Cockroach Labs
高质量分布式数据库公司,私募市场价格要求很高
Cockroach Labs 看起来是真正的后期基础设施赢家候选,但与公开收入和效率证据相比,当前私募市场价格显得偏高。
封面要素
公司概况
Cockroach Labs 是一家总部位于纽约的基础设施公司,成立于 2015 年,围绕 CockroachDB 打造业务。CockroachDB 是兼容 PostgreSQL 的分布式 SQL 数据库,以全托管云服务和企业软件销售。公开证据显示,公司在支付、订单管理、欺诈、游戏和内部数据库平台工作负载中拿到了罕见强的客户证明,2026 年二级市场兴趣仍在。核心尽调张力在于:公司质量看起来真实,但当前财务披露仍过于稀疏,难以高置信度承销估值。
- 成立时间
- 2015-01-01
- 创始人
- Spencer Kimball, Ben Darnell, Peter Mattis
- 创立地点
- New York, NY
- 总部
- New York, NY
- 产品
- Cockroach Labs 销售 CockroachDB Cloud、自托管 CockroachDB Enterprise,以及面向全球分布、高韧性事务型工作负载的支持和迁移能力。
- 客户
- 运行关键任务、多区域、合规敏感或高可用应用的企业和技术成熟软件团队。
- 商业模式
- 数据库收入来自订阅和按量消费,覆盖托管云、企业软件、支持,以及迁移驱动的扩张。
- 阶段
- late-stage private
- 融资情况
- 2021 年 12 月 Series F 融资 $278M,估值 $5B;2026 年二级市场指标暗示估值约 $6.9B,但未披露新的一级资本。
执行摘要
主要优势
- 在关键任务分布式事务型工作负载上,产品和客户证据都很强。
- DoorDash、Netflix、Route、Riskified、Form3、Hard Rock 等高质量标杆客户,支撑了产品相关性和切换成本耐久性。
- 二级市场兴趣延续,溢价叙事也没有散,说明市场仍把 Cockroach 视作差异化后期资产。
- PostgreSQL 兼容性叠加多区域韧性和可迁移性,在大型基础设施品类里形成可信战略楔子。
主要风险
- 公开财务披露太薄,无法按当前价格验证 ARR、NRR、毛利率、集中度或现金跑道。
- 分布式数据库的正确性、升级纪律和支持强度,带来实质运营和执行风险。
- 客户组合质量看起来高,但偏向成熟复杂的大客户,可能拉长销售周期,并需要更重的售后支持。
- 除非隐藏指标显著强于公开锚点,否则当前隐含估值相对上市软件基础设施可比公司的倍数偏贵。
未决问题
- 当前 2025-2026 ARR 或收入运行率,以及分段净留存 / 总留存。
- 毛利率、支持成本强度,以及支持重的部署是否稀释经营杠杆。
- 客户集中度和头部企业账户收入占比。
- 现金余额、烧钱速度、现金跑道,以及除二级流动性外是否有新股融资计划。
目录
01公司概况
1.1 身份定位、产品边界与当前位置
Cockroach Labs 将自己定位为一家基础设施公司,解决的是一个狭窄但很痛的问题:事务数据库如何在节点、可用区甚至区域故障中仍保持在线,又不逼应用团队手工分片,或围绕脆弱的故障切换模式重做设计。可访问公开资料一致显示,公司位于纽约、成立于 2015 年;官方产品界面强调的是兼容 PostgreSQL、云原生的分布式 SQL 数据库,而不是专有开发者体验。产品栈已经不再是单一产品。2026 年,公司销售全托管 CockroachDB Cloud、自托管 Enterprise、迁移工具和支持,并通过 Basic、Standard、Advanced 云层级变现。分层很关键,因为它说明 Cockroach Labs 试图用同一平台同时吃下自下而上的试用和受监管生产工作负载。公开信任与安全材料强化了同一条 GTM 信息:韧性、合规和操作简化。因此,最强身份信号不是原始规模,而是公司坚持把事务正确性、多区域韧性和 PostgreSQL 熟悉度放进同一个现代应用记录源产品。[CO001, CO002, CO003, CO004, CO005, CO006]
| 指标 | 数值 / 状态 | 日期 | 置信度 | 缺口 |
|---|---|---|---|---|
| 成立时间 | 2015 | 2015 | 高 | |
| 总部 | 纽约州纽约市 | 2026-07-20 | 高 | |
| 当前阶段 | 私营 / Series F 轮 | 2026-07-20 | 高 | |
| 最近一次新股融资 | $278M Series F 轮,估值 $5B | 2021-12-16 | 高 | |
| 累计融资 | $633M | 2021-12-16 | 高 | |
| 2026 年老股交易价格信号 | Nasdaq Private Market 每股 $7.79 | 2026-07-02 | 中 | 老股交易定价不等同于新一轮新股融资。 |
| Notice 老股交易价格信号 | 每股 $8.18 | 2026-07-20 | 中 | 市场挂牌信息,而非已定价融资。 |
| 2023 年收入 / ARR | $128.3M | 2023-12-31 | 中 | 最新公开财务数字来自聚合方,而非管理层在 2026 年披露。 |
| 员工数估计 | ~715 名员工 | 2025-11-28 | 中 | 第三方估计;官方 2026 年员工数未披露。 |
| 客户规模证据 | 2021 年 200+ 家客户,加上 2026 年具名企业部署 | 2026-07-20 | 中 | 当前官方客户数未公开披露。 |
| 开源信号 | 公开 GitHub 仓库,历史 22k+ 星标 | 2026-07-20 | 中 | 当前星标数会随时间变化;2021 年数字为官方披露。 |
| 云可用性 | AWS、GCP、Azure;BYOC 和自托管选项 | 2026-07-20 | 高 | |
| 安全态势 | SOC 2、ISO 27001/27017/27018、PCI DSS、HIPAA-ready 声明 | 2026-07-20 | 高 | 认证范围取决于套餐和部署模式。 |
| 董事会披露 | 可获取来源未公开披露 | 2026-07-20 | 中 | 需要管理层或资料室尽调。 |
混合使用公司官方陈述、市场交易价格指标和第三方估计。null 或缺口文本表示信息尚未以足够精度公开披露。
[CO001, CO004, CO005, CO006, CO007, CO008]Cockroach Labs 把 PostgreSQL 兼容的分布式 SQL 核心,与托管云交付、合规工具和宕机敏感行业的客户证明连在一起;老股市场定价是投资人最主要的未解信号。
[CO002, CO004, CO005, CO006, CO007, CO012]1.2 融资历程、私募市场信号与规模标记
最后一轮一级融资仍是 2021 年 12 月 Series F,当时 Cockroach Labs 融资 $278 million、估值 $5 billion,并称累计融资已达 $633 million。按成长型软件公司的标准,这一轮已经偏旧,但它仍是最干净的官方估值锚点,因为信息直接来自公司,并点名了由 Greenoaks 领投的一线投资人阵容。此后变化的不是新一级融资,而是更可见的二级市场。Nasdaq Private Market 和 Notice 都显示,2026 年中股价位于高个位数区间;PM Insights 则披露,2026 年 6 月价格相较 2021 年一轮有溢价,暗示企业标记远高于官方 $5 billion 头条估值。这些信号确认了流动性和投资人兴趣,但没有解决核心尽调问题:公开收入证据仍然很薄。GetLatka 的 2023 年 ARR 数字为 $128.3 million,2025 年员工数估计约 715 人,说明公司已有有意义的规模;但公司 2026 年进展公告仍回避硬财务披露。由此得到的图景是:公司规模已足够重要,显然仍是私有公司,二级市场价格也可能高于最后一轮一级融资;但透明度仍不足,承销方不能把这些标记等同于一轮新定价融资。[CO008, CO009, CO010, CO011, CO012, CO013]
| 人物 | 职务 | 背景 | 创始人-市场匹配或职能覆盖 | 关键人依赖 |
|---|---|---|---|---|
| Spencer Kimball | CEO 兼联合创始人 | 在融资、产品和韧性叙事中最常露面的公开高管 | 主导融资、产品方向和企业定位叙事 | 高 |
| Sailesh Munagala | 首席财务官 | 在 2026 年动能更新中被公开列为新增高级领导层 | Cockroach Labs 从纯创始人主导的产品销售走向更成熟阶段时,补上财务领导力 | 中 |
| Kimball 之外的公开联合创始人名单 | 部分披露 | 可获取公开来源明显以 Kimball 为中心,但没有清楚列出完整的当前创始人名单 | 创始人连续性关系到基础设施公司可信度,但此处透明度不足 | 中 |
| 董事会与治理名单 | 可获取来源未披露 | 公开证据没有展示当前董事会构成或观察员席位 | 老股交易活跃之后,治理不透明更重要 | 中 |
本领导层表有意区分已清楚披露的高管,以及可获取来源中仍不透明的治理领域。
[CO021, CO022, CO035, CO020]| 利益相关方 | 角色 | 控制权或经济重要性 | 尽调问题 |
|---|---|---|---|
| Greenoaks | Series F 领投方 | 领投最近一次官方定价轮,锚定 $5B 估值 | 厘清董事会席位、持股比例,以及任何要约收购或流动性限制。 |
| Benchmark | 早期投资者,可见的长期支持者 | 代表公司从最早融资阶段到后期规模化的一条连续支持线 | 确认 Series F 后的当前持股和治理权利。 |
| Tiger Global | Series F 投资团中的成长型投资者 | 显示 2021 年基础设施定价高峰期交叉基金仍有胃口 | 评估未来内部人支持或老股流动性的意愿。 |
| Nasdaq Private Market 私募市场 | 老股流动性交易场所 | 发布估算股价,并为符合条件的卖方撮合员工 / 投资者转让 | 了解买卖价差深度、转让审批规则和买方集中度。 |
| Notice.co / 其他私募市场聚合方 | 老股市场情报层 | 提供指示性私募股价,但不具备新融资的确定性 | 将报价的每股价格与 409A 标记或实际要约文件核对。 |
| 员工和期权持有人 | 私募流动性计划中的潜在卖方 | 活跃老股交易可能影响留任、士气和稀释预期 | 索取期权池规模、刷新授予政策,以及当前任何流动性窗口。 |
本表聚焦影响估值和控制权的利益相关方,而不是试图用公开碎片重建完整股权结构表。
[CO008, CO009, CO010, CO012, CO013, CO014]Cockroach Labs 从 2015 年创立走到 2021 年 $5B 估值的 Series F 轮,此后进入 2026 年阶段:老股市场价格发现、企业客户证明和以韧性牵引的 AI 叙事共同定义公司。
客户里程碑使用公开案例研究的发布时间作为日期锚点,而不是内部最初上线日期。
[CO001, CO008, CO010, CO019, CO023, CO024]Cockroach Labs 有强企业部署证据,但公开财务披露有限;估值语境更多由老股交易塑造,而不是新的一级融资轮。
混合公司表述、市场报价和第三方估计;这些数字能指方向,但不等同于公司新披露的经审计数据。
[CO008, CO010, CO012, CO013, CO015, CO016]1.3 领导层集中度与治理可见度
领导层可见度异常集中在 Spencer Kimball 身上。他仍是 2021 年融资材料和 2026 年进展叙事绑定的公开高管,这有利于叙事一致,但也带来关键人物集中度。公司在 2026 年进展更新中任命 Sailesh Munagala 为首席财务官,释放了补强管理梯队的信号;这是可访问来源中最显眼的公开高管新增。除此之外,公开治理披露很薄。现成的公司资料和第三方资料并未清楚列出当前董事会名单、保护性条款或详细股权结构表。这不等于治理薄弱,但意味着外部投资者必须从融资财团而不是直接治理文件中推断很多信息。实际含义是,Cockroach Labs 现在像一家创始人主导的后期基础设施公司,拥有可信机构背书,但董事会构成和投资人控制的公开透明度有限。作为私有独角兽可以接受,但仍是尽调缺口,因为当公司开始在二级市场活跃交易,并把自己定位为长期企业平台时,治理强度变得更重要。[CO021, CO022, CO035, CO008, CO009]
| 日期 | 事件 | 类型 | 金额 / 状态 | 参与方 | 含义 |
|---|---|---|---|---|---|
| 2015 | Cockroach Labs 成立 | 创立 | 公司创立 | Spencer Kimball 与早期工程团队 | 开启面向云优先世界的分布式 SQL 投资逻辑。 |
| 2021-12-16 | 宣布 Series F 轮 | 融资 | $278M,估值 $5B;累计融资 $633M | Greenoaks 及后期投资团 | 后来所有定价讨论的官方估值锚点。 |
| 2021-12-16 | 披露 Serverless 采用里程碑 | 规模 | 10,000+ 名 Serverless 新用户 | Cockroach Labs | 显示自下而上的开发者漏斗开始与企业销售并行出现。 |
| 2021-12-16 | 披露 Cloud 渗透率里程碑 | 规模 | 50%+ 客户使用 Dedicated | Cockroach Labs | 确认到 2021 年,托管云变现正在成为核心。 |
| 2024 | Booking.com 公开迁移订单平台用例 | 客户 | ~20TB 订单平台 | Booking.com | 为关键任务工作负载增加头部旅行行业验证。 |
| 2024 | 发布 Netflix 集群案例研究 | 客户 | 380+ 个集群,160+ 个生产集群 | Netflix | 验证超大规模下多团队内部平台使用。 |
| 2025 | 发布 State of Resilience 研究 | 市场 | 1,000 名企业调研受访者 | Cockroach Labs + Wakefield Research | 强化韧性主导叙事和监管相关性。 |
| 2026-02 | 发布 CockroachDB v26.1 安全 / 合规版本 | 产品 | 扩展安全与合规控制 | Cockroach Labs | 显示公司强调受监管工作负载和身份感知访问。 |
| 2026-05 | 公开安全公告包含权限提升和数据完整性问题 | 反向 | 多项安全与正确性公告 | Cockroach Labs | 提醒投资者,分布式数据库信任取决于披露和补丁纪律。 |
| 2026 | 动能公告突出 AI 规模韧性和伙伴扩张 | 商业化 | 吞吐量最高提升 50%;与 IBM 的 OEM 关系;任命 CFO | Cockroach Labs | 在没有新一轮新股融资的情况下,将公司定位到下一阶段企业扩张。 |
里程碑有意把融资、产品、客户和反向披露放在一起,方便后续章节复用同一条记录时间线。
[CO001, CO008, CO010, CO018, CO019, CO023]1.4 客户证据、韧性可信度与反向信号
Cockroach Labs 最强的非财务证据,是具名客户集合的质量。Booking.com 将 CockroachDB 用在一个约 20 TB、覆盖多个欧洲区域的订单平台;SumUp 称它迁移了服务超过 4 million 商户的核心支付系统;Form3 在 AWS、GCP、Azure 之上运行多云支付骨干;Netflix 报告拥有超过 380 个 CockroachDB 集群。这些不是摆门面的 logo。它们是支付、旅行、媒体里的高可用工作负载,正确性和容灾能力真正重要。这些客户证据支撑公司 2026 年围绕 AI 规模韧性和永久在线运营的定位。与此同时,反向记录并非空白。Cockroach Labs 公开发布技术公告,2026 年清单包括权限提升问题、部分索引损坏、实时数据删除边界情况和备份完整性问题。状态页也显示公司愿意公开暴露云事故。合并来看,公司有真实企业证据和异常强的韧性资质,但这些优势与复杂分布式数据库的普通现实并存:产品面很宽,bug 会发生,信任既靠架构,也靠披露纪律。[CO023, CO024, CO025, CO026, CO027, CO028]
02市场分析
2.1 市场边界与品类逻辑
不能把 Cockroach Labs 放进整个数据库宇宙来测算。它的产品是分布式 SQL 数据库,面向云原生、关键任务事务型应用,需求是韧性、多区域可用性和 PostgreSQL 熟悉度。因此,最干净的品类锚点是分布式 SQL;更宽的云数据库和分布式数据库数字只能作为边界检查。 应纳入的支出,是面向应用的事务型数据基础设施:生产 OLTP 集群、韧性升级、云数据库运营、从传统关系型系统迁移,以及相邻的重合规现代化工作。排除项包括数据仓库、搜索引擎、纯文档存储,以及不直接替换事务记录源系统的开发者工具品类。这个边界很重要,因为一个很宽的“数据库 TAM”会让机会显得巨大,却几乎无法说明 Cockroach Labs 实际能赢下多少支出。 现状替代方案也比年轻软件品类更强。团队可以留在 PostgreSQL,留在 Aurora,用应用层分片,或选择 Spanner 这样的超大规模云厂商原生系统。因此,市场不是仅靠认知就能创造出来;只有当买方感受到足够的宕机、规模、合规或地理分布痛点,愿意承担迁移成本时,市场才会打开。这个框架能让后续估值更诚实:Cockroach Labs 卖的是一个有价值但竞争高度激烈的现代化决策,而不是无人认领的绿地需求。[CM001, CM002, CM003, CM018, CM033]
| 细分市场 / 类别 | 纳入支出 | 排除支出 | 买方 / 付款方 | 意义 |
|---|---|---|---|---|
| 分布式 SQL 数据库 | 全球事务集群、韧性升级、托管 SQL 运维、兼容模式迁移 | 分析型数据仓库、搜索引擎、文档数据库 | 平台工程 / CTO / CIO | 最适合 Cockroach Labs 的核心市场口径。 |
| 更广义的分布式数据库 | 事务型及部分非事务型分布式数据基础设施 | 单节点数据库和无关开发者工具 | 架构与基础设施负责人 | 有助于做市场合理性校验,但比公司范围更宽。 |
| 云数据库市场 | 跨 SQL、NoSQL 和多种部署模式的托管数据库服务 | 纯本地部署遗留支持合同 | 云平台负责人和财务赞助方 | 外层相邻市场,不是精确可服务市场。 |
| 维持现状的 PostgreSQL 资产 | 自管 OLTP、重度使用扩展的安装、应用侧分片 | 专为全球一致分布式 SQL 打造的功能 | 工程经理和 DBA | 主要的不作为替代方案。 |
在使用任何市场规模数字前,先界定范围。
[CM001, CM002, CM003, CM033]| 选项 | 买方为什么选择 | 为什么不是同一个市场 | 给 Cockroach Labs 的启示 |
|---|---|---|---|
| 自托管 PostgreSQL | 使用熟悉,扩展生态成熟 | 缺少多区域韧性所需的原生分布式 SQL 架构 | 最大惯性来源,也是迁移起点。 |
