的模型 API 端点。产品线现已囊括采购方看重的控制功能、垂直封装方案以及分发机制。这种平台定位成了客户评估公司的核心框架,采购方如今会拿这套组合拳与更广泛的企业级软件平台横向对比。[CE001, CE007, CE008]
5.2 Codex 与工作流深度集成
Codex 是这一平台化转型的最鲜明范例。OpenAI 将其定义为能够并行处理多项任务的云端软件工程智能体,周边生态也日益将其视为一款工作流产品,而非局限在狭义的代码辅助工具。这使得类似 GitHub 的界面与具备代码仓库感知能力的工具链成了竞争参照物。这之所以关键,是因为工程团队采购工具看重的是能否融入其代码库管理和评审习惯,而非仅仅靠漂亮的基准测试跑分。外部生态已经在主动教导团队如何上手使用,这是极具说服力的渗透信号;这也解释了为什么紧贴工作流程往往比炫耀基准跑分更为重要。这标志着产品评估逻辑发生了实质性转变。[CE002, CE003, CE009]
5.3 安全合规、访问限制与部署控制
公司最新的前沿能力配备了新的安全护栏。Astra 明确主打关键网络安全能力,Daybreak 则允许管理员按项目或用户精细划分访问权限。产品演进方向明确指向更可控的企业级部署,即便这会让正常操作遭遇更多阻力。前沿产品叙事日益与治理层绑定,由后者裁定谁能调用性能最顶尖的模型。与此同时,更严格的管控也给客户带来了更多技术支持与治理管理负担。治理机制正在转变为一大产品卖点,而非纯粹的合规包袱,这也促使公司更加向受监管软件的标准靠拢。[CE004, CE005, CE006, CE010, CE011]
| 能力项 | 业务收益 | 潜在风险 | 尽调要求 |
|---|---|---|---|
| 并行任务智能体 | 吞吐效率更快 | 监管监督难度加大 | 检查运行日志与任务边界 |
| 网络安全管控 | 治理机制更强 | 摩擦阻力增加 | 评估误报率与技术支持负担 |
| 行业定制打包 | 转化率更高 | 复杂度增加 | 跟踪各细分行业的成单胜率 |
| CLI 命令行集成 | 开发者粘性 | 工作流依赖 | 按代码仓类型审查留存率 |
注重治理的平台必须兼顾能力拓展与管控指标。
[CE004, CE005, CE010, CE011]OpenAI 以模型层为基础,产品已覆盖对话、编程、垂直工作流和治理控制。
| 节点 / 连接 | 节点 / 起点 | 终点 | 说明 |
|---|---|---|---|
| 节点 1 | 模型层 [model] | 前沿模型与 API | |
| 节点 2 | ChatGPT [chat] | 消费级与办公交互界面 | |
| 节点 3 | Codex [codex] | 编程智能体与 CLI | |
| 节点 4 | Daybreak / Astra [controls] | 安全与访问控制机制 | |
| 节点 5 | 垂直行业打包方案 [verticals] | 金融服务与企业工作流 | |
| 节点 6 | 文档与生态 [docs] | 开发教程与社区工具链 | |
| 连接 1 | 模型层 [model] | ChatGPT [chat] | |
| 连接 2 | 模型层 [model] | Codex [codex] | |
| 连接 3 | 模型层 [model] | Daybreak / Astra [controls] | |
| 连接 4 | ChatGPT [chat] | 垂直行业打包方案 [verticals] | |
| 连接 5 | Codex [codex] | 垂直行业打包方案 [verticals] |
5.4 附图与附表
06客户情况
6.1 客户客群与采购动线
目前的获客动线横跨个人消费者、个体职场人、企业席位以及大型企业级部署。采购决策者与实际使用者往往不同:IT 或业务条线主管负责审批开支,一线员工使用产品,系统管理员负责权限管控。最有说服力的公开证据不仅是席位总数,更是高频复用、内部 GPTs 构建以及部门级的工作流集成。理解这一现象的最佳视角是“先落地后扩张”(land-and-expand)策略:始于熟练上手,终于部门级规模部署。这种模式让客户盘子看起来远比单纯的个人应用安装量要宽广得多。这也构筑了更具想象力的多席位增长叙事。这也构筑了更具想象力的多席位增长叙事。[CU001, CU002, CU006, CU007]
6.2 投资回报率(ROI)与垂直行业渗透
OpenAI 自己的企业报告称,员工平均每天可节省 40–60 分钟,75% 的受访用户反映产出速度加快或质量提高。该报告与公司的垂直行业案例研究均显示,金融、医疗健康、制造业、技术以及专业服务是最明显的应用领跑者。这是平台从探索试验走向可复制日常运营的典型路径。合作伙伴主导的案例研究指出,公司正逐步打入采购流程繁琐、变革管理严苛的复杂企业级采购流程。核心考验在于,一旦新鲜感褪去,此类效率收益能否持续。公司依然需要更长周期的实践来验证这种模式。公司依然需要更长周期的实践来验证这种模式。[CU003, CU004, CU005, CU009]
6.3 业务持久性、集中度与信息缺口
客户维度的叙事虽然强劲,但并不完整。OpenAI 至今仍未以可勾稽的方式公开净留存率(NRR)、流失率、合同期限或头部客户集中度。这就留下了关键的尽调缺口:由于用户持续回访,业务看似具备持久性,但公开数据尚未量化这种持久性究竟集中度如何、黏性究竟有多强。这些缺失的指标,恰恰是将一款高增长产品与真正持久的企业级平台区分开来的试金石。缺乏这些数据,市场只能单纯凭使用量和席位增长推断业务质量。这正是客户投资逻辑中遗留的尽调缺口。这正是客户投资逻辑中遗留的尽调缺口。[CU008, CU009, CU010]
| 指标 | 状态 | 为何重要 | 下一步尽调要求 |
|---|---|---|---|
| 净留存率(NRR) | 反映业务扩张与客户粘性 | 要求提供收入队列分析 | |
| 客户流失率 | 反映客户留存水平 | 要求提供客户数与席位流失率 | |
| 合同期限 | 反映收入耐久性 | 要求提供标准条款清单 | |
| 头部客户集中度 | 反映下行风险集中度 | 要求提供头部客户收入占比 |
上述四项指标仍无公开数据。
[CU008]OpenAI 的客户采纳路径始于尝鲜好奇,逐步推进至席位部署,继而向深度工作流整合及内部自定义 GPT 构建拓展。
[CU001, CU002, CU003, CU006, CU010]6.4 附图与附表
07风险
7.1 可靠性与客户体验风险
OpenAI 自身的状态监控历史记录显示,ChatGPT、API、登录及特定模型服务曾多次发生事故,外部评价渠道也依然充斥着大量不满情绪。一家已蜕变为日常工作基础设施的公司,哪怕只有轻微的宕机率或技术支持滞后,都会成为阻碍企业级客户采纳的绊脚石。即便确切的事故次数尚未公开,各大评测网站和用户论坛中仍能看到更广泛的投诉。尽调的核心关注点不仅在于故障是否发生,更在于这些事故是否已波及那些将 OpenAI 深度嵌入日常工作、无法容忍长时间服务中断的客户。这不仅是产品运维问题,更直接关乎收入留存。[CR003, CR004, CR009]
7.2 网络安全风险与访问控制
《Path to Astra》一文明确提出了网络安全风险,指出 Astra 已达到关键的网络安全门槛,而 Daybreak 则增加了分级访问控制。这固然是走向成熟的标志,但 OpenAI 如今操盘的是一款既强大又脆弱的产品:各类管控虽降低了滥用风险,却可能拖慢正常工作节奏,或在合规工作流中引发误报。实际运作中,每当用户尝试执行敏感任务或涉及网络安全的关联流程时,OpenAI 都必须在安全性与易用性之间权衡取舍。真正的挑战在于:如何在抑制滥用的同时,不让产品成为正常用户频繁遭遇干扰的源头。[CR001, CR002, CR005, CR007, CR008]
7.3 治理结构与法律风险
OpenAI 的治理叙事依然非同寻常,而这种特殊架构屡次招致公众的放大审视。公司由非营利组织控制的公共利益公司(PBC)形态、围绕该架构旷日持久的法律与使命纷争,以及与 IPO 相关的法律评议,均说明治理风险绝非细枝末节,而是核心投资逻辑不可分割的一部分。面向未来的上市企业,这类架构可能引发信披摩擦、利益相关方关系紧绷,并引来投资方与监管机构的双重审视。这让治理风险超越了单纯的新闻标题范畴,演变为制约融资步伐与战略落地的经营层面的硬约束。在庞大业务体量下,该问题已具备重大实质影响。[CR006]
| 风险项 | 影响领域 | 严重程度 | 尽调要求 |
|---|---|---|---|
| 使命宗旨与董事会争议 | 控制权与战略走向 | 高 | 审查公司章程、细则及董事会保留事项 |
| IPO 监管审查 | 信息披露与上市就绪度 | 高 | 评估所有未决诉讼与监管暂缓事项 |
| 公共利益公司架构复杂性 | 治理形象与利益相关方平衡 | 中 | 梳理基金会(Foundation)与 PBC 之间的权利划分 |
该治理架构本身就是风险画像的一部分。
[CR006]7.4 附图与附表
08估值
8.1 估值观察框架
最新公开估值为 $852 billion,公司自己也表示单月收入已达到 $2 billion。两项数据叠加,核心问题直指:极速膨胀的体量能否转化为持久的经营杠杆。市场目前愿意将其视作定义赛道的标杆平台出资支持,但高估值仍需消化巨额资金消耗、沉重的资本承诺,以及未来伴随信息披露增多必然面临的上市折价。尽调的关键要害在于:市场当下定价买入的是远期平台图景,而非眼下的现金造血能力。这固然保留了上行空间,但只要增长或利润率指标出现任何闪失,估值就会面临快速重估。[CV001, CV004, CV008]
8.2 增长与现金消耗博弈
Sacra 的测算显示,其年化收入达 $40 billion,毛利率约 33%,且 2026 年资金消耗仍可能高达 $27 billion,2027 年消耗可能达 $63 billion。这足以撑起庞大的增长叙事,但尚未足以支撑起成熟软件企业的高纯度估值倍数。换言之,该估值既在为未来的利润率扩张买单,也是在赌其对依然高资本密集的商业模式的落地执行。真正该问的并非体量是否存在,而是体量扩张的步伐能否快到跑赢推理开支、技术支持重负以及融资拖累。一旦这些增长曲线发生背离,即便收入持续走高,估值倍数也会遭遇挤压。[CV002, CV005, CV009]
OpenAI 庞大的收入体量暂时撑住了高估值,但持续沉重的烧钱轨迹,使估值倍数始终承压。
| 节点 / 连接 | 节点 / 起点 | 终点 | 说明 |
|---|---|---|---|
| 节点 1 | 估值 [val] | 当前估值标尺 $852B | |
| 节点 2 | 收入 [rev] | 年化 ~$40B | |
| 节点 3 | 毛利率 [gm] | 预计 ~33% | |
| 节点 4 | 烧钱速度 [burn] | 2026 年 $27B / 2027 年 $63B | |
| 节点 5 | 流动性 [liq] | 战略资本 + 循环信贷 | |
| 节点 6 | 自由现金流 [fcf] | 至今仍未经证实 | |
| 连接 1 | 估值 [val] | 收入 [rev] | |
| 连接 2 | 收入 [rev] | 毛利率 [gm] | |
| 连接 3 | 毛利率 [gm] | 烧钱速度 [burn] | |
| 连接 4 | 烧钱速度 [burn] | 流动性 [liq] | |
| 连接 5 | 流动性 [liq] | 自由现金流 [fcf] |
该桥接分析仅具方向参考意义,旨在展示从业务规模向现金转化的实现路径。
[CV001, CV002, CV003, CV005]8.3 IPO 路径与尽调推导
舆论叙事已然全面转向 IPO 框架。FT 和 Tech Insider 均认为公开上市已在运作推进,融资结构中如今引入了更多战略参与方和更宽广的流动性退出方案。这虽然降低了资金链断裂风险,却并未消除核心的尽调疑虑:优先股堆叠(preference stack)究竟如何构成、资金消耗中有多少属于刚性兑付承诺,以及公司多快能将庞大体量转化为自由现金流?即便在运营数据尚未彻底理顺之前,融资故事仍能讲通,这也是治理条款、抽屉协议(side letters)和优先权利在此时格外关键的原因。正是这些具体条款,决定了上行红利究竟能有多少真正落入普通股股东囊中。[CV003, CV006, CV007]
| 投资方类别 | 角色定位 | 关注原因 | 尽调要求 |
|---|---|---|---|
| 战略投资方 | 资金加算力支持 | 可支撑基础设施建设与渠道分发 | 确认商业补充协议条款 |
| 财务投资方 | 成长型资本 | 支撑 IPO 路径与前期烧钱消耗 | 审查优先股堆叠架构 |
| 散户与员工持股渠道 | 扩大股权持有面 | 放大市场信号效应与流动性 | 核查流转机制与锁定期安排 |
| 真实的优先股堆叠结构为何? | 待明确 | 影响下行保护与 IPO 转股经济性 | 待明确 |
| 已承诺的固定烧钱规模有多大? | 待明确 | 决定后续融资跑道长度 | 待明确 |
| 自由现金流(FCF)何时能转正? | 待明确 | 决定估值能否长期支撑 | 待明确 |
本表将融资参与方类别与尚未解决的核心估值问题合并呈现。
[CV001, CV006, CV007, CV010, CV011, CV012]8.4 附图与附表
免责声明
本报告仅基于公开及预先获取的证据更新;缺乏证据支撑的非公开指标明确列为未决项,不作无依据的主观估计。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | OpenAI describes itself as an AI research and deployment company with a mission to ensure that artificial general intelligence benefits all of humanity. | 高 | SO001, SO003, SO007 |
| CO002 | The OpenAI Foundation controls OpenAI Group PBC and can appoint and replace the Group board. | 中 | SO001 |
| CO003 | OpenAI was founded in 2015 as a nonprofit research organization. | 高 | SO001, SO007 |
| CO004 | OpenAI created its for-profit subsidiary in 2019 to scale research and deployment. | 中 | SO001 |
| CO005 | The 2025 recapitalization left the Foundation with a 26% equity stake and aligned the for-profit around public-benefit corporation rules. | 中 | SO001 |
| CO006 | OpenAI’s current board materials name Sam Altman as CEO and list Bret Taylor, Adam D’Angelo, Paul Christiano, Sue Desmond-Hellmann, Zico Kolter, Paul Nakasone, Adebayo Ogunlesi, Nicole Seligman, David Vélez, and Robin Vince as directors. | 中 | SO001, SO007 |
| CO007 | OpenAI closed a $122 billion funding round at an $852 billion valuation in March 2026. | 高 | SO002, SO004 |
| CO008 | OpenAI says more than 1 million business customers now use its tools. | 高 | SO005, SO006 |
| CO009 | OpenAI says ChatGPT for Work has more than 7 million seats and ChatGPT Enterprise seats have grown roughly 9x year over year. | 中 | SO006 |
| CO010 | OpenAI says it now reaches more than 1 billion weekly active users and 2.5 million businesses. | 中 | SO002 |
| CO011 | OpenAI says it is generating $2 billion in revenue per month in the latest funding announcement. | 中 | SO002 |
| CO012 | OpenAI’s status history shows repeated incidents across ChatGPT, the API, login, and model-specific services. | 中 | SO011, SO012 |
| CO013 | OpenAI’s enterprise usage report says weekly enterprise messages grew about 8x and average reasoning-token consumption per organization grew about 320x year over year. | 中 | SO006 |
| CO014 | OpenAI’s latest public disclosure package still does not fully reconcile headcount, locations, or lifetime capital raised. | 中 | SO001, SO002, SO004 |
| CM001 | OpenAI’s market is best framed as a blend of consumer assistants, enterprise workflow software, and developer/API infrastructure rather than a single narrow product category. | 中 | SM001, SM002, SM003, SM005, SM006 |
| CM002 | OpenAI says more than 1 million business customers already use its tools directly. | 高 | SM002, SM001 |
| CM003 | OpenAI’s enterprise report says the company now serves more than 7 million ChatGPT workplace seats. | 中 | SM001 |
| CM004 | OpenAI says weekly enterprise messages grew about 8x and average reasoning-token consumption per organization grew about 320x year over year. | 中 | SM001 |