| Amazon Aurora PostgreSQL | AWS 内托管体验 | 往往优先优化单云便利性,而不是跨云可移植性 | 买方希望尽量少改时,它最难打。 |
| Google Spanner | Google Cloud 客户可获得强架构 | 把买方更紧地绑在超大规模云厂商和另一套生态选择上 | 大型战略交易里的强劲对手。 |
| 文档或分析型数据库 | 可能解决邻近数据问题 | 不能直接替代作为记录系统的事务型 SQL | 不应纳入核心总可用市场(TAM)主张。 |
现状替代方案很关键,因为多数买方可以接受更窄架构取舍,从而推迟迁移。
[CM002, CM003, CM018, CM020, CM021]2.2 市场规模口径及其分歧
公开市场规模证据覆盖三个嵌套但并不相同的市场。DataIntelo 的分布式 SQL 估算最贴近窄口径,因为它明确匹配 Cockroach Labs 的架构品类。Business Research Insights 给出更宽的分布式数据库口径,Coherent Market Insights 则覆盖大得多的云数据库大伞。这些口径放在一起有用,正是因为它们彼此不一致:范围本身揭示了品类边界扩张时带来的不确定性。 因此,投资者不应把单一头条 TAM 当作既定事实。$26B 的云数据库估算也许能为长期相邻市场提供方向感,但它并不等同于分布式 SQL 的可寻址资金池。反过来,低于 $5B 的分布式数据库估算,也可能漏掉全球一致 SQL 工作负载的特定价值溢价。使用多重口径,是为了保留矛盾,而不是把矛盾洗成虚假的精确。 这也限制了公开 SAM 和 SOM 工作。没有披露客户数、转化漏斗或按区域和工作负载拆分的组合,外部人无法负责任地推导公司特定的可服务市场。正确的尽调姿态,是保留这条失败的测算路径,并向管理层索取他们内部真正使用的细分。换句话说,这里的市场数字适合作为边界条件,而不是替代真实公司运营数据。[CM004, CM005, CM006, CM007, CM026, CM031]
| 发布方 | 年份 | 地域 | 数值 | CAGR | 方法 | 置信度 | 局限 |
|---|---|---|---|---|---|---|---|
| DataIntelo | 2025/2034 | 全球 | $7.2B 至 $24.8B | 14.7% | 分布式 SQL 品类估计 | 中 | 方法仅摘要披露,并不完全透明。 |
| Coherent Market Insights 市场研究 | 2026/2033 | 全球 | $26.0B 至 $73.0B | 15.9% | 更广义的云数据库市场 | 中 | 口径过宽,不适合作为公司 SAM。 |
| Business Research Insights | 2026/2035 | 全球 | $4.48B 至 $9.96B | 10.5% | 分布式数据库市场 | 中 | 不限于分布式 SQL。 |
| 内部尽调含义 | 2026 | n/a | 以分布式 SQL 作为基础口径 | n/a | 采用嵌套市场框架 | 高 | 真正的 SAM/SOM 需要管理层协助。 |
这些口径有意采用嵌套,而非混合,因为各品类不能互换。
[CM004, CM005, CM006, CM007, CM031, CM035]| 问题 | 证据 | 为什么会改变估值工作 | 暂定结论 |
|---|---|---|---|
| 品类口径不一致 | 云数据库、分布式 SQL、分布式数据库的估算差异很大 | 边界不同,总可用市场(TAM)可能被放大数倍 | 采用嵌套口径,不用单一混合估算。 |
| 方法论不透明 | 分析机构页面更多总结结果,而非方法 | 难以核验纳入哪些细分市场和部署假设 | 将数字视为方向性参考,不视为精确值。 |
| 缺少公开可服务市场(SAM) | 公司未公开按工作负载适配度划分的细分 | 仅靠外部数据无法隔离可捕获区域 | 向管理层索取内部细分。 |
| 缺少公开可获取市场(SOM) | 未披露客户数或赢率 | 无法支撑可信的份额假设 | 公开信息模型不应做精确市场份额假设。 |
本登记表保留矛盾,不把它们压成一个标题式总可用市场(TAM)。
[CM007, CM026, CM031, CM032, CM035]狭义分布式 SQL 视角最适合作为基准情景;更宽口径仍可用来做边界校验。
三个市场层级使用的来源方法和年份不同;该图只是框架工具,不是数学上可相加的漏斗。
[CM004, CM005, CM006, CM007, CM031, CM033]将点估算展示为上下限相同的区间,让差异可见,同时避免制造虚假精度。
每个已发布估算都是点值,这里用低 / 高相同的区间呈现;最后一行是解读,不是数值。
[CM007, CM026, CM035]2.3 买方地图、采用驱动与门槛约束
CockroachDB 的经济买方很少是单个开发者。能为这个产品证明合理性的数据库改造,通常涉及平台团队、数据库专家、安全 / 合规负责人,以及愿意用迁移投入换取更低宕机风险或更好地理扩展能力的高管赞助人。日常用户是工程师,但付款方往往是对韧性、云架构或受监管上线计划负责的领导者。 公开证据中的采用触发点很一致:多区域增长、无法容忍停机、摆脱脆弱分片的迁移需求,以及对合规或本地性控制的需求。PostgreSQL 兼容性有帮助,因为相比全新数据库模型,它降低了组织切换成本。客户案例反复说明,第一个工作负载通常是很痛的点——支付流、预订路径、欺诈控制服务或其他对延迟敏感的系统——之后才进入更广泛的标准化。 约束同样重要。迁移成本高,分布式系统中的运营行为变化并不简单,任何正确性公告或公开事故都可能拖慢企业信心。竞争也压缩了紧迫感:Aurora、Spanner、TiDB、YugabyteDB、Neon 和 PlanetScale 都给买方提供了推迟或重塑问题的办法。这让市场有价值,但并不无摩擦。因此,采用路径不像一次标准软件升级,更像一个董事会可见的可靠性项目,只有痛点足够具体时才拿得到预算。[CM008, CM009, CM010, CM011, CM012, CM013]
| 细分市场 | 买方 | 用户 | 付款方 | 工作流 | 预算负责人 | 采用触发点 |
|---|---|---|---|---|---|---|
| 金融服务 / 支付 | CTO 或平台副总裁 | 数据库工程师和应用团队 | 工程 / 转型预算 | 账本、支付、风险、受监管系统 | CIO / CTO | 韧性加合规要求。 |
| 旅行 / 预订 | 平台负责人 | SRE 和应用工程师 | 工程预算 | 预订和库存系统 | 工程副总裁 | 全球流量峰值和停机敏感性。 |
| 欺诈 / 商务基础设施 | 安全或数据平台负责人 | 应用工程师 | 产品 + 工程 | 欺诈决策和结账基础设施 | CTO / GM | 需要低延迟,且不能有单区域故障模式。 |
| 媒体 / 流媒体 / 消费平台 | 基础设施负责人 | 数据库和后端团队 | 平台预算 | 用户状态、目录和全球会话数据 | 基础设施副总裁 | 规模加多区域覆盖。 |
| AI 应用平台 | 架构负责人 | 平台和数据工程师 | 创新 / 平台预算 | 贴近向量工作流的操作型存储 | CTO | 需要事务核心加新 AI 功能。 |
买方和付款方角色基于产品定位及公开客户案例推断,并非来自已披露管线数据。
[CM010, CM011, CM012, CM013, CM016, CM027]| 驱动因素 / 约束 | 方向 | 时点 | 影响 | 尽调问题 |
|---|---|---|---|---|
| 多区域韧性需求 | 正向 | 当前 | 支撑关键任务应用的溢价价值叙事 | 衡量多少新增胜单明确提到避免宕机。 |
| PostgreSQL 兼容性 | 正向 | 当前 | 降低切换成本,扩大开发者相关性 | 索取 PostgreSQL 重度账户的迁移转化数据。 |
| 合规与数据主权 | 正向 | 当前 | 打开受监管工作负载和更大交易规模 | 要求披露受监管行业管线占比和销售周期长度。 |
| AI / 向量功能扩展 | 正向但仍早期 | 当前 | 可能扩大产品相关性,但不改变核心切入点 | 衡量实际与 AI 相关的生产部署。 |
| 迁移复杂度 | 负向 | 当前 | 放慢转化,也抬高验证负担 | 要求披露专业服务投入强度和上线投产时间。 |
| 正确性公告 / 信任审视 | 负向 | 当前 | 如果韧性承诺显得脆弱,采用速度会放慢 | 要求披露事件复盘流程,以及事件后的客户背书可用性。 |
正负力量并存:同一套韧性叙事既创造需求,也抬高信任审视。
[CM014, CM015, CM016, CM017, CM024, CM025]| 替代方案 | 定价姿态 | 买方吸引力 | Cockroach Labs 承压点 |
|---|---|---|---|
| Aurora | 按用量计费的托管服务 | 熟悉的 SQL,加上 AWS 原生运维 | 降低 AWS 中心团队的更换紧迫感。 |
| Spanner | 高端超大规模云厂商服务 | 全球规模,加上 Google 托管运维 | 在战略性云交易中树立很高的架构标杆。 |
| Neon / PlanetScale | 低摩擦无服务器入口 | 低成本试验,开发者主导评估 | 能截流更小或更早期的工作负载。 |
| TiDB / YugabyteDB | 分布式 SQL 替代方案 | 开源或多模式评估路径 | 抬高逐项功能比较的负担。 |
定价姿态是定性判断,因为公开资料没有把按工作负载同口径比较的价格标准化。
[CM019, CM021, CM022, CM023]| 阶段 | 主导角色 | 所需验证 | 摩擦点 |
|---|---|---|---|
| 识别问题 | CTO / 平台 VP | 故障痛点、地理扩展需求、合规要求 | 内部优先级可能仍然不高。 |
| 技术评估 | 平台 / 数据库团队 | 兼容性、性能、故障切换表现 | 基准测试难服众,迁移投入也高。 |
| 试点工作负载 | 应用团队 + SRE | 一个高痛点工作流证明价值 | 运维学习曲线。 |
| 标准化 | 工程领导层 | 可背书赢单和平台经济性 | 更大的迁移积压和组织产能。 |
这些阶段描述基于公开信息最可信的采用路径,不是公司披露的正式销售流程。
[CM013, CM014, CM015, CM024, CM034]将代表性细分市场映射到公开信息中最可能的买方、用户和付款方角色。
角色是根据客户故事和产品定位推断,而非来自披露的组织架构图。
[CM011, CM013, CM028, CM034]采用路径通常始于可靠性痛点或地理规模需求;首个生产验证跑通后,再向外扩展。
这是从多个公开案例研究推断出的通用流程,并非公司披露的销售方法论。
[CM014, CM015, CM024, CM034]03竞争格局
3.1 竞争集合:现状、超大规模云厂商与真正的分布式 SQL 同行
Cockroach Labs 并不是在空白市场销售。第一个、也常常是最强的竞争者,就是现状:团队可以继续运行 PostgreSQL,增加副本,并把迁移整体往后推。这一点重要,因为 PostgreSQL 仍然广受欢迎且可移植。即便 CockroachDB 对全球工作负载来说架构上可能更干净,组织层面采用更重分布式系统的紧迫感也会下降。 下一层竞争来自超大规模云厂商。Aurora、Spanner 和 AlloyDB 各自提供一条采购捷径:买方可以留在 AWS 或 Google Cloud 内,复用商业关系,并采用托管服务,而不引入独立数据库厂商。对许多企业账户来说,这种组织便利本身就是功能,甚至早于产品差异进入考虑。 只有穿过这两层之后,买方才会来到最直接的分布式 SQL 同行集合:YugabyteDB、TiDB,以及部分 CockroachDB 对 Spanner 的决策。PlanetScale、Vitess、Neon 和 SingleStore 又进一步拓宽决策集合,分别提供更低摩擦的开发者入口、分片 MySQL、serverless Postgres,或交易加分析混合定位。实际结论是,Cockroach Labs 争夺的不是泛泛的“数据库支出”,而是一类狭窄的、对正确性敏感的工作负载;只有多区域韧性足够痛,迁移才值得。[CP001, CP002, CP003, CP005, CP008, CP010]
| 竞争对手 | 类别 | 规模 / 所有权信号 | 目标细分 | 差异化 | 局限 |
|---|---|---|---|---|---|
| PostgreSQL | 现状方案 / 开源 | 社区主导的 OSS 生态 | 通用 OLTP 和广泛开发者使用场景 | 锁定效应最低,工具熟悉度最广 | 核心项目没有原生双活多区域写入路径 |
| Amazon Aurora | 既有托管关系型服务 | AWS 原生托管服务 | 需要托管 PostgreSQL / MySQL 兼容性的 AWS 客户 | 采购便利、无服务器选项、副本、Global Database | 仅 AWS 覆盖,多段式定价 |
| Google Spanner | 全球一致性既有方案 | Google 托管分布式数据库 | 大型 GCP 优先的全球应用 | 全球规模强一致性,加上托管运维 | 锁定效应高,基于副本的定价复杂 |
| AlloyDB | 托管 PostgreSQL 性能升级路径 | Google Cloud 托管产品 | 希望在 GCP 上获得更高性能和 AI 功能的 PostgreSQL 用户 | 100% PostgreSQL 兼容定位,配合读取池和 AI 叙事 | 仍绑定 Google 环境和节点式定价 |
| PlanetScale / Vitess | 分片 MySQL / 低摩擦云数据库 | 构建在 Vitess 之上的商业平台 | 优化开发速度或 MySQL 横向扩展的团队 | 低成本进入、HA 集群、BYOC 选项、经过验证的分片控制平面 | MySQL / Vitess 血统限制与 PostgreSQL 的直接重叠 |
| Neon | 无服务器 Postgres | 商业化托管 Postgres 平台 | 需要快速启动、由开发者主导的应用和智能体 | 缩至零、分支、计算与存储分离 | 不太主打全球双活事务语义 |
| YugabyteDB | 直接分布式 PostgreSQL 对手 | 开源加商业支持打法 | 以可移植性为目标的关键任务云原生 OLTP | 多主分布式 PostgreSQL 和开源反锁定叙事 | 叙事高度相似,带来正面替换风险 |
| TiDB | 带 HTAP 角度的分布式 SQL | PingCAP 托管或自托管 | 同时需要事务和实时分析的团队 | MySQL 兼容、计算与存储解耦、HTAP 框架 | PostgreSQL 迁移叙事较弱 |
| SingleStore | 邻近实时数据平台 | 托管云 DBaaS | 融合 SQL、JSON、向量和分析的实时应用 | 事务与分析混合广度,加上强性能主张 | 不是每类工作负载都能干净替代 PostgreSQL 记录系统 |
| MongoDB Atlas | 邻近现代应用替代方案 | 托管文档数据库平台 | 偏好模式灵活性和现代应用工具的构建者 | 强现代应用和 AI 品牌 | 文档模型不是关系型数据库的即插即用替代品 |
画像聚焦公开所有权、平台姿态和目标细分,不试图为每个对手还原精确的私募融资情况。
[CP001, CP002, CP005, CP008, CP010, CP012]基于证据支持的序位判断,把主要替代方案放在「部署自由度」与「捆绑式运营简便性」两条轴上。
轴值是根据公开产品、定价和部署描述得出的序位放置,不是公司披露的基准测试分数。
[CP002, CP005, CP008, CP010, CP015, CP017]3.2 能力重叠:Cockroach 仍突出的地方与对手追赶的地方
能力重叠真实存在,但并不均匀。CockroachDB 最清楚的主张,仍是把 PostgreSQL 兼容性、分布式 ACID 行为、主动 - 主动多区域设计,以及跨云或自托管部署自由组合在一起。这套组合仍相对少见,尤其是与标准 PostgreSQL 和 Aurora 相比;后两者更容易采用,但原生分布式能力更弱。 问题在于,卖点中的重要部分已经不再独一无二。AlloyDB 打着 PostgreSQL 标签,同时在 Google Cloud 上加入托管性能和 AI 功能。Yugabyte 用开源和多云语言销售分布式 PostgreSQL,叙事非常接近 Cockroach 的核心故事。TiDB 销售强一致性加水平扩展,但走的是 MySQL 兼容口径。PlanetScale、Vitess 和 Neon 并不在每个正确性或多区域维度上匹配 Cockroach,但它们能靠更低运营摩擦和起步成本赢下早期项目或单团队决策。 因此,仅靠功能广度不是护城河。买方越来越能组出一张短名单,每个厂商各赢一个取舍切片:可移植性、超大规模云厂商便利、serverless 开发者速度、避免开源锁定,或多模型数据处理。当工作负载需要全球事务正确性且不愿接受单云锁定时,Cockroach Labs 仍显得最强;但公司现在必须证明这种需求,而不能假设需求自然存在。[CP004, CP009, CP014, CP015, CP017, CP020]
| 维度 | CockroachDB | PostgreSQL | Aurora | Spanner / AlloyDB | Yugabyte | PlanetScale / Neon / TiDB |
|---|---|---|---|---|---|---|
| PostgreSQL 兼容性 | 核心定位强 | 原生基线 | 兼容性高,但受 AWS 形态影响 | 混合:AlloyDB 强,Spanner 只有部分 PG 接口 | YSQL 姿态下较强 | 混合:Neon 强,PlanetScale / TiDB 较弱 |
| 双活多区域写入 | 核心定位 | 核心项目较弱 | 受限 / 取决于拓扑 | Spanner 强,AlloyDB 较弱 | 竞争重叠强 | 通常较弱或不是核心 |
| 云中立性 | 强 | 强 | 弱 | 弱到中 | 强 | 中 |
| 开放生态 / 锁定安心度 | 中 | 强 | 弱 | 弱 | 强 | 混合 |
| 开发者进入摩擦 | 中 | 留在原位则低 | AWS 内部低 | 中到高 | 中 | Neon 和 PlanetScale 低 |
| 分析 / 多模型邻近性 | 中 | 中 | 中 | 借 AI 和读取扩展增强 | 中 | TiDB 和 SingleStore 强 |
| AI / 向量营销强度 | 上升 | 核心项目低 | 这些来源显示有限 | 高 | 高 | 高 |
矩阵使用有证据支撑的序数标签,不假装所有功能都能在不同架构和部署模型之间直接比较。
[CP004, CP009, CP010, CP014, CP015, CP017]展示主要替代方案各自把优势押在哪里,而不是假装所有产品都为同一任务优化。
[CP010, CP014, CP015, CP017, CP020, CP021]3.3 定价、锁定与护城河耐久性