| CM005 | OpenAI says ChatGPT supports more than 900 million weekly active users and more than 2.5 million businesses. | 高 | SM008, SM004 |
| CM006 | OpenAI says it is generating $2 billion in revenue per month, which implies a monetized base that is already massive before any public listing. | 中 | SM004 |
| CM007 | Public-facing product descriptions show OpenAI selling into knowledge work, coding, data analysis, and multimodal assistance rather than only chat. | 中 | SM003, SM005, SM006 |
| CM008 | The enterprise report says users save 40–60 minutes per day on average and that 75% report improved speed or quality. | 中 | SM001 |
| CM009 | The same report shows technology, healthcare, manufacturing, finance, and professional services as the clearest early adoption sectors. | 中 | SM001 |
| CM010 | International adoption is no longer optional: the enterprise report says customer growth is broad-based and global, not limited to the United States. | 中 | SM001 |
| CM011 | OpenAI’s market position is supported by workflow depth, not only user count, because repeated use and custom GPTs expand inside firms over time. | 中 | SM001 |
| CM012 | The main constraints on adoption remain trust, compliance, budget ownership, and integration effort rather than raw awareness. | 中 | SM001, SM006 |
| CM013 | The company’s own public materials do not yet isolate a clean SAM/SOM for a single product line, so market sizing remains a diligence exercise rather than a precise public number. | 中 | SM001, SM002, SM004 |
| CM014 | OpenAI’s market is expanding in layers, with consumer reach, enterprise seats, and workflow depth all reinforcing one another. | 中 | SM001, SM002, SM003, SM009 |
| CP001 | OpenAI now overlaps with developer tooling, code hosting, and workflow-automation surfaces, not just model hosting. | 中 | SP001, SP002, SP003, SP004, SP005 |
| CP002 | Codex is positioned as a cloud software-engineering agent that runs tasks in parallel and works through CLI and ChatGPT surfaces. | 高 | SP001, SP004 |
| CP003 | TechRepublic explicitly frames OpenAI as eyeing a GitHub alternative, which is a clear signal that the company is moving into GitHub-adjacent territory. | 中 | SP003 |
| CP004 | OpenAI’s public GitHub presence and open-source footprint reinforce distribution into the developer ecosystem. | 中 | SP002, SP005 |
| CP005 | OpenAI’s own product lineup spans chat, coding, images, deep research, agents, and browser/desktop workflows, which increases bundle power. | 高 | SP005, SP006, SP007 |
| CP006 | External observers such as ZenML and Teachfloor already describe Codex as a production workflow platform rather than a narrow model endpoint. | 中 | SP004, SP008 |
| CP007 | The most important substitute is still in-house build, because enterprise teams can swap between hosted models, open-source models, and internal agents. | 中 | SP005, SP007, SP009 |
| CP008 | GitHub remains a central distribution surface because OpenAI now needs to meet developers where repositories and pull requests already live. | 中 | SP002, SP003, SP004 |
| CP009 | The moat question is shifting from raw model quality toward workflow integration, governance, and the ability to ship a coherent product suite. | 中 | SP001, SP005, SP006 |
| CP010 | Adjacent tooling such as MCP, CLI automation, and repository assistants shows that coding workflows are becoming more modular and more contestable. | 中 | SP004, SP009 |
| CP011 | OpenAI’s bundle is strongest when it reduces the number of places a developer has to switch to replace one workflow. | 中 | SP001, SP005, SP006, SP008 |
| CP012 | The competitive risk is less about one rival than about many partial substitutes eroding one layer at a time. | 中 | SP003, SP004, SP007, SP009 |
| CI001 | OpenAI closed its latest funding round at $122 billion of committed capital and an $852 billion post-money valuation. | 高 | SI001, SI002 |
| CI002 | Sacra estimates OpenAI reached $40 billion in annualized revenue in July 2026 and reported $6.7 billion of Q2 2026 revenue. | 中 | SI003 |
| CI003 | Sacra says enterprise now represents more than half of revenue, with ads and coding/work products adding new monetization lanes. | 中 | SI003 |
| CI004 | Sacra estimates gross margin at roughly 33% while inference costs remain the dominant drag on economics. | 中 | SI003 |
| CI005 | Sacra and related analysis put projected cash burn at about $27 billion in 2026 and about $63 billion in 2027. | 中 | SI003, SI004, SI006 |
| CI006 | OpenAI expanded its revolving credit facility to about $4.7 billion and said it was undrawn at close. | 高 | SI001, SI002 |
| CI007 | OpenAI’s March 2026 financing brought in strategic participation from Amazon, NVIDIA, SoftBank, and Microsoft. | 高 | SI002, SI004 |
| CI008 | OpenAI has framed the business as a two-billion-dollar-per-month revenue engine that is still in high-growth mode. | 中 | SI001 |
| CI009 | Third-party commentary consistently describes the business as capital-intensive and compute-constrained rather than capital-light. | 中 | SI004, SI005, SI006 |
| CI010 | The public record still does not fully resolve exact preference stack, secondary mechanics, or the complete cumulative capital-raise total. | 中 | SI002, SI004, SI005 |
| CI011 | The latest public valuation implies an extremely rich multiple even before the company proves durable margin expansion. | 中 | SI002, SI003, SI004 |
| CI012 | A lot of the valuation debate is really a debate about whether OpenAI can turn scale into free cash flow faster than infrastructure spending expands. | 中 | SI003, SI006, SI007 |
| CI013 | The Financial Times corroborates that OpenAI’s scale is now large enough to make losses and margin trajectory the central financial question. | 中 | SI009, SI003, SI006 |
| CE001 | OpenAI now sells across ChatGPT, Codex, API, enterprise workflows, financial-services use cases, and security/access-control tooling. | 高 | SE001, SE003, SE004, SE005, SE006 |
| CE002 | Codex is positioned as a cloud software-engineering agent that runs many tasks in parallel and can read, edit, and test code. | 中 | SE001 |
| CE003 | OpenAI now exposes Codex through both ChatGPT and CLI-style workflows, which blurs the line between model and product. | 中 | SE001, SE008 |
| CE004 | The Path to Astra post says Astra reaches a critical cybersecurity threshold and requires stronger safeguards during development and deployment. | 中 | SE002 |
| CE005 | Daybreak access controls now let admins scope Cyber-model access at the project or user level across API and Codex. | 中 | SE003 |
| CE006 | OpenAI is moving product packaging toward enterprise controls, auditability, and role-based access rather than only raw model access. | 中 | SE003, SE004, SE005 |
| CE007 | The financial-services and enterprise pages show that OpenAI is productizing industry-specific workflows rather than staying generic. | 高 | SE004, SE005 |
| CE008 | The news feed and launch posts show rapid iteration across model, agent, and safety features. | 中 | SE006, SE005, SE002 |
| CE009 | TechRepublic and ZenML both frame Codex as part of a broader workflow platform rather than a narrow coding toy. | 中 | SE007, SE008 |
| CE010 | OpenAI’s platform now combines product breadth with more explicit guardrails, which should improve enterprise adoption but can also slow legitimate work. | 中 | SE002, SE003 |
| CE011 | OpenAI’s public product stack is becoming easier to buy and govern, but the company has not yet eliminated the early-stage friction of frontier capability controls. | 中 | SE002, SE003, SE004 |
| CE012 | OpenAI’s coding and control story is now reinforced by a broader developer ecosystem that treats Codex as a workflow tool, not a demo. | 中 | SE001, SE008, SE009 |
| CU001 | OpenAI says more than 1 million business customers use its tools and more than 7 million ChatGPT workplace seats are deployed. | 高 | SU001, SU002 |
| CU002 | OpenAI says enterprise usage is scaling with deeper workflow integration, not just one-off prompts. | 中 | SU002 |
| CU003 | OpenAI’s enterprise report says average workers save 40–60 minutes per day and 75% report faster or higher-quality output. | 中 | SU002 |
| CU004 | OpenAI partner and enterprise case studies show large-account adoption is moving into repeatable workflows. | 中 | SU009, SU010, SU011 |