定价模型说明,竞争危险不止来自一个方向。Aurora、Spanner 和 AlloyDB 大多通过实例、节点、副本、存储和网络复制等企业式基础设施构件变现。这适合稳态生产工作负载和企业采购习惯,但也会产生多行项目账单和依赖拓扑的成本。相比之下,Neon 和 PlanetScale 用免费或低成本起点与按量扩张,让入门更便宜、更适合开发者。SingleStore 和部分 TiDB 产品介于两端之间,把托管服务经济性与对工作负载敏感的用量或云打包混在一起。 这些定价差异强化了锁定格局。PostgreSQL、Vitess、Yugabyte 和 TiDB 等开源或开放生态选项,可以缓解买方对不可逆平台依赖的焦虑。Aurora 和 Google 技术栈则用其中一部分自由,换取采购便利和相邻云集成。PlanetScale 和 Neon 的强项并不是完美替代 CockroachDB;而是能在买方得出“必须要 Cockroach 级分布语义”之前,先截获工作负载。 因此,护城河是有条件的,不是绝对的。Cockroach Labs 并不独占分布式数据库理念、AI 叙事或 PostgreSQL 熟悉度。它的护城河来自证明:对一部分关键任务应用来说,一个云无关、正确性优先、多区域的 PostgreSQL 系统,明显优于留在现状或走打包的超大规模云厂商路线。这可以很耐久,但前提是迁移工具、客户证据和销售聚焦能让公司领先 Yugabyte 等直接同行,同时守住 Neon 和 PlanetScale 提供的更便宜试点体验。[CP006, CP007, CP011, CP012, CP013, CP016]
| 竞争对手 | 入门模式 | 扩展模式 | 合同 / 隐性成本信号 | 给 Cockroach Labs 的启示 |
|---|---|---|---|---|
| PostgreSQL | 免费 OSS 软件 | 基础设施 + 人力 | HA 和工具需另行拼装 | 仅靠成本最难撬动的现状方案 |
| Aurora | 按需或无服务器 ACU | 实例 + 存储 + I/O + 副本 + Global DB 成本 | 特定区域承诺和复制写入 I/O | 买方更看重 AWS 便利性而非可移植性时胜出 |
| Spanner | 处理单元 / 节点 | 计算 + 存储 + 备份 + 复制 + 带宽 | 副本数量和最低计费窗口很重要 | 强大但企业形态明显的全球扩展选项 |
| AlloyDB | 预配置托管集群 | vCPU + 内存 + 存储 + 备份 + 网络 | HA 使用两个节点;CUD 可改善经济性 | 面向已经在 GCP 上的严肃 PostgreSQL 工作负载 |
| PlanetScale | Postgres $5/mo 起,或低成本 HA | 按资源月度定价,加企业附加项 | BYOC 将基础设施成本转入买方账户 | 低成本截流开发者试点 |
| Neon | 免费层和付费 CU-hour 用量 | 按用量计费的计算 + 存储 + 历史存储 | 自动扩缩和分支会改变实际成本形态 | 让无服务器 Postgres 试点非常容易 |
| Yugabyte / TiDB | 免费 OSS 加托管云产品 | 随集群规模或托管服务增长 | 商业支持选择决定真实成本 | 不靠超大规模云厂商锁定,直接打可移植性 |
| SingleStore | 按需用量 | 积分 + 存储 + 可选 Flow CDC 费用 | 高摄取量用例会迅速推高成本 | 实时 AI / 分析覆盖面重要时参与竞争 |
定价行把供应商定价计算器简化成最影响买方行为和获客摩擦的计费原语。
[CP005, CP006, CP007, CP008, CP011, CP012]| 护城河主张 | 威胁 | 严重性 | 为何真实存在 | 缓解动作 / 尽调问题 |
|---|---|---|---|---|
| PostgreSQL 熟悉度为迁入 CockroachDB 输送需求 | 同样的熟悉度也支持继续留在 PostgreSQL,或改选 Aurora / AlloyDB | 高 | 现状方案和托管 Postgres 选项让组织变化维持低位 | 向管理层索取按来源数据库拆分的具体替换数据 |
| 云中立分布能力具备差异化 | Spanner、AlloyDB Omni、Yugabyte 和 BYOC 产品会在许多账户中缩小这一差距 | 高 | 多个对手现在公开主打可移植性或混合部署叙事 | 检验买方是否真的需要跨云写入,还是只需要云采购灵活性 |
| 分布式 SQL 构建难度高 | Yugabyte 和 TiDB 已提供可信的分布式替代方案,Vitess / PlanetScale 也能更早解决许多扩展问题 | 高 | 概念独特性已经被削弱 | 要求提供对 Yugabyte、TiDB、PlanetScale 和 Aurora 的赢单 / 输单明细 |
| AI 时代让数据库重新相关,需求面会变宽 | AlloyDB、Yugabyte、TiDB、SingleStore 都在讲 AI 和向量能力 | 中 | 公开话术显示,AI 定位已是全品类叙事 | 把真实客户用量和营销叙事拆开看 |
| 企业级证明应能守住定价 | 超大规模云厂商捆绑销售、低门槛无服务器试点,仍可能压低获客成本 | 中 | 架构优势生效前,分发能力和低价切入更要紧 | 按竞品类别检查销售周期数据 |
| 迁移工具带来粘性 | 公开证据无法证明具体竞争胜率或转化经济性 | 中 | 外部无法量化工具是否实质改变成交率 | 按迁移路径和部署目标索取管线转化数据 |
风险清单关注的是护城河能否经受真实买方行为检验,而不是单看 CockroachDB 技术上是否足够亮眼。
[CP026, CP028, CP029, CP030, CP032, CP033]一张按证据加权的评分卡,评估 Cockroach Labs 相对当前替代方案的竞争耐久度。
[CP019, CP028, CP029, CP030, CP033, CP035]3.4 展示材料
04财务情况
4.1 变现架构横跨开发者漏斗、生产集群与高接触服务
Cockroach Labs 卖的不是一个干净的 SKU。它的变现栈有意横跨低摩擦开发者漏斗、预置云生产计划,以及一组帮助客户迁移和运营数据库的企业服务层。Basic 计划是最清楚的自助入口:按请求单元用量计费,并给出有意义的免费额度,降低了无重承诺起步的难度。Standard 和 Advanced 随后把初始漏斗转成明确基础设施支出,Standard 销售预置 vCPU 容量加用量项目,Advanced 则通过专用节点、存储以及某些情况下的 IOPS 变现。 一旦把支持和迁移纳入,模型就更有意思。公开支持页面显示,Cockroach Labs 销售分层订阅,对企业云支持收取额外百分比费用,并把技术顾问、具名成功覆盖和根因分析打包进更高层级。迁移文档和 MOLT 工具也说明,公司预期许多大客户会通过替换项目进入,而不只是通过全新绿地应用进入。 这对可变现客户深度是好信号,但也意味着业务并非纯软件自动化。相当一部分价值创造似乎绑定了解决方案工程、迁移专长和企业运营信任。从财务上看,Cockroach 更像云基础设施经常性收入与高接触企业交付的混合体,而不是简单的免费增值数据库工具。[CI001, CI002, CI004, CI009, CI012, CI013]
| 收入流 | 计费单位 | 客户类型 | 扩张驱动因素 | 含义 |
|---|---|---|---|---|
| Basic 云 | 请求单位(RU)+ 存储,含免费额度 | 开发者和小型工作负载 | 请求负载和存储数据增加 | 承担漏斗顶端获客和轻量用量变现 |
| Standard 云 | 预置 vCPU-hours 加用量明细项 | 生产级云工作负载 | 更高预留容量、存储、备份、CDC、传输 | 类似带扩张杠杆的经常性基础设施收入 |
| Advanced 云 | 按节点计费的计算、存储、IOPS,加用量项 | 企业级和受监管工作负载 | 更多节点、更大节点、复制副本、安全功能、支持 | ACV 可能最高,但交付也最定制 |
| 自托管企业版 | 许可 / 订阅式商业动作 | 需要自有环境的客户 | 更多集群、支持和架构范围 | 在 Cockroach 托管云之外保留企业级变现 |
| 支持订阅 | 分层服务订阅;云企业版费用百分比 | 生产工作负载运营方 | 更高严重级别保障和企业运营需求 | 增加粘性服务收入,也强化留存 |
| 专业服务 / 迁移 | 项目制或限定范围服务 | 替换和现代化项目 | 迁移复杂度和工作负载关键性 | 提高成交率,但增加交付成本 |
各行概括定价、支持、部署和迁移材料中公开可见的变现界面。
[CI001, CI002, CI004, CI009, CI012, CI013]| 方案 / 杠杆 | 公开计费机制 | 鼓励的客户行为 | 隐性或可变成本信号 | 财务含义 |
|---|---|---|---|---|
| Basic | 按用量计费的 RU,含每月免费额度 | 原型验证和突发式试用 | RU 套餐吸收备份和传输 | 有利于获客漏斗,但小账户利润率更不透明 |
| Standard 计算 | 预置 vCPU-hours 按预留容量计费 | 容量规划和长期生产使用 | 超配缓冲可能产生闲置支出 | 支撑可预测的经常性收入 |
| Standard 存储 / CDC / 传输 | 按用量计费的明细项 | 工作负载增长或全球化后扩张 | 跨区域流量和被监控数据可能让买方意外 | 落地后产生扩张收入 |
| Advanced 节点 | 按节点计费的计算和存储 | 企业架构规划和区域设计 | AWS IOPS 和安全附加项增加复杂度 | ACV 更高,商业动作也更定制 |
| 企业支持 | 带云支持附加费的订阅层 | 运营信任和更快响应预期 | 服务强度可能随云消费增长 | 增加经常性收入,也增加服务成本 |
| 迁移服务 | 围绕替换项目的高接触度服务 | 复杂切换和数据库整合 | 交付和售前工程都吃人力 | 推高 CAC,但也提高大单胜率 |
定价机制从文档简化为它最可能驱动的商业行为。
[CI002, CI003, CI004, CI009, CI010, CI013]展示 Cockroach Labs 如何把低摩擦云入口转化为更高价值的生产基础设施和服务收入。
[CI001, CI002, CI004, CI009, CI012, CI014]4.2 公开单位经济信号强在定价机制,而非效率披露
最好的公开财务证据是机制性的,而非完整的。Cockroach Labs 清楚解释 Standard 和 Advanced 客户如何计费、40 percent 容量缓冲是什么样、跨区域定价何时适用,以及哪些扩张杠杆位于基础计算之外。这很重要,因为它告诉投资者,初始采用之后收入可以从哪里扩张:备份留存、watched-data CDC、跨区域传输、额外副本和专用企业支持,都可以叠加在集群本身之上。 但同样的公开清晰度,也揭示了外部人很难承销单位经济的原因。Standard 按预留容量而非实际用量计费,这有利于收入可预测性,但也可能带来客户闲置或优化压力。Advanced 更加定制化,定价会受区域、云服务商、安全要求,有时还受 IOPS 影响。这些项目可能在成熟账户里产生有吸引力的净扩张,但也让毛利率对基础设施组合和多区域流量更敏感。 可见的 GTM 动作进一步放大了这种模糊性。MOLT、迁移服务和顾问式支持帮助 Cockroach 拿下困难的替换项目,但也暗示未单独披露的获客和实施成本。因此,公开证据支持 Cockroach 拥有多条变现杠杆的判断,却不足以支撑精确的 CAC、回本周期或毛利率模型。[CI005, CI006, CI007, CI008, CI010, CI011]
| 经济性领域 | 公开信号 | 可能收入驱动 | 可能成本驱动 | 置信度 |
|---|---|---|---|---|
| 开发者漏斗 | Basic 免费用量和请求单位计费 | 从试用转向生产 | 免费层支持和被吸收的基础设施成本 | 中 |
| Standard 云生产 | 预留 vCPU 容量,建议留 40% 缓冲 | 经常性计算预留 | 即使客户未充分使用预留容量,也要承担云基础设施 | 中 |
| Advanced 企业版 | 按节点定价,并随安全和 IOPS 变化 | 更大的受监管或多区域账户 | 专用基础设施和支持强度 | 中 |
| 支持订阅 | 分层 SLA,含指定服务和云支持附加费 | 运营附加项和留存 | 熟练支持人力和事件管理 | 中 |
| 迁移工具包 | Schema 转换、加载、复制、验证、回退 | 大型现代化项目 | 解决方案工程和服务投入 | 中 |
| 扩张用量 | 备份、CDC 被监控数据、跨区域传输 | 部署后工作负载增长 | 云出站、存储和运营复杂度 | 中 |
这张表带有推断性质,因为公开来源对计费机制的描述远比利润率结果精确。
[CI006, CI008, CI013, CI014, CI015, CI026]追踪客户工作负载增长如何转化为可计费扩张杠杆,同时标出公开利润率披露止步之处。
[CI006, CI010, CI013, CI014, CI026, CI027]将可见收入杠杆与可能的交付成本、披露清晰度对应起来。
[CI001, CI013, CI014, CI026, CI029, CI035]4.3 牵引力与资本背景真实存在,但当前承销输入仍稀疏
Cockroach Labs 不是一家还在等待第一份证明的小公司。2021 年 Series F 公告披露了强劲 ARR 增长、云收入快速加速,以及超过 200 个客户,其中逾半数已在 Dedicated 上运行。第三方汇总后来披露,2023 年 ARR 或收入数字为 128.3 million dollars,2025 年员工数估计约 715 人。这些数据点支持一种判断:Cockroach 已经在有意义的后期规模上运营。 缺的是投资者现在最想要的东西:新收入、烧钱速度、现金余额、毛利率、留存和效率数据。2026 年进展公告强调产品和伙伴进展,却没有发布更新后的财务。二级市场定价显示投资人兴趣仍在,估值高于 2021 年一级融资;但二级交易不会注入新的运营现金,也不能回答现金跑道问题。 结果是一条定性信号强、精度弱的财务故事。收入质量看起来有希望,因为产品位于关键任务基础设施,云收入占比早期已有改善,支持或迁移服务似乎可以变现。但缺少当前财务报表意味着,盈利路径和 2021 年资本基础是否足够,仍只能推断,不能证明。这足以支持继续尽调,但不足以高置信度承销。[CI018, CI019, CI020, CI021, CI022, CI023]
| 信号 | 公开数值 | 日期 | 它说明什么 | 局限 |
|---|---|---|---|---|
| 最近一轮新股融资 | $278M Series F 轮,估值 $5B | 2021-12-16 | 大额后期资产负债表融资和估值锚 | 作为运营资金指标已经滞后 |
| 官方累计融资 | $633M | 2021-12-16 | 公司进入 2022 年后仍有实质资本缓冲 | 当前现金余额未公开 |
| 老股估值信号 | 隐含约 $6.9B | 2026-06 | 投资人仍给予高溢价私募市场价值 | 没有新增现金进入公司 |
| 私募市场流动性 | Nasdaq Private Market 和 Notice 上有活跃报价 | 2026-07 | 不像明显冻结或承压 | 持有人流动性不等于现金跑道披露 |
| 员工数代理指标 | 约 715 名员工 | 2025-11-28 | 暗示可观运营成本基数 | 仅为第三方估计 |
| 2026 年财务披露 | 未公开更新 ARR、烧钱速度或现金指标 | 2026-07-20 | 核心承销缺口仍在 | 无法精确判断现金跑道 |
由于缺少当前资产负债表披露,资本充足性只能靠融资历史、老股定价和规模代理指标三角校验。
[CI020, CI021, CI022, CI023, CI024, CI025]| 指标 | 最佳公开数值 | 来源质量 | 重要性 | 尽调要求 |
|---|---|---|---|---|
| 当前 ARR / 收入 | 2023 年 $128.3M | 第三方聚合方 | 增长和估值精度需要该指标 | 索取 2024 和 2025 年 ARR 及云收入占比 |
| 毛利率 | 未公开披露 | 缺失 | 决定基础设施软件质量 | 按方案索取云毛利率 |
| 烧钱速度 / 现金跑道 | 未公开披露 | 缺失 | 评估融资依赖需要该指标 | 索取现金余额、烧钱速度和现金跑道假设 |
| NRR / GRR | 未公开披露 | 缺失 | 证明扩张驱动模型韧性需要该指标 | 索取队列和续约指标 |
| CAC / 回本周期 | 未公开披露 | 缺失 | 判断顾问式 GTM 效率需要该指标 | 索取全口径获客和部署成本指标 |
| 支持 / 服务组合 | 定性可见,但未披露数字 | 部分 | 需要把软件利润率和服务利润率拆开 | 索取按云、许可、支持和服务划分的收入结构 |
缺口清单才是真正的承销阻塞点:公开证据显示 Cockroach 如何销售,但不足以说明它卖得有多高效。
[CI020, CI022, CI035, CI036, CI038]给少数公开财务锚点划出区间,并标出 2026 年证据集仍然稀薄的地方。
[CI013, CI020, CI021, CI023, CI024, CI025]4.4 展示材料
05产品与技术
5.1 产品定义与 SKU 地图
CockroachDB 的产品定义比“分布式 SQL”更具体。公开材料始终把它描述为面向永久在线、云原生应用的事务记录源数据库;这些应用同时需要 PostgreSQL 熟悉度、韧性和全球规模。这种工作流取向很重要,因为它把 CockroachDB 同纯分析系统,以及解决不同任务的文档或多模型平台区分开来。 产品面也比单一引擎更宽。Cockroach Labs 围绕同一个数据库核心,销售托管云、自托管企业软件、迁移工具和支持。Basic 或类似 serverless 的入口用于降低试用摩擦;Standard 和 Advanced 则面向企业账户中要求逐步提高的生产和受监管工作负载、采购路径与安全预期。 这张模块地图揭示了公司的战略意图:用熟悉度和低门槛起步进入工作流,再扩张到更高价值的生产、合规和迁移工作。因此,产品故事不只是功能数量,而是把一个困难的替换品类,转成一项可以从小处开始、仍能扩展成关键任务控制平面的平台决策。[CE001, CE002, CE003, CE026, CE027, CE031]
| 模块 / SKU | 部署形式 | 买方核心需求 | 重要能力 | 证据状态 |
|---|---|---|---|---|
| Basic / 无服务器入口 | 托管云 | 快速启动和低摩擦实验 | 以用量计费切入分布式 SQL | 公开可见 |
| Standard | 托管云 | 带预置容量的生产工作负载 | 预置多区域集群规划 | 公开可见 |
| Advanced | 托管云 | 受监管或高控制生产环境 | 专用节点、更深安全和合规能力 | 公开可见 |
| Enterprise 自托管 | 客户环境 | 控制权、可移植性和自有环境运营 | Cockroach 托管云之外的完整功能面 | 公开可见 |
| MOLT 工具包 | CLI / 迁移工具 | 替换遗留数据库 | Schema 转换、加载、复制、验证 | 公开可见 |
| 支持 / 服务 | 运营叠加层 | 更快问题解决和迁移帮助 | 专属 Slack、咨询、性能调优 | 公开可见 |
模块图聚焦客户实际购买或使用什么,而不是每一个内部包装细节。
[CE002, CE003, CE024, CE031]| 工作流或用例 | Cockroach 为什么适配 | 最适合方案 / 模块 | 运营取舍 |
|---|---|---|---|
| 始终在线 OLTP | 强一致性加多活韧性 | Standard 或 Advanced | 需要容量和地域属性规划 |