| CU005 | OpenAI’s financial-services and BBVA case studies show that the company is already landing in regulated, workflow-heavy environments. | 高 | SU002, SU009, SU010 |
| CU006 | OpenAI says consumer familiarity with ChatGPT lowers enterprise adoption friction and shortens rollout cycles. | 高 | SU003, SU008 |
| CU007 | The customer base is therefore not just broad; it is increasingly embedded in repeat work, internal GPTs, and department-level deployments. | 中 | SU001, SU002, SU006 |
| CU008 | Diligence still lacks a clean public view of NRR, churn, contract length, and top-account concentration. | 中 | SU001, SU002, SU006 |
| CU009 | OpenAI’s current customer story is strongest where users can see immediate time savings and workflow reuse. | 中 | SU002, SU003, SU006 |
| CU010 | The public materials suggest a classic land-and-expand motion: seat deployment, custom GPT creation, and then broader workflow adoption. | 中 | SU001, SU002, SU006 |
| CU011 | OpenAI’s enterprise motion is increasingly partner-led in large accounts, which shows the company is moving beyond simple self-serve adoption. | 中 | SU009, SU010, SU011 |
| CU012 | The same public record shows a growing support surface, which means customer satisfaction and reliability perception are now part of the buying motion. | 中 | SU006, SU012 |
| CR001 | OpenAI says Astra meets a critical cybersecurity threshold and needs stronger safeguards during development and release. | 中 | SR001 |
| CR002 | OpenAI is now using Daybreak access controls to scope Cyber-model access by project and by user. | 中 | SR002 |
| CR003 | OpenAI’s status page shows recurring incidents across ChatGPT, the API, login, and model-specific services. | 中 | SR004, SR005, SR006, SR009 |
| CR004 | Public customer reviews and community posts show a non-trivial support and reliability complaint surface. | 中 | SR004, SR005 |
| CR005 | The company’s product controls may reduce misuse risk, but they can also slow or stop legitimate work. | 中 | SR001, SR002 |
| CR006 | OpenAI’s governance and legal profile still includes recurring dispute risk around the company’s mission and public-benefit structure. | 中 | SR007, SR008 |
| CR007 | Enterprise users should expect stricter monitoring and possible pauses when models are operating in cyber-sensitive contexts. | 高 | SR001, SR002 |
| CR008 | The public record does not yet show how often safety controls trigger false positives in legitimate enterprise workflows. | 中 | SR001, SR002 |
| CR009 | Reliability, cyber safety, and governance are now first-order diligence items rather than edge cases. | 中 | SR001, SR004, SR007 |
| CR010 | OpenAI’s product and safety updates are moving fast enough that risk controls have to be treated as part of the release process rather than post-launch cleanup. | 中 | SR001, SR002, SR009 |
| CR011 | Customer complaints in forums and review sites suggest that support and reliability perception can become a commercial issue even when the platform remains heavily used. | 中 | SR004, SR005, SR006 |
| CR012 | Governance complexity and IPO scrutiny mean that listing readiness is only one part of the diligence work. | 中 | SR007, SR008, SR009 |
| CR013 | The broader ecosystem around OpenAI includes developer-side integration patterns that make the platform feel more embedded and more exposed at the same time. | 中 | SR009, SR010 |
| CV001 | OpenAI closed a $122 billion financing round at an $852 billion valuation and says it is now generating $2 billion in revenue per month. | 高 | SV001, SV002 |
| CV002 | Sacra estimates annualized revenue at $40 billion and gross margin around 33%, while burn remains about $27 billion in 2026 and $63 billion in 2027. | 中 | SV003 |
| CV003 | The public market narrative is now explicitly IPO-driven, with the FT and Tech Insider both describing OpenAI as preparing for or weighing a public listing. | 中 | SV004, SV005 |
| CV004 | The market appears willing to pay a very rich multiple for growth because the company combines massive revenue growth with category-defining distribution. | 中 | SV001, SV002, SV006 |
| CV005 | The current valuation remains stretched because compute intensity, burn, and capital commitments are still very high. | 中 | SV003, SV007, SV009 |
| CV006 | The funding stack now includes strategic investors and a broader set of financing participants, which lowers liquidity risk but not operating risk. | 中 | SV001, SV008, SV009 |
| CV007 | OpenAI’s public materials do not fully disclose the preference stack, the exact secondary mechanics, or the full cap-table economics. | 中 | SV001, SV002, SV008 |
| CV008 | A reasonable diligence conclusion is that OpenAI has a world-class growth story but a valuation that remains expensive relative to current public economics. | 中 | SV001, SV002, SV003 |
| CV009 | The main risk to the valuation is that revenue can keep compounding while free cash flow remains elusive because of infrastructure and compute obligations. | 中 | SV003, SV007 |
| CV010 | Public commentary frames the business as financeable now, but not yet cleanly self-funding at the current growth rate. | 中 | SV004, SV009, SV011 |
| CV011 | The newest financing and revenue disclosures make common-equity upside possible, but only if burn bends down materially. | 中 | SV001, SV003, SV011 |
| CV012 | The valuation case ultimately depends on the company proving that revenue growth can outrun infrastructure intensity for several more quarters. | 中 | SV001, SV003, SV007 |
| 编号 | 出版方 | 标题 | 引文 |
|---|---|---|---|
| SO001 | OpenAI | Our structure | The OpenAI Foundation continues to control OpenAI Group and holds a 26% equity stake. |
| SO002 | OpenAI | OpenAI raises $122 billion to accelerate the next phase of AI | Today, we closed our latest funding round with $122 billion in committed capital at a post money valuation of $852 billion. |
| SO003 | OpenAI | OpenAI | Research & Deployment | OpenAI is built to turn this research progress into broad benefit for people and organizations around the world. |
| SO004 | CNBC | OpenAI closes record-breaking $122 billion funding round as anticipation builds for IPO | OpenAI announced it closed its record-breaking funding round at a post-money valuation of $852 billion. |
| SO005 | OpenAI | 1 million business customers: the fastest-growing business platform in history | More than 1 million business customers around the world are directly using OpenAI. |
| SO006 | OpenAI | The state of enterprise AI | OpenAI now serves more than 7 million ChatGPT workplace seats. |
| SO007 | Wikipedia | OpenAI | OpenAI was founded December 8, 2015 and lists Sam Altman and Greg Brockman among the founders. |
| SO011 | Trustpilot | OpenAI is rated "Bad" with 1.3 / 5 on Trustpilot | OpenAI is rated "Bad" with 1.3 / 5 on Trustpilot. |
| SO012 | OpenAI Community | Catastrophic Failures of ChatGpt that's creating major problems for users | Catastrophic Failures of ChatGpt that's creating major problems for users. |
| SM001 | OpenAI | The state of enterprise AI | Enterprise usage is scaling, with deeper workflow integration. |
| SM002 | OpenAI | 1 million business customers: the fastest-growing business platform in history | More than 1 million business customers around the world are directly using OpenAI. |
| SM003 | OpenAI | OpenAI | Research & Deployment | OpenAI is built to turn this research progress into broad benefit for people and organizations around the world. |
| SM004 | OpenAI | OpenAI raises $122 billion to accelerate the next phase of AI | OpenAI says it is generating $2 billion in revenue per month. |
| SM005 | Wikipedia | OpenAI | OpenAI lists ChatGPT, Codex, and GPT-6 among its products. |
| SM006 | Teachfloor | OpenAI: What It Is, Key Products, Technology, and How to Get Started | OpenAI spans text, code, image, and audio use cases. |
| SM007 | Tech Insider | OpenAI IPO: $850B Valuation, $25B Revenue [2026] | OpenAI revenue run-rate and valuation discussion now sit inside a public-IPO narrative. |
| SM008 | CNBC | OpenAI closes funding round at an $852 billion valuation | OpenAI said ChatGPT supports more than 900 million weekly active users. |
| SM009 | Sacra | OpenAI revenue, valuation & funding | Sacra estimates that OpenAI hit $40B in annualized revenue in July 2026. |
| SP001 | OpenAI | Introducing Codex | Codex is a cloud-based software engineering agent that can work on many tasks in parallel. |