| 全球消费者应用 | 按行 / 表做区域化,加生存目标 | Advanced | 跨区域流量和合规选择很关键 |
| 摆脱遗留关系型资产的现代化 | PostgreSQL 兼容性加 MOLT 迁移路径 | Enterprise + MOLT | 切换和 schema 转换仍需严谨执行 |
| 受监管工作负载 | CMEK、合规姿态、行级安全、身份集成 | Advanced | 安全功能增加设置和治理工作 |
| 靠近运营数据的 AI 语义搜索 | 内置向量索引,带分布式存储语义 | Standard / Advanced | 功能集仍在成熟,且由厂商自述 |
| 开发者评估或原型验证 | 低摩擦托管启动 | Basic / 无服务器 | 单独不能证明适合企业生产 |
用例映射把数据库功能翻译成买方真正想完成的任务。
[CE001, CE003, CE009, CE015, CE019, CE020]连接面向客户的产品模块、底层数据库核心和服务层。
[CE001, CE002, CE003, CE024, CE031]5.2 核心架构与部署模型
CockroachDB 的技术身份从架构开始。数据库在任意节点接收 SQL,把这些语句转换成键值操作,再通过分布在集群中的 range 进行协调和复制。数据增长时 range 会拆分,写入通过 Raft quorum 流动,本地性元数据加 range 放置让系统在不要求应用自行管理分片的情况下,为韧性和地理位置优化。 多区域控制不是事后补丁。文档暴露了主区域、表本地性、超级区域、次级区域和生存目标等明确抽象。客户可以用 SQL 可见的术语在延迟、合规和韧性之间取舍,而不只是看网络拓扑图。集群虚拟化沿着同一方向推进,用系统集群和虚拟集群分离控制平面与数据平面;这对物理集群复制、资源隔离和更可治理的运营模型很重要。 这一层最难由客户自己重建。PostgreSQL 熟悉度相对容易理解;真正困难的是在一个产品里把这种熟悉度与多区域拓扑、一致复制、可移植性、低接触运营和更新的 AI-ready 索引结合起来。底层架构的差异化仍在这里。[CE004, CE005, CE006, CE007, CE008, CE009]
| 层级 | 作用 | 公开证据 | 重要性 |
|---|---|---|---|
| SQL 层 | 任意节点都接收 SQL,并规划分布式执行 | 架构概览 | 让产品保持关系型和熟悉感 |
| KV 存储 | 把 SQL 状态转换为分布式键值范围 | 架构概览 | 支持弹性拆分和再平衡 |
| 复制层 | 通过 Raft quorum 复制 range | 架构概览 | 驱动一致性和可生存性 |
| 多区域抽象 | 增加区域、locality、super region、secondary region、survival goal | 多区域概览 | 让地理位置和韧性可以用产品术语配置 |
| 集群虚拟化 | 把控制平面和数据平面拆成 system cluster 与 virtual cluster | 集群虚拟化概览 | 支持更隔离的运营拓扑 |
| 向量索引层 | 在 range 中存储分区向量索引结构 | C-SPANN 材料 | 不用单独引擎,也把产品延展到 AI 原生搜索 |
这张表把实现层抽象成买方和运营方关心的运营概念。
[CE004, CE005, CE006, CE008, CE011, CE015]展示从迁移或原型到受监管多区域运营的路径。
[CE003, CE008, CE020, CE024, CE035]映射让 CockroachDB 强大、但也让运营变复杂的主要产品依赖。
[CE006, CE008, CE011, CE016, CE019, CE022]5.3 信任、质量与运营控制
产品成熟度故事不仅关乎原始性能,同样关乎控制面。安全和信任材料描述了一套包含加密、合规项目、身份集成、CMEK、行级安全和负责任披露流程的栈。Release 26.1 进一步推进,加入 Azure 合规参与、自动用户预配、JWT 和 OIDC 角色同步,以及原生 FIPS 140-3 支持——这些信号说明公司争取的是受监管企业,而不只是成长型初创公司。 运营透明度是公开可见的。Cockroach Labs 发布技术公告、负责任披露政策和公开事故历史。这种透明度是优势,但也确认了复杂分布式数据库的普通现实:系统足够强大,bug、安全问题和细微运营边界情况就不可避免。因此,信任取决于响应质量、可重复工具和平台纪律,而不是假装架构很简单。 同样的张力也出现在更宽的产品广度中。迁移、CDC、多区域设置、向量索引、AI-agent 治理、虚拟化和韧性基准测试,都会提高严肃买家的产品价值,也会扩大运营表面积。客户真正该问的不是 CockroachDB 是否技术丰富——显然是——而是这种丰富度能否降低目标工作负载的净复杂度,还是只是把复杂度搬进一个更集中的平台。公开材料有力说明韧性已被深度产品化;但它没有消除谨慎运营、分阶段上线纪律和懂行操作人员的必要性。[CE019, CE020, CE021, CE022, CE023, CE024]
| 控制领域 | 公开控制项 | 买方为什么关心 | 剩余注意点 |
|---|---|---|---|
| 身份和访问 | JWT / OIDC 角色同步、自动用户开通、LDAP/AD 体系继承 | 减少手工安全运营 | 配置仍然关键 |
| 加密与密钥管理 | CMEK 与加密控制 | 支撑受监管云部署 | 云端专属配置仍有复杂度 |
| 合规态势 | HIPAA、PCI、SOC 2、ISO、FIPS 合规口径 | 缩短企业安全评审 | 范围随计划和云而异 |
| 数据主权 | 行级地域化、超级区域、司法辖区绑定 | 支撑数据驻留规则 | 延迟与拓扑的取舍必须谨慎选择 |
| 披露纪律 | 负责任披露政策与安全公告 | 释放成熟度和透明度信号 | 也暴露反复出现的缺陷面 |
| 事件可见性 | 公开云事件历史 | 体现运营问责 | 证实宕机风险不是零 |
信任态势不只看合规徽章,而是产品控制与运营披露共同撑起。
[CE019, CE020, CE021, CE022, CE023, CE034]| 领域 | 2025-2026 年公开信号 | 阶段或成熟度线索 | 战略含义 |
|---|---|---|---|
| 吞吐量改进 | 25.2 称九类工作负载平均吞吐量提升 50% | 近期发布口径 | 性能仍是路线图主轴 |
| 缓冲写入 | 25.2 中预览,称提升 15-40% | 预览 | 写入路径效率仍在提升 |
| 向量索引 | 25.2 中预览,并披露更深的 C-SPANN 内部机制 | 预览阶段,但战略重要 | AI 搜索用例正成为一等场景 |
| 行级安全 | 25.2 GA | 已 GA | 受监管应用的租户隔离更强 |
| 身份感知访问 | 26.1 加入 JWT 与 OIDC 自动化 | 近期发布 | 强化企业安全态势 |
| FIPS / 合规 / Azure 支持 | 26.1 合规与原生 FIPS 支持 | 近期发布 | 进一步打入受监管工作负载 |
路线图信号来自厂商自写发布说明,更适合判断方向和成熟度线索,不能直接证明客户层面的结果。
[CE013, CE014, CE015, CE017, CE019, CE020]区分产品内成熟核心能力和较新的扩展面。
[CE015, CE019, CE022, CE024, CE025, CE031]5.4 展示材料
06客户情况
6.1 客户基础 logo 面广,但集中在关键任务用例
Cockroach Labs 并不缺具名客户证明。公开客户页面覆盖金融服务、零售、软件、媒体、游戏、制造和博彩;更早的官方披露在 2021 年就已经指向超过 200 个客户和数万个已部署集群。这个组合很重要,因为它同时说明了广度和时间长度:这不是一个仍在等待第一批严肃背书的产品。 但更重要的模式不是 logo 数量,而是工作负载类型。在可见客户名单中,CockroachDB 反复用于订单管理、支付、欺诈筛查、设备控制平面、游戏元数据、发货跟踪和 sportsbook 账本。这些都是运营系统,停机、错误写入或区域故障会立刻伤害收入或客户体验。 这种集中度有两面性。它强有力地证明,CockroachDB 找到了有真实痛点和真实预算的买方;也说明产品最适合技术要求高、愿意为了韧性、一致性和拓扑控制承受分布式数据库复杂度的客户。换句话说,这份客户名单更像企业基础设施客户账本,而不是轻量、广谱的 SMB 工具基础。这个模式有利于 ACV 和留存,但不那么明显支持大众市场规模或无摩擦自助广度。[CU001, CU002, CU003, CU004, CU005, CU006]
| 垂直行业 | 具名案例 | 核心工作负载 | 重要性 |
|---|---|---|---|
| 银行与金融科技 | Form3、SumUp、未具名 Fortune 50 银行 | 支付、清算网络连接、受监管交易处理 | 说明适配对正确性敏感的受监管工作负载 |
| 零售与电商 | Shipt、Route、Riskified | 支付、反欺诈、订单与配送数据 | 证明能顶住 7x24 消费流量的运营需求 |
| 旅游出行 | Booking.com | 预订订单管理 | 展示复杂全球交易编排能力 |
| 媒体与流媒体 | Netflix | 设备平台与工作流编排 | 显示顶尖内部平台团队采用 |
| 游戏与博彩 | Hard Rock Digital、Superbet、Netflix 游戏 | 状态敏感的博彩和游戏控制平面 | 支撑监管合规与多区域定位 |
| 软件 / 技术平台 | DoorDash 内部 DBaaS、Route 平台服务 | 内部开发者平台与运营控制平面 | 暗示可从单一应用部署向外扩张 |
垂直覆盖强调具名或可合理推断的生产环境细分,而不是营销网站上的每个客户标识。
[CU001, CU006, CU014, CU021, CU024, CU029]| 来源 | 公开数据点 | 支撑的判断 | 注意事项 |
|---|---|---|---|
| F 轮新闻稿 | 200+ 客户;数万个集群 | 说明历史商业基础不小 | 到 2026 年已过时 |
| F 轮博客 | 50%+ 客户使用 Dedicated | 托管云采用深度 | 同样是历史数据 |
| 客户页面 | 覆盖多行业的五页名单 | 具名证据广度 | 没有准确数量或支出 |
| Apps Run The World | 方向性部署清单与企业案例 | 独立佐证覆盖广度 | 方法论不完全透明 |
| ReadyContacts | 营销口径的 226 家公司名单 | 方向性账户范围 | 获客线索来源,不是审计披露 |
| 业务进展发布稿 | 强调企业客户与伙伴扩张 | 商业活动仍在推进 | 没有最新付费客户数 |
此表把客户名单广度与准确客户数分开看。
[CU002, CU003, CU004, CU005, CU040]展示高要求企业客户如何从既有数据库痛点出发,走向账户内部扩张的常见路径。
这是分析师基于公开案例研究中反复出现的模式归纳出的客户旅程,不是 Cockroach Labs 披露的生命周期漏斗。
[CU001, CU006, CU024, CU029, CU032, CU034]6.2 最强客户背书显示深度生产采用,而非轻量试点
具名案例研究异常具体。DoorDash 将 CockroachDB 作为内部服务运行,峰值查询约每秒 1.2 million 次、超过 300 个集群,数据量约 1.9 petabytes。Netflix 运营超过 380 个集群,并把使用扩展到设备管理、工作流编排和游戏。Route 支撑超过 1 billion 个订单,Booking.com、Form3、Shipt、Riskified 和 Hard Rock 都描述的是业务关键交易流,而不是边缘项目。 这很重要,因为它降低了基础设施软件中常见的尽调担忧:纸面上好看的背书,其实只是小试点或非关键影子部署。这里的公开故事指向事实源系统、支付引擎、面向客户的订单流和内部数据库平台。这些工作负载更难赢下、更难治理;一旦成功运行,也更难被替换。 反复出现的主题是,CockroachDB 常常从一个困难架构问题切入——全球交易、监管本地性或传统数据库瓶颈——随后逐步变得更核心。多则案例还显示,初始决策之后部署继续扩张,可能是更多集群、更多区域、更多节点,或更广泛的内部自助服务。因此,这些背书不仅表明采用,也表明在高要求环境中采用后的运营信任。客户证据因此有真实分量。[CU007, CU008, CU009, CU010, CU011, CU012]
| 客户 | 公开规模信号 | 工作负载类型 | 信号质量 |
|---|---|---|---|
| DoorDash | 1.2M 峰值 QPS;300+ 集群;1.9PB;~900 条 changefeed | 内部数据库平台 | 高,但由厂商撰写 |
| Netflix | 380+ 集群;160+ 生产;60+ 多区域 | 内部 DBaaS 与控制平面 | 高,但由厂商撰写 |
| Route | 1B+ 订单;13,000 个品牌;52TB | 订单与配送数据 | 中高 |
| Riskified | 10,000+ 在线 TPS;99%+ 留存 | 反欺诈交易系统 | 中高 |
| Shipt | 1-2M 笔支付交易 / 天;四个区域 | 分布式支付系统 | 中 |
| Hard Rock Digital | 峰值约 100 个节点 x 32 vCPUs | 多区域体育投注平台 | 中 |
| Form3 | 700 TPS;严格 P99 延迟 SLA | 多云支付引擎 | 中 |
| Booking.com | Order Platform 约 20TB | 预订订单管理 | 中 |
规模信号由案例研究口径归一化,应视为指示性信息,而非审计数据。
[CU004, CU009, CU012, CU016, CU019, CU021]| 指标或信号 | 数值 / 状态 | 细分 | 置信度 | 尽调追问 |
|---|---|---|---|---|
| Riskified 留存 | 99%+ 客户留存 | 反欺诈 / 电商 | 中 | 确认该数字是客户数留存还是收入留存,以及时效性 |
| Route 支持 ROI | 付费支持带来正向定性 ROI | 电商平台 | 中 | 索取续约率与支持附加率 |
| 零停机迁移 | Riskified、DoorDash、SumUp 多次出现的定性信号 | 迁移任务重的企业买家 | 中 | 索取迁移成功率与价值实现时间数据 |
| 公开 NRR / GRR | 未披露 | 全组合 | 高 | 按细分索取 NRR、GRR、流失和续约数据 |
| 客户集中度 | 未披露 | 全组合 | 高 | 索取前 10 和前 20 大客户收入占比 |
| 独立满意度覆盖 | 生产账户样本薄 | 全组合 | 中 | 安排客户访谈,并索取第三方评价分布 |
留存证据大多是个案或单个客户层面,不是组合层面的证据。
[CU019, CU023, CU037, CU039, CU040, CU041]展示证据可得性如何从历史披露的客户基数,收窄到少数有深度量化的公开参考案例。
各阶段混合了不同证据口径,并不是真正的商业转化漏斗;图表旨在说明,公开证据会从宽泛客户说法收窄到少数描述丰富的参考账户。
[CU018, CU024, CU028, CU035, CU041]定性比较旗舰客户的参考质量、结果具体度、留存可见度和部署成熟度。
评分是基于公开来源深度和独立性的分析判断,不是供应商披露的评分体系。
[CU019, CU023, CU038, CU039, CU040, CU041]6.3 客户质量强,但数量精度和独立性较弱
从投资者视角看,客户故事强在质量,而不是精确度。Riskified 的 99+ percent 留存、Route 的支持 ROI,以及多次零停机迁移或多区域生存案例,都支持一个判断:CockroachDB 一旦嵌入,就能变得粘。这些都是扩张和长期收入的好信号。 公开证据没有给出一个干净的 2026 年答案:Cockroach Labs 现在到底有多少付费客户,净留存是什么水平,客户支出如何在自助和大型企业账户之间分布。官方客户数披露已经陈旧,第三方客户数据库有帮助,但方法论较弱。 因此,审慎结论是建设性但不天真。Cockroach Labs 似乎拥有高价值客户,切换成本有意义,售后支持参与也可见。不过,公开证据集仍严重依赖公司撰写的成功故事,所以客户深度比客户广度、留存质量或队列经济更容易验证。投资者应把客户章节视为困难账户适配度的强证据,而不是替代真实队列披露或集中度分析。缺失输入大多是商业性的,不是技术性的;这仍是一个有意义的尽调区分。[CU013, CU017, CU020, CU023, CU031, CU037]
| 客户 | 起点或约束 | Cockroach 胜出的原因 | 落地后的影响 |
|---|---|---|---|
| 平台标准化 | DoorDash 和 Netflix 把 CockroachDB 做成内部服务,放大了大客户内部扩张空间 | 一旦获批,可能撬动很大的钱包份额 | 也会依赖少数成熟旗舰账户 |
| 受监管地域扩张 | Form3 和 Hard Rock 在合规和数据地域性重要的场景扩张 | 支撑支付和博彩领域的高 ACV | 销售周期可能拉长,部署支持需求可能上升 |
| 迁移驱动的切入打法 | Riskified、Booking.com、DoorDash 和 SumUp 显示,替换项目可以作为切入点 | 可能拿下持久的核心记录系统位置 | 迁移可能需要大量支持和解决方案工程 |
| 托管服务采用 | 2021 年披露 50%+ 客户使用 Dedicated | 提升云端扩张潜力 | 当前云业务结构未披露 |
| 支持服务附加 | Route 和 Hard Rock 提到供应商支持有明显价值 | 可能提高扩张和续约韧性 | 如果支持强度高,可能推高服务成本 |
| 客户集中度 | 无公开集中度数据 | 大型标杆账户可能有重要经济意义 | 直接索取集中度和队列数据 |
扩张潜力看得见,但公开证据仍缺集中度和附加率精度。
[CU003, CU023, CU035, CU037, CU038, CU041]Cockroach Labs 未公开披露真实留存队列,因此这里只给示意性留存曲线。
Cockroach Labs 未披露队列留存数据。这些曲线只是说明框架,用来衡量价值有多大程度取决于真实续约和扩张行为,并非公司特定测量值。
[CU039, CU040, CU041, CU044]6.4 展示材料
07风险
7.1 最大近期风险是运营正确性和版本纪律
公开公告记录让一点无法忽视:CockroachDB 的核心挑战不是证明架构在原则上可行,而是在众多版本、工作负载和部署模式中,让一个非常雄心勃勃的分布式数据库保持安全。仅 2026 年公告就覆盖权限提升、部分索引损坏、实时数据删除,以及静默导入或备份失败场景。即便许多 bug 罕见且很快修复,它们也直接击中客户在记录源系统中最在意的信任假设。 运营模型会放大这种风险。Cockroach Labs 自己的生产和升级文档强调拓扑纪律、复制健康、备份验证、热点感知 schema 设计、负载均衡和发布定版意识。这是成熟文档,但也说明韧性取决于客户执行。数据库拿掉了一部分应用层负担,同时在运营工程中创造了另一种负担。 这让公司暴露在一个经典基础设施挑战下:如果正确性缺陷或升级错误击中旗舰账户,损害会很快从支持负载传导到续约风险和品牌声誉。因此,卓越运营对 Cockroach Labs 不是可有可无,而就是产品本身。主要风险不是某个灾难性未知,而是许多小运营故障在一个客户预期完全可信的数据库中不断叠加。[CR001, CR002, CR003, CR004, CR005, CR006]