| SP002 | OpenAI | OpenAI | https://github.com/OPENAI | OpenAI maintains a public GitHub presence and open-source tooling footprint. |
| SP003 | TechRepublic | OpenAI Reportedly Eyes a GitHub Alternative | OpenAI reportedly eyes a GitHub alternative. |
| SP004 | ZenML | Building Production-Ready AI Agents: OpenAI Codex CLI Architecture and Agent Loop Design | The article frames Codex CLI as a production-ready AI agent architecture. |
| SP005 | OpenAI | OpenAI | Research & Deployment | OpenAI is built to turn this research progress into broad benefit for people and organizations around the world. |
| SP006 | OpenAI | OpenAI News | OpenAI News aggregates ongoing launches and product changes. |
| SP007 | Wikipedia | OpenAI | OpenAI’s product list now spans ChatGPT, Codex, GPT Image, Deep Research, ChatGPT agent, and ChatGPT Atlas. |
| SP008 | Teachfloor | OpenAI: What It Is, Key Products, Technology, and How to Get Started | OpenAI spans text, code, image, and audio use cases. |
| SP009 | QueryBear | SQLite + OpenAI Codex MCP: CLI Database Access | OpenAI’s coding stack is being evaluated in adjacent developer workflows. |
| SI001 | OpenAI | OpenAI raises $122 billion to accelerate the next phase of AI | Today, we closed our latest funding round with $122 billion in committed capital at a post money valuation of $852 billion. |
| SI002 | CNBC | OpenAI closes record-breaking $122 billion funding round as anticipation builds for IPO | OpenAI announced it closed its record-breaking funding round at a post-money valuation of $852 billion. |
| SI003 | Sacra | OpenAI revenue, valuation & funding | Sacra estimates that OpenAI hit $40B in annualized revenue in July 2026. |
| SI004 | Tech Insider | OpenAI IPO: $850B Valuation, $25B Revenue [2026] | OpenAI has taken its first formal step toward the public markets. |
| SI005 | Medium | The $852 Billion Bet: Reading the OpenAI VC Pricing Model the Way Investors Actually Should | The analysis frames the $852 billion valuation as an investors-should model question. |
| SI006 | Substack | The Impossible Business Model of OpenAI | The article argues OpenAI’s business model is constrained by compute and burn. |
| SI007 | OpenAI | OpenAI | Research & Deployment | OpenAI is built to turn this research progress into broad benefit for people and organizations around the world. |
| SI008 | OpenAI | OpenAI News | OpenAI News surfaces continuing product launches and commercial updates. |
| SI009 | Financial Times | OpenAI financials, loss and revenue analysis | The FT corroborates the scale of OpenAI’s revenue and cost structure. |
| SE001 | OpenAI | Introducing Codex | Codex is a cloud-based software engineering agent that can work on many tasks in parallel. |
| SE002 | OpenAI | Path to Astra: critical capabilities and frontier safeguards | GPT-6 Astra is a major step forward in AI capability. |
| SE003 | OpenAI Help Center | OpenAI Daybreak Migration & Org consolidation Guide | OpenAI is introducing access controls for Daybreak across API and Codex. |
| SE004 | OpenAI | Introducing ChatGPT for Financial Services | OpenAI has a long history of collaborating with innovators to unlock novel AI solutions. |
| SE005 | OpenAI | OpenAI | Research & Deployment | OpenAI is built to turn this research progress into broad benefit for people and organizations around the world. |
| SE006 | OpenAI | OpenAI News | OpenAI News aggregates launches and product updates. |
| SE007 | TechRepublic | OpenAI Reportedly Eyes a GitHub Alternative | OpenAI reportedly eyes a GitHub alternative. |
| SE008 | ZenML | Building Production-Ready AI Agents: OpenAI Codex CLI Architecture and Agent Loop Design | The article frames Codex CLI as a production-ready AI agent architecture. |
| SE009 | DataCamp | OpenAI Codex Tutorial: A Guide With 3 Practical Examples | OpenAI Codex Tutorial: A Guide With 3 Practical Examples. |
| SU001 | OpenAI | 1 million business customers: the fastest-growing business platform in history | More than 1 million business customers around the world are directly using OpenAI. |
| SU002 | OpenAI | The state of enterprise AI | Enterprise usage is scaling, with deeper workflow integration. |