| 失效模式 | 可能性 | 严重性 | 缓解成熟度 | 剩余暴露 | 未解决缺口 |
|---|---|---|---|---|---|
| 权限提升或认证控制缺陷 | 低至中 | 高 | 可修补,已有文档 | 中等 | 需要客户群补丁采用证据 |
| 数据损坏或在线数据删除缺陷 | 低 | 很高 | 可修补,但影响严重 | 重大 | 需要事件频率和客户影响历史 |
| 静默备份 / 导入失效模式 | 低至中 | 高 | 已有文档化缓解措施 | 重大 | 需要大型账户备份验证实践证据 |
| 热点与糟糕的工作负载设计 | 中 | 中高 | 文档充分,但依赖客户执行 | 中等 | 需要售前工作负载筛选纪律的证据 |
| 升级或支持政策漂移 | 中 | 中高 | 文档扎实 | 中等 | 需要存量客户版本分布可见性 |
| 云或控制平面事件处理 | 低至中 | 中高 | 公开状态页与 SLA 结构 | 中等 | 公开披露的事故历史细节仍偏薄 |
各行按剩余严重性排序,而不是只看原始发生概率。
[CR001, CR002, CR003, CR004, CR005, CR007]定性评估主要风险集群的发生可能性、严重性、缓解成熟度和剩余暴露。
[CR001, CR013, CR018, CR031, CR036, CR042]7.2 受监管客户提高了产品经济价值,也提高了失败严重性
Cockroach Labs 越来越多卖进支付、银行和博彩用例;数据库不只是存信息,还帮助满足法律、运营韧性和位置控制要求。DORA 提高了金融业 ICT 风险管理、事故报告、测试和第三方监督的门槛;sportsbook 和游戏部署则可能继承州内本地性和 Wire Act 逻辑,迫使客户做定制拓扑决策。 这在战略上有吸引力,因为这些是高价值工作负载,切换成本大。但也危险,因为失败会被用更高标准审视。受监管行业客户会期待发布纪律、事故透明度、数据处理控制和清晰合同边界的证明。公司自身法律材料也强化了这种共同责任模型:客户必须保护账户、遵循最佳实践,并接受围绕 beta 服务、服务连续性和补救的实质限制。 实际上,这意味着 Cockroach 上探高端市场,也是在进入后果更严重的场景。公司赢下越多关键任务受监管账户,就越没有空间容忍模糊宕机、薄弱文档或草率版本政策。这种不对称尤其重要:在最好客户细分中的成功,也会抬高每一次未来错误的成本。[CR016, CR017, CR018, CR019, CR020, CR021]
| 规则 / 义务 | 辖区 / 范围 | Cockroach 面临的风险 | 可能性 | 严重性 | 缓解措施 | 剩余暴露 | 尽调路径 |
|---|---|---|---|---|---|---|---|
| DORA ICT 韧性与第三方监督 | 欧盟金融实体及其供应商 | 金融科技 / 银行账户的尽调负担、事件响应预期和审计压力更高 | 中 | 高 | 维持强控制,并提供面向客户的合规证据 | 重大 | 索取受监管客户审计结果和控制项映射 |
| Wire Act 与州内博彩地域要求 | 美国体育投注 / 游戏部署 | 游戏客户需要定制拓扑,并承担法律合规失败风险 | 中 | 高 | 使用地域感知架构和部署指引 | 重大 | 索取博彩客户的法律 / 合规手册 |
| 云服务条款与网站条款 | 所有云客户和网站用户 | 服务暂停、修改或终止权可能带来采购摩擦 | 中 | 中 | 谈判企业条款,明确 SLA 补救措施 | 中等 | 审查企业 MSA 和头部客户合同例外 |
| 隐私与跨境传输义务 | 全球用户,尤其是 GDPR / 英国 / 加州场景 | 注重隐私的买家会加强数据处理审查,并增加合同负担 | 中 | 中 | 提供 DPA / 隐私控制和区域设计指引 | 中等 | 审查 DPA 采用情况和隐私事件历史 |
此登记表聚焦可能给客户或供应商带来重大义务的外部规则或合同。
[CR018, CR019, CR020, CR021, CR022, CR023]| 依赖项 | 交易对手 | 角色 | 集中度 / 锁定问题 | 失效情景 | 严重性 | 缓释措施 | 剩余敞口 |
|---|---|---|---|---|---|---|---|
| 公有云基础设施 | AWS / GCP / Azure | 托管客户集群或相邻服务 | 可迁移性无法消除对超大规模云厂商的依赖 | 区域性宕机、定价变化或服务降级影响客户 | 高 | 多区域设计和多云架构模式 | 实质性 |
| 对象存储和备份目标 | 云存储提供商 | 备份 / 导入 / 导出 | 存储层问题会外溢成数据保护风险 | 后端间歇性故障会导致备份或导入结果不完整 | 高 | 校验、恢复测试和补丁版本 | 实质性 |
| 托管 Kubernetes 或云控制平面 | 云厂商 | 部分部署的运营底座 | 会在 Cockroach 之外再叠一层控制 | 控制平面退化会拖慢客户恢复或扩容 | 中 | 架构审查和兜底预案 | 中等 |
| 客户账户与支持团队 | Cockroach Labs 内部组织 | 知识转移和运营指导 | 服务介入重的部署可能依赖隐性知识 | 支持人手不足或人员流失会推高客户流失风险 | 中 | 固化作战手册,并把更多运维自动化 | 中等 |
相比 API 包装层,公司对合作伙伴的依赖较低,但仍然明显暴露在基础设施和服务依赖上。
[CR016, CR017, CR023, CR033, CR040]即使产品承诺可移植性,外部依赖仍会塑造 Cockroach Labs 风险;本图映射这些依赖。
[CR018, CR020, CR023, CR033, CR042]7.3 商业风险来自选择性适配、支持强度和有限公开可见度
Cockroach Labs 面临的风险看起来不是简单的“没人需要这个”。更大的商业问题是,真正有吸引力的买方池有多宽。PostgreSQL 仍然免费且极其熟悉,Aurora 和 Spanner 对接受云锁定的买方来说,运营上可能更简单。这意味着 Cockroach 往往在工作负载特别苛刻时赢单——全球交易、多区域生存能力或合规敏感的可移植性——但在主流关系型机会中可能遇到更多摩擦。 客户故事显示,许多胜利由迁移驱动且高度依赖支持。这可以形成耐久的记录源部署和强 ACV,但也会拉长销售周期、抬高服务成本。围绕隐藏实施成本的独立批评,也指向同一方向。如果扩张经济性强,这笔交易仍然成立;公开来源只是没有披露足够信息来证明今天已经成立。 这种不透明很重要。没有新的留存、集中度、毛利率或现金跑道披露,投资者能看见风险地图的轮廓,却看不见各项权重。正确姿态是选择性乐观,同时硬尽调客户集中度、支持强度和发布相关事故历史。如果这些隐藏指标强,许多可见风险看起来可管理;如果它们弱,同样的可见风险会很快威胁投资逻辑。[CR028, CR029, CR030, CR031, CR032, CR034]
| 角色 / 职能 | 依赖或缺口 | 发生概率 | 严重性 | 缓释措施 | 尽调路径 |
|---|---|---|---|---|---|
| 支持与解决方案工程 | 复杂账户可能需要深入上手支持 | 中 | 高 | 投入自动化和作战手册 | 索取支持人员配比和升级处理指标 |
| 产品 / 发布工程 | 正确性缺陷可能造成过大的信任损伤 | 中 | 很高 | 严格补丁纪律和强 QA | 索取缺陷严重性和补丁延迟仪表盘 |
| GTM / 企业销售 | 工作负载适配面较窄会拉长销售周期 | 中 | 中高 | 聚焦 ICP,并借力合作伙伴 | 按细分市场索取销售周期和赢率数据 |
| 财务 / 领导层 | 烧钱速度、利润率和集中度不透明,会限制投资测算 | 中 | 中高 | 在尽调中提高披露 | 索取董事会材料指标和情景分析 |
| 客户成功 | 扩张取决于续约和采用管理 | 中 | 高 | 规范客户健康度评分 | 索取续约手册和流失复盘 |
执行风险集中在需要用人的专业能力弥补产品不够简单的环节。
[CR016, CR017, CR031, CR032, CR036, CR037]| 风险 | 可监控触发点 | 阈值 / 事件 | 行动含义 |
|---|---|---|---|
| 正确性 / 安全缺陷风险 | Severity-1 或大范围安全公告模式 | 多个受支持主版本缺陷对真实客户造成影响 | 升级技术尽调,或暂停投资 |
| 版本政策 / 支持漂移 | 大量存量安装暴露在不受支持版本或创新版本上 | 相当比例收入落在不受支持路径上 | 形成确信前,要求补救计划 |
| 支持强度风险 | 支持负担上升,但缺少利润率证明 | 企业客户支持成本增速快于 ARR | 下压毛利率假设 |
| 客户集中风险 | 尽调发现头部客户依赖 | 前 10 大客户收入占比过高 | 下调估值,并索取合同韧性证明 |
| 竞争对手简化风险 | 核心 ICP 中输给 Aurora / Postgres | 极端用例之外赢率走弱 | 重估 TAM 和销售效率假设 |
| 受监管账户事故风险 | 金融科技或博彩客户出现高关注宕机 | 受监管工作流出现客户可见的服务或正确性故障 | 视为威胁投资逻辑的事件 |
每行把一个可见风险与可操作的尽调或监控阈值配对。
[CR038, CR039, CR040, CR041, CR042]展示技术或合同问题如何传导到客户、支持成本、留存和估值。
[CR032, CR039, CR040, CR041, CR042]7.4 展示材料
08估值
8.1 当前价格信号指向溢价估值,但公开财务支撑有限
最具体的估值锚点很直接:Cockroach Labs 上一次一级融资是在 2021 年末,估值 $5 billion;2026 年中二级市场指标暗示更接近 $6.9 billion。这些价格并非小成就。它们表明,市场仍把 Cockroach 视为严肃的后期基础设施资产,而不是正在褪色的私募轮次。 问题在于,公开数字到底能支撑什么。最清楚的收入锚点仍来自第三方 2023 年约 $128.3 million 的估算。即便投资者慷慨测试 2025-2026 年 $170-200 million 的收入区间,隐含收入倍数仍然很高。这意味着当前价格已经假设收入规模、留存或毛利率质量存在有意义的隐藏进展。因此,公开证据支持方向,但精度不足,无法让价格显得保守。 换句话说,公开市场面对的问题不是 Cockroach 好不好,而是这家公司是否足够好、足够大、足够高效,能在披露不完整的情况下支撑溢价倍数。这是难得多的问题。因此,投资问题主要是估值纪律,而不是公司识别或简单认知。[CV001, CV003, CV004, CV005, CV006, CV007]
| 建议 | 信心 | 风险评级 | 估值立场 | 决策含义 |
|---|---|---|---|---|
| 继续尽调 / 不追高 | 中 | 高 | 偏满到苛刻 | 公司质量确实存在,但当前公开证据尚不能清楚支持激进入场价 |
| 看好公司质量 | 高 | 中高 | 原则上可支撑溢价 | 客户证明和产品深度支持继续跟进 |
| 不认为便宜 | 高 | 高 | 并非明显低估 | 需要更好的私有指标,才可把当前价格视为有吸引力 |
| 只有数据更强时才有条件看多 | 中 | 中 | 乐观情景需要未显露的上行空间 | 收入、留存和利润率证据会改变判断 |
该建议把对公司的热情与对当前隐含价格的热情分开。
[CV038, CV039, CV042]| 情景 | 假设 | 估值 / 回报逻辑 | 关键风险 | 概率信号 |
|---|---|---|---|---|
| 乐观 | 收入明显高于 $200M,NRR 强,毛利率好,集中支持负担可控 | 当前老股交易价格可被证明合理,或小幅跑赢 | 执行仍然关键,但质量得到验证 | 可能,但尚未证明 |
| 基准 | 收入已有增长,但不足以抹掉不透明折价;客户质量强 | 价值集中在数十亿美元中段,接近或略低于当前老股交易价格标记 | 当前价格已计入大部分利好 | 公开证据下最站得住脚 |
| 悲观 | 增长更慢、留存更弱,或支持强度高于预期 | 价值落向公开可比公司的低区间,当前老股交易价格过高 | 不透明掩盖脆弱经济性 | 不能排除 |
| 上行释放 | 新一轮一级融资或类公开市场披露验证指标 | 溢价倍数压缩幅度可能小于担忧 | 需要目前未公开的数据 | 取决于尽调是否跑通 |
情景逻辑围绕当前隐藏指标必须达到什么水平,而不是围绕任意叙事。
[CV033, CV034, CV035, CV036, CV037, CV039]展示假设收入上调时,隐含收入倍数会多快下降。
[CV008, CV009, CV010, CV011, CV012]8.2 公开可比公司支持质量溢价,但不支持无上限溢价
上市软件和数据库可比公司给出一个不舒服的现实校准。MongoDB 和 Confluent 的收入倍数大约在 10 倍,Snowflake 接近 20 倍,Elastic 明显更低。放在这个背景下,即使用乐观收入假设去测,Cockroach 隐含的老股交易倍数也偏贵。若当前价格要落入类似 Snowflake 的上市高溢价区间,Cockroach 的收入必须远高于最近一次公开锚点;若要落入类似 MongoDB 的区间,收入还得大幅更高。 这不说明估值不理性。Cockroach 有少见的强客户验证、扎实的技术深度,也有一个围绕韧性、可移植性和全球分布式事务工作负载展开的品类叙事。这些特质配得上高于平均基础设施软件的溢价,也大概率高于较弱数据库资产。优质资产和容易下手的价格是两回事;本章的结论是,公司更明显属于前者,而不是后者。 但质量再高,也不能取消估值纪律。公司还背负运营复杂度、支持强度和披露不透明,这些都不支持支付无限稀缺性溢价。合理结论是,Cockroach 值得一定溢价,但当前定价似乎已经吃掉其中很大一部分。投资者应假设,公司需要用更强的私有指标挣回估值,而不是仅靠产品质量让人赞叹。[CV013, CV014, CV015, CV016, CV017, CV018]
| 论点 | 什么会改变判断 |
|---|---|
| 高质量分布式数据库,客户证明强,长期相关性稳 | 若当前旗舰客户消费浅、合同短或扩张弱,该判断会走弱 |
| 品类质量稀缺,因此可支撑溢价估值 | 若当前收入远高于公开锚点,且 NRR 属于顶尖水平,该判断会加强 |
| 相对公开可比公司和可见数据,估值过满 | 若私有财务数据显示 Snowflake 式增长质量,或利润率结构异常强,该判断会走弱 |
| 阻碍形成确信的主要因素是不透明,而非产品可信度 | 若拿到当前 ARR、利润率、留存和集中度数据,障碍会缓解 |
反向逻辑主要围绕价格纪律和缺失数据,而不是不相信核心产品。
[CV021, CV023, CV029, CV038, CV042]| 可比对象 | 指标 | 倍数 / 估值 / 状态 | 相关性 | 局限 |
|---|---|---|---|---|
| MongoDB | 公开市值 / TTM 收入 | ~10.2x 收入 | 最接近的公开市场高溢价数据库可比公司 | 产品组合和上市公司成熟度不同 |
| Snowflake | 公开市值 / TTM 收入 | ~19.9x 收入 | 数据基础设施溢价的上限参考 | 规模大得多,分析定位也更宽 |
| Confluent | 公开市值 / TTM 收入 | ~9.6x 收入 | 带平台叙事的基础设施软件可比公司 | 不是数据库,毛利率画像也不同 |
| Elastic | 公开市值 / TTM 收入 | ~3.9x 收入 | 倍数压缩的下限参考 | 搜索 / 可观测性混合业务 |
| Cockroach Labs 2021 年一级融资 | 基于公开 2023 年锚点的 $5B | ~39x(基于 $128.3M);~29x(基于 $170M 测试) | 显示上一轮私有市场定价已经很高 | 使用已过时的公开收入锚点 |
| Cockroach Labs 2026 年老股交易 | 基于公开锚点 / 测试情形的 $6.9B | ~53.8x(基于 $128.3M);~40.6x(基于 $170M);~34.5x(基于 $200M) | 当前私有市场参考 | 老股交易不是广泛公开市场清算价 |
公开可比公司测算为近似值,基于 2026 年 7 月市值和 TTM 收入快照。
[CV008, CV009, CV010, CV011, CV012, CV013]展示公司质量、价格偏高和指标缺失如何共同导向谨慎但保持跟进的建议。
[CV021, CV023, CV030, CV038, CV042]8.3 正确姿态:认可公司质量,谨慎看待入场价
估值争论最终落在几个看不见的变量上。如果 Cockroach 收入已经远高于 $200 million,净留存强,支持成本可控,并且证明基础设施利润率可以长期守住,私有估值区间上沿就说得通。否则,当前老股价格已经走在公开证据能支撑的位置之前。 因此,建议应是有条件的,而不是二元判断。这个业务看起来具备后期基础设施公司应有的客户质量和技术相关性。但价格已经不再明显早期,也不宽容。当前价格预设了多个尽调问题都有好答案,而这些答案还没有公开披露。纪律严明的投资者仍然可以喜欢这个局面,但在把当前入场点称为有吸引力之前,应坚持看到更强证据。 因此,最佳投资姿态是保持接触,紧盯当前指标,不要把对产品的欣赏误当成当前估值有吸引力的证明。Cockroach Labs 看起来可投,但还不像便宜。这个区别正是纪律型投资者在现阶段做建议时应遵循的边界。[CV024, CV025, CV026, CV027, CV028, CV029]
| 触发点 | 阈值 | 对投资逻辑的传导 | 行动含义 |
|---|---|---|---|
| 收入远低于乐观私有市场预期 | 当前 ARR / 收入未能令人信服地高于 $200M | 溢价倍数理由大幅走弱 | 不要以当前老股交易价格投资测算 |
| 留存弱或集中度高 | NRR 低于同类水平,或头部客户依赖高 | 客户质量逻辑部分破裂 | 大幅打折或暂停 |
| 支持负担重的经济性 | 毛利率和支持成本差于预期 | 运营复杂性变成财务拖累 | 降低估值容忍度 |
| 正确性 / 信任事故 | 严重安全公告,或核心账户出现客户可见故障 | 质量溢价快速压缩 | 立即重估投资逻辑 |
| 私有市场热度消退 | 没有新基本面支撑,老股交易需求走弱 | 业务证明出现前,价格支撑先消失 | 避免因动量入场 |
触发点聚焦最可能快速改变估值的少数隐藏变量。
[CV039, CV040, CV042]| 主题 | 缺失证据 | 重要性 | 负责人或尽调路径 |
|---|---|---|---|
| 当前 ARR / 收入 | 2025-2026 年实际收入运行率 | 决定当前倍数只是偏高,还是大幅偏高 | 财务领导层 / 董事会材料 |
| 净留存和总留存 | 分细分市场的 NRR / GRR | 显示高质量客户证明能否转化为持续扩张 | FP&A / RevOps |
| 毛利率和支持成本 | 支持 / 服务强度和云毛利率 | 检验复杂性会增厚收益,还是稀释利润率 | 财务 + 支持领导层 |
| 客户集中度 | 前 10 / 前 20 大账户收入占比 | 高集中度会显著抬高下行风险 | 财务 + 销售运营 |
| 现金跑道和烧钱速度 | 当前现金余额和烧钱轨迹 | 老股交易价格不等于新增现金 | 财务 / CFO |
| 合同保护 | 头部客户协商后的 SLA / MSA 相比标准云条款的差异 | 可能降低部分网页条款风险假设 | 法务 / 销售领导层 |
这些问题按最可能改变估值立场的顺序排序,而不是只为填补好奇缺口。
[CV029, CV030, CV039, CV040, CV041, CV042]围绕当前私募市场参考价,勾勒基于公开证据可辩护的估值区间。
[CV033, CV034, CV035, CV036, CV037, CV042]IC 风格,从七个估值维度打分。
[CV021, CV024, CV029, CV030, CV038, CV042]8.4 图表
免责声明
这份报告元数据仅基于截至 2026-07-20 在 Cockroach Labs 章节 YAML 中审阅的公开信息,不构成投资建议。投资建议和估值判断仍高度依赖未披露的当前财务指标、客户集中度、支持成本强度以及谈判后的企业合同条款。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | Cockroach Labs was founded in 2015 and is based in New York, New York. | 高 | SO014, SO008 |
| CO002 | CockroachDB is a cloud-native distributed SQL database built for high availability, effortless scale, and control over data placement. | 高 | SO001, SO002, SO024 |
| CO003 | CockroachDB combines a PostgreSQL-compatible SQL interface with distributed architecture, ACID transactions, and migration tooling. | 高 | SO002, SO002, SO024 |
| CO004 | Cockroach Labs sells fully managed CockroachDB Cloud, self-hosted Enterprise, and support or migration services rather than a single deployment form factor. | 高 | SO003, SO004, SO005 |
| CO005 | CockroachDB Cloud Basic includes 50 million request units and 10 GiB of storage free each month, while Standard and Advanced tiers target progressively heavier production workloads. | 高 | SO005, SO003 |
| CO006 | Advanced CockroachDB Cloud clusters support AWS, GCP, and Azure multi-region deployments with an advertised availability target of up to 99.999 percent. | 高 | SO005, SO003 |
| CO007 | Cockroach Labs states that CockroachDB Cloud Advanced is HIPAA-ready and PCI DSS capable, and the trust materials also cite SOC 2 and ISO certifications. | 高 | SO007, SO006, SO002 |
| CO008 | Cockroach Labs raised $278 million in Series F financing in December 2021 at a $5 billion valuation. | 高 | SO008, SO009 |
| CO009 | The Series F round was led by Greenoaks with participation from Altimeter, BOND, Benchmark, Coatue, FirstMark, GV, Index Ventures, J.P. Morgan, Lone Pine, Redpoint, and Tiger Global. | 中 | SO008 |
| CO010 | Cockroach Labs said the Series F round brought lifetime funding to $633 million. | 高 | SO008, SO014 |
| CO011 | CB Insights labels Cockroach Labs as a Series F company that is still alive and privately held. | 中 | SO014, SO017 |
| CO012 | Nasdaq Private Market estimated Cockroach Labs shares at $7.79 per share on July 2, 2026, confirming active private-market price discovery. | 中 | SO017 |
| CO013 | Notice.co listed Cockroach Labs at roughly $8.18 per share in July 2026, broadly corroborating the high-single-digit secondary pricing range. | 中 | SO016, SO017 |
| CO014 | A PM Insights snapshot surfaced through web search described Cockroach Labs secondary pricing in June 2026 as roughly a 38 percent premium to the 2021 primary round, implying valuation around $6.9 billion. | 低 | SO015, SO017, SO016 |
| CO015 | GetLatka reports Cockroach Labs generated $128.3 million of revenue or ARR in 2023. | 中 | SO013 |
| CO016 | GetLatka estimates Cockroach Labs employed about 715 people by November 2025, up from 590 in 2023 and 483 in 2022. | 中 | SO013 |
| CO017 | At the time of the Series F announcement, Cockroach Labs said annual recurring revenue had tripled year over year and cloud revenue had risen 500 percent in the prior quarter. | 中 | SO008 |
| CO018 | The same 2021 announcement said Cockroach Labs had more than 200 customers, tens of thousands of deployed clusters, and more than 50 percent of customers already on CockroachDB Dedicated. | 高 | SO008, SO009 |
| CO019 | Cockroach Labs said the Serverless beta had already brought in more than 10,000 new users only weeks after launch. | 高 | SO008, SO009 |
| CO020 | The Series F blog says the company started in 2015 as a small team frustrated by the lack of sophisticated open-source database technology for a cloud-first world. | 中 | SO009 |
| CO021 | Spencer Kimball remains CEO and the most visible public executive voice for Cockroach Labs in both the 2021 funding materials and the 2026 momentum update. | 高 | SO008, SO010 |
| CO022 | The 2026 momentum release says Cockroach Labs strengthened its leadership team by appointing Sailesh Munagala as chief financial officer. | 中 | SO010 |
| CO023 | The 2026 momentum release positions CockroachDB around AI-scale resilience, claiming throughput improvements of up to 50 percent in release 25.2 and new C-SPANN vector indexing for PostgreSQL-compatible search. | 中 | SO010, SO010 |
| CO024 | The 2026 update also says Cockroach Labs expanded its partner ecosystem, including an OEM relationship with IBM. | 中 | SO010 |
| CO025 | Cockroach Labs cites Form3, Hard Rock Digital, and Shipt among customers in the 2026 momentum release, signaling continued concentration in regulated and always-on workloads. | 中 | SO010 |
| CO026 | Booking.com uses CockroachDB Cloud for an order platform of roughly 20 TB spanning Frankfurt, London, and Ireland, showing relevance for global travel transaction systems. | 中 | SO020 |
| CO027 | SumUp says CockroachDB underpins a global payments platform serving more than 4 million merchants across 36 markets and processed over 1 billion transactions in 2024. | 中 | SO021 |
| CO028 | Form3 describes CockroachDB as the backbone of a multi-cloud payments engine spanning AWS, GCP, and Azure with 100 to 200 millisecond p99 service-level targets and up to 700 TPS. | 中 | SO022 |
| CO029 | Netflix says it now operates more than 380 CockroachDB clusters, including over 160 production clusters and more than 60 multi-region clusters. | 中 | SO023 |
| CO030 | The GitHub repository describes CockroachDB as open source distributed SQL designed for high availability and effortless scale, reinforcing the company’s open-core developer motion. | 中 | SO024 |
| CO031 | The 2021 funding materials cited more than 22,000 GitHub stars and more than 500 open-source contributors, indicating substantial early developer adoption. | 中 | SO008, SO009 |
| CO032 | Stack Overflow’s 2024 developer survey reported PostgreSQL as the most-used database, strengthening Cockroach Labs’ strategic choice to emphasize PostgreSQL compatibility rather than a proprietary query model. | 中 | SO025, SO002 |
| CO033 | Cockroach Labs discloses technical advisories publicly, and the 2026 advisory list includes privilege escalation, partial-index corruption, live-data deletion, and backup-integrity bugs. | 中 | SO011 |
| CO034 | CockroachDB Cloud also maintains a public incident-history page, showing the company chooses transparency on operational events rather than keeping status private. | 中 | SO012 |
| CO035 | Accessible public materials do not clearly disclose the current board roster or full cap-table governance terms, leaving investors without direct evidence on board composition, liquidation preferences, or information rights. | 中 | SO014, SO008, SO015 |
| CM001 | Cockroach Labs should be analyzed inside the narrower distributed-SQL category nested within broader cloud-database and distributed-database markets, because its product is explicitly a distributed SQL database rather than a generic NoSQL or analytics system. | 高 | SM002, SM018, SM011 |
| CM002 | The company's practical capture zone includes globally distributed OLTP databases, resilience-driven application modernization, and managed transactional database services, but excludes general-purpose analytics warehouses and document stores as primary market definitions. | 中 | SM002, SM003, SM017 |
| CM003 | Status-quo substitutes remain self-managed PostgreSQL, Amazon Aurora PostgreSQL, Google Spanner for hyperscaler buyers, and application-level sharding on incumbent relational databases. | 高 | SM018, SM014, SM015, SM030 |
| CM004 | DataIntelo publishes a 2025 distributed SQL database market estimate of about $7.2B growing toward roughly $24.8B by 2034 at about 14.7% CAGR, giving the cleanest narrow-category external sizing anchor. | 中 | SM011 |
| CM005 | Coherent Market Insights places the broader cloud database market around $26.0B in 2026 and about $73.0B by 2033 at roughly 15.9% CAGR, which is directionally useful but much broader than Cockroach Labs' product category. | 中 | SM012 |
| CM006 | Business Research Insights estimates the distributed database market at about $4.48B in 2026 growing to about $9.96B by 2035, a lens narrower than cloud databases but still broader than distributed SQL. | 中 | SM013 |
| CM007 | The spread between roughly $4.5B, $7.2B, and $26.0B market lenses shows why a single generic TAM number would overstate precision for Cockroach Labs. | 中 | SM011, SM012, SM013 |