| SU003 | OpenAI | The Work Now Within Reach | Our products reach more than one billion weekly active users and 2.5 million businesses. |
| SU004 | OpenAI | Introducing ChatGPT for Financial Services | Financial services firms and developers can use our API. |
| SU006 | OpenAI | OpenAI News | OpenAI News aggregates launches and customer-facing updates. |
| SU007 | OpenAI | OpenAI | Research & Deployment | OpenAI is built to turn this research progress into broad benefit for people and organizations around the world. |
| SU008 | OpenAI | OpenAI raises $122 billion to accelerate the next phase of AI | OpenAI says ChatGPT supports more than 900 million weekly active users. |
| SU009 | CliffsNotes | CASE 17 Microsoft's Strategic Alliance with OpenAI, Inc. | Microsoft's Strategic Alliance with OpenAI. |
| SU010 | Course Hero | Case Study - OpenAI - Microsoft Partnership Since its founding | OpenAI and Microsoft partnership case study. |
| SU011 | Inveritasoft | OpenAI Select Partner for Enterprise AI Development | OpenAI Select Partner for Enterprise AI Development. |
| SU012 | Trustpilot | OpenAI is rated "Bad" with 1.3 / 5 on Trustpilot | OpenAI is rated "Bad" with 1.3 / 5 on Trustpilot. |
| SR001 | OpenAI | Path to Astra: critical capabilities and frontier safeguards | GPT-6 Astra is a major step forward in AI capability. |
| SR002 | OpenAI Help Center | OpenAI Daybreak Migration & Org consolidation Guide | OpenAI is introducing access controls for Daybreak across API and Codex. |
| SR009 | OpenAI | OpenAI News | OpenAI News aggregates launches and product updates. |
| SR004 | Trustpilot | OpenAI is rated "Bad" with 1.3 / 5 on Trustpilot | OpenAI is rated "Bad" with 1.3 / 5 on Trustpilot. |
| SR005 | OpenAI Community | Catastrophic Failures of ChatGpt that's creating major problems for users | Catastrophic Failures of ChatGpt that's creating major problems for users. |
| SR006 | WhatJobs News | Microsoft Outlook and OpenAI’s ChatGPT Work Hit Outages on Same Day | Microsoft Outlook and OpenAI’s ChatGPT Work Hit Outages on Same Day. |
| SR007 | Wikipedia | OpenAI | OpenAI has faced multiple lawsuits and governance disputes. |
| SR008 | Klover.ai | OpenAI IPO: Regulatory, Political, and Legal Risks [In-Depth Analysis] [2026] | The analysis highlights legal and regulatory risks around the OpenAI IPO. |
| SR010 | Logarithmic Spirals | Cache OpenAI Embeddings with SQLite for Low-Cost Astro Builds | Cache OpenAI Embeddings with SQLite for Low-Cost Astro Builds. |
| SV001 | OpenAI | OpenAI raises $122 billion to accelerate the next phase of AI | Today, we closed our latest funding round with $122 billion in committed capital at a post money valuation of $852 billion. |
| SV002 | CNBC | OpenAI closes record-breaking $122 billion funding round as anticipation builds for IPO | OpenAI announced it closed its record-breaking funding round at a post-money valuation of $852 billion. |
| SV003 | Sacra | OpenAI revenue, valuation & funding | Sacra estimates that OpenAI hit $40B in annualized revenue in July 2026. |
| SV004 | Financial Times | OpenAI weighs funding round at $1.2tn valuation before IPO | OpenAI weighs funding round at $1.2tn valuation before IPO. |
| SV005 | Tech Insider | OpenAI IPO: $850B Valuation, $25B Revenue [2026] | OpenAI has taken its first formal step toward the public markets. |
| SV006 | Medium | The $852 Billion Bet: Reading the OpenAI VC Pricing Model the Way Investors Actually Should | The analysis frames the $852 billion valuation as an investors-should model question. |
| SV007 | Substack | The Impossible Business Model of OpenAI | The article argues OpenAI’s business model is constrained by compute and burn. |
| SV008 | Tracxn | OpenAI - 2026 Funding Rounds & List of Investors | Tracxn tracks the 2026 funding rounds and investors. |
| SV009 | Pinggy | Racing to a Trillion: OpenAI and Anthropic's Funding History | The article tracks OpenAI’s funding history against a trillion-dollar market narrative. |
| SV010 | OpenAI | OpenAI | Research & Deployment | OpenAI is built to turn this research progress into broad benefit for people and organizations around the world. |
| SV011 | Yahoo Finance | OpenAI’s own forecast predicts 14... | OpenAI’s own forecast predicts losses. |