| CM008 | DataIntelo attributes nearly 59% of distributed-SQL demand to cloud-based deployment models, reinforcing why Cockroach Labs' managed cloud products sit in the growth center of the category. | 高 | SM011, SM003 |
| CM009 | Coherent Market Insights likewise describes public-cloud deployment as the leading mode in the broader cloud-database market, corroborating the directional cloud bias even if the category definition is broader. | 高 | SM012, SM003 |
| CM010 | DataIntelo identifies BFSI as the leading distributed-SQL vertical at roughly a quarter of the market, aligning with Cockroach Labs' heavy marketing into financial-services resilience and compliance use cases. | 高 | SM011, SM005, SM007 |
| CM011 | Official customer evidence also shows travel, commerce, fraud prevention, and media/streaming workloads as credible adjacent verticals through Booking.com, Riskified, and Netflix references. | 高 | SM006, SM009, SM008 |
| CM012 | Cockroach Labs' 2026 momentum release adds AI to the target-vertical mix, implying category expansion into AI infrastructure workloads rather than only classic payment and booking systems. | 高 | SM004, SM025 |
| CM013 | The daily user is typically a platform, database, or application-infrastructure engineer, while the economic buyer is an engineering VP, CIO, CTO, or transformation owner sponsoring migration and resilience budgets. | 中 | SM002, SM024, SM006, SM007 |
| CM014 | Adoption usually starts when teams need multi-region writes, survive zonal outages, or remove operational fragility from manual sharding and failover designs. | 高 | SM002, SM034, SM006, SM007 |
| CM015 | PostgreSQL compatibility matters because PostgreSQL remains the most-used database in Stack Overflow's 2024 survey, and CockroachDB repeatedly positions compatibility as a migration and developer-adoption lever. | 高 | SM010, SM002, SM024 |
| CM016 | Compliance-sensitive buyers gain another adoption trigger when regulated workloads require data controls, resilience evidence, and auditable security posture. | 高 | SM005, SM035, SM007 |
| CM017 | AI/vector-search positioning is real but still an extension rather than the company's core historical market wedge, because the stronger public proof remains resilience-heavy transactional workloads. | 中 | SM025, SM036, SM006, SM007 |
| CM018 | Migration tooling is central to category adoption because the practical alternative is not “no database” but staying on PostgreSQL, Aurora, Oracle, or a legacy sharded estate. | 高 | SM024, SM037, SM038, SM014 |
| CM019 | Pricing pressure comes from hyperscaler databases and newer developer-native serverless products that make entry points cheap even when capabilities are narrower. | 高 | SM014, SM015, SM021, SM022, SM019, SM020 |
| CM020 | Aurora and PostgreSQL often win when buyers prioritize familiarity and single-cloud optimization over cross-region consistency and cloud portability. | 中 | SM014, SM018, SM031 |
| CM021 | Spanner is a stronger architectural alternative for large Google Cloud-centric deployments but implies hyperscaler dependence that some buyers explicitly want to avoid. | 中 | SM015, SM032 |
| CM022 | Newer serverless SQL products such as Neon and PlanetScale compress evaluation cycles by lowering experimentation cost, even if they do not fully match CockroachDB on resilience and consistency positioning. | 中 | SM022, SM021, SM003 |
| CM023 | TiDB and YugabyteDB matter most where teams want open-source-flavored distributed SQL alternatives rather than hyperscaler-managed databases. | 中 | SM019, SM020, SM033 |
| CM024 | Migration complexity remains a first-order adoption barrier because distributed SQL asks buyers to change database operations, performance expectations, and failure testing discipline rather than only switch vendors. | 高 | SM013, SM024, SM023 |
| CM025 | The presence of official advisories and public resilience messaging indicates that trust is both a driver and a gating constraint: buyers want resilience, but they scrutinize correctness failures intensely. | 高 | SM023, SM026, SM005 |
| CM026 | Market reports themselves are a diligence risk because they often bundle incompatible categories, geographies, and methodologies under one headline number. | 中 | SM011, SM012, SM013 |
| CM027 | North America appears to be the leading geography in the available market studies, which is directionally consistent with Cockroach Labs' U.S. headquarters and enterprise go-to-market footprint. | 中 | SM011, SM012, SM001 |
| CM028 | Large-enterprise buyers dominate the economic value pool because multi-region resilience projects usually require cross-team migration effort and executive sponsorship. | 中 | SM012, SM006, SM007 |
| CM029 | SMB adoption is more plausible through serverless and self-serve cloud entry points than through classic field-sold dedicated deployments. | 中 | SM003, SM022, SM021 |
| CM030 | Cockroach Labs' market opportunity is more adoption-timing sensitive than raw-TAM sensitive, because the biggest gating variables are migration urgency, outage pain, and willingness to pay for resilience. | 高 | SM034, SM026, SM024 |
| CM031 | Public evidence is insufficient to isolate a clean company-specific SAM without making aggressive assumptions about cloud preference, geography, and workload fit. | 中 | SM011, SM012, SM013 |
| CM032 | Public evidence is even less sufficient for SOM because no disclosed conversion funnel, win rate, or live paid-customer count lets an outsider ground market share responsibly. | 中 | SM028, SM029, SM003 |
| CM033 | The most credible market narrative is therefore not “massive generic database TAM,” but “high-value slice of mission-critical cloud-native transactional workloads with expensive failure modes.” | 高 | SM002, SM005, SM006, SM007 |
| CM034 | Customer examples show that the adoption path often begins in one latency- or resilience-sensitive workflow before expanding into broader platform standardization. | 中 | SM006, SM009, SM027 |
| CM035 | Later valuation work should treat the narrow distributed-SQL estimate as the base lens, the distributed-database estimate as a sanity check, and the cloud-database estimate as an outer bound rather than a core TAM. | 中 | SM011, SM013, SM012 |
| CP001 | Cockroach Labs competes across five distinct buyer choices: staying on PostgreSQL, moving to hyperscaler-managed PostgreSQL, adopting global-consistency distributed SQL, using sharded MySQL-style systems, or choosing adjacent modern data platforms. | 中 | SP008, SP010, SP012, SP017, SP028 |
| CP002 | PostgreSQL is open source, self-hostable anywhere, and backed by a large ecosystem, making it the default status-quo alternative rather than a niche competitor. | 高 | SP008, SP009 |
| CP003 | The 2024 Stack Overflow survey still lists PostgreSQL among the most popular database technologies, reinforcing that many teams can postpone migration by staying on familiar tooling. | 中 | SP009 |
| CP004 | Cockroach’s own comparison material argues that standard PostgreSQL remains primarily a single-primary architecture with HA, sharding, and failover assembled through extensions or third-party tooling rather than native distributed writes. | 中 | SP004 |
| CP005 | Amazon Aurora is positioned as a managed relational database on AWS with instance, storage, and optional I/O-based billing rather than a cloud-agnostic database layer. | 高 | SP010, SP011 |
| CP006 | Aurora Serverless prices capacity in ACUs and can scale database capacity automatically, lowering pilot friction for teams that want PostgreSQL compatibility without self-management. | 中 | SP010 |
| CP007 | Aurora Global Database adds cross-region replicated write I/O charges plus region-specific infrastructure, so global resilience on AWS still comes with AWS-native complexity and cost layers. | 高 | SP011, SP010 |
| CP008 | Google Spanner pricing is multi-component, charging for compute capacity, database storage, backups, cross-region replication, and some network bandwidth. | 中 | SP012 |
| CP009 | Spanner’s multi-region and replica-based pricing model is well suited to large committed GCP workloads but pushes buyers toward Google procurement, topology design, and cross-region cost planning. | 中 | SP012, SP006 |
| CP010 | AlloyDB markets itself as a 100 percent PostgreSQL-compatible managed database with AI features, up to 20 read replicas in a read pool, and a 99.99 percent uptime SLA. | 中 | SP013 |
| CP011 | AlloyDB pricing is built from vCPU, memory, storage, backup-storage, and networking charges, and a high-availability primary instance uses two nodes in-region. | 中 | SP014 |
| CP012 | PlanetScale now spans both Vitess and Postgres motions, with a base plan that starts at 5 dollars per month for single-node Postgres and 15 dollars for a three-node high-availability cluster. | 中 | SP015 |
| CP013 | PlanetScale’s enterprise offering includes single-tenant and bring-your-own-cloud deployment, migration assistance, and upgraded support, which makes it more credible with larger regulated customers than a pure hobby tool would be. | 中 | SP015 |
| CP014 | Vitess is a CNCF-graduated, MySQL-compatible sharding and failover layer that emphasizes transparent sharding, query rewriting, and near-zero-downtime resharding instead of PostgreSQL semantics. | 高 | SP017, SP016 |
| CP015 | Neon positions itself as a Postgres backend for apps and agents and documents a serverless architecture that separates compute from durable storage via streamed WAL. | 高 | SP018, SP020 |
| CP016 | Neon’s pricing is usage-based: paid plans bill CU-hours and storage, idle compute can scale to zero, and branching plus history retention change storage economics relative to a conventional managed cluster. | 高 | SP019, SP020 |
| CP017 | Yugabyte markets itself as an AI-ready distributed PostgreSQL database that is 100 percent open source, multi-master, and built for resilience across zones, regions, and clouds. | 中 | SP021 |
| CP018 | Yugabyte’s commercial motion pairs open-source software with managed or commercial support options, which lowers perceived lock-in relative to hyperscaler-only services. | 中 | SP021, SP022 |
| CP019 | A customer quote on Yugabyte’s homepage says one buyer consolidated Cassandra, Neo4j, Microsoft SQL Server, and CockroachDB systems into one YugabyteDB cluster, which is direct but vendor-authored evidence that Cockroach workloads can be displaced. | 中 | SP021 |
| CP020 | TiDB describes itself as a MySQL-compatible distributed SQL platform that unifies high-volume transactions and real-time analytics in one system with decoupled compute and storage. | 中 | SP023 |
| CP021 | TiDB publicly emphasizes automatic sharding, 99.99 percent availability, strong ACID consistency, and multi-cloud flexibility, which makes it a credible alternative for buyers who prize operational plus analytical consolidation more than PostgreSQL fidelity. | 中 | SP023 |
| CP022 | PingCAP’s cloud packaging spans starter or serverless entry points plus managed and self-managed deployment options, reinforcing a broad funnel from experimentation to enterprise. | 中 | SP024, SP025, SP023 |
| CP023 | SingleStore Helios is a cloud database service with separate compute and storage, vector search, MySQL and MongoDB wire-protocol compatibility, and multi-AZ high-availability positioning. | 中 | SP026 |
| CP024 | SingleStore pricing is explicitly usage-based and can add separate Flow CDC charges on top of compute and storage, which means financial predictability depends heavily on workload shape and ingestion patterns. | 中 | SP027 |
| CP025 | MongoDB Atlas is an adjacent substitute for modern app and AI workloads, but its public messaging centers on document-model agility and modern app builders rather than PostgreSQL-compatible distributed SQL. | 中 | SP028, SP029 |
| CP026 | Review and market-summary surfaces show that buyers evaluating CockroachDB are exposed to a crowded alternative set rather than to one single benchmark rival. | 低 | SP030, SP031 |
| CP027 | Cockroach’s strongest direct peer overlap remains with Yugabyte, Aurora, and Spanner because those options all compete on mission-critical transactional workloads rather than only on developer convenience or analytics adjacency. | 中 | SP007, SP005, SP006, SP002 |
| CP028 | Cockroach’s differentiator is not generic “cloud database” branding but the combination of PostgreSQL compatibility, distributed ACID, active-active multi-region design, and cloud-agnostic deployment. | 高 | SP002, SP004, SP005, SP006 |
| CP029 | That differentiation is narrower than it used to be because AlloyDB, Yugabyte, TiDB, SingleStore, and Neon all now market some mix of performance, AI, vector, or scale features that blunt a simple “modern database” pitch. | 中 | SP013, SP021, SP023, SP026, SP018 |
| CP030 | Hyperscaler products enjoy embedded procurement, support, and adjacent-service bundling, so Aurora, Spanner, and AlloyDB can win on organizational convenience even when Cockroach’s architecture is more portable. | 中 | SP011, SP012, SP013 |
| CP031 | Open-source or source-available ecosystems reduce lock-in anxiety for PostgreSQL, Vitess, Yugabyte, and TiDB relative to AWS-only Aurora or Google-centric Spanner and AlloyDB. | 中 | SP008, SP017, SP021, SP023, SP011, SP012, SP013 |
| CP032 | Switching-cost pressure cuts both ways: PostgreSQL familiarity lowers migration friction into CockroachDB, but the same familiarity also makes “stay on Postgres” or “move to managed Postgres” the easiest no-change answer. | 中 | SP008, SP002, SP010, SP013 |
| CP033 | Low-friction pricing from Neon and PlanetScale means developer-led experiments can start with little commitment, which pressures Cockroach Labs to justify why its heavier distributed model should be adopted before scale pain becomes acute. | 中 | SP019, SP015 |
| CP034 | Enterprise-style node or replica pricing from Spanner, AlloyDB, and Aurora suits steadier production workloads but can look expensive or complex next to simpler free-tier or usage-based serverless offers. | 中 | SP012, SP014, SP010, SP019, SP015 |
| CP035 | The market’s AI narrative is now table stakes: AlloyDB advertises AlloyDB AI, Yugabyte highlights vector indexing for RAG, TiDB frames mixed OLTP and AI workloads, and SingleStore markets vector and real-time AI use cases. | 高 | SP013, SP021, SP023, SP026 |
| CP036 | Public evidence does not reveal clean win-rate data for Cockroach Labs versus Aurora, Spanner, or Yugabyte, so competitive confidence must rely on product positioning and customer proof rather than on transparent market-share disclosures. | 低 | SP005, SP006, SP007, SP003 |
| CP037 | Public evidence also does not cleanly disclose funding or revenue scale for every private competitor in one comparable format, which limits how precise an outsider can make the private-competitive ranking. | 低 | SP021, SP018, SP015, SP023 |
| CI001 | Cockroach Labs monetizes multiple layers around the same core database: Basic cloud usage, Standard provisioned clusters, Advanced dedicated clusters, self-hosted enterprise deployment, support subscriptions, and professional or migration services. | 高 | SI004, SI002, SI003, SI006, SI011 |
| CI002 | CockroachDB Basic is billed purely on usage through request units, and the first monthly spending band equivalent to 50 million request units and 10 GiB of storage is credited back for monthly customers. | 高 | SI005, SI004 |
| CI003 | On the Basic plan, backups and data transfer are included in request-unit pricing rather than charged as separate line items. | 中 | SI005 |
| CI004 | CockroachDB Standard monetizes production workloads through provisioned vCPU-hours plus usage-based storage, data transfer, backups, and change data capture charges. | 高 | SI005, SI004 |
| CI005 | For multi-region Standard clusters, the price of the most expensive region is applied to the cluster’s provisioned compute capacity. | 高 | SI005, SI012 |
| CI006 | Cockroach Labs recommends a 40 percent capacity buffer when planning Standard clusters, which means customers are encouraged to reserve more compute than current observed demand alone would imply. | 中 | SI012 |
| CI007 | CockroachDB Standard is publicly described as well suited for workloads requiring 12 or fewer vCPUs, while Advanced has a minimum production configuration of 3 nodes times 4 vCPUs or 12 total vCPUs. | 中 | SI012 |
| CI008 | CockroachDB Standard is not recommended for analytical or hybrid OLTP/OLAP workloads, which suggests the company protects gross margin and customer fit by steering expensive mixed workloads toward other plans or architectures. | 中 | SI012 |
| CI009 | CockroachDB Advanced monetizes compute and storage on a per-node basis, and on AWS each node can also incur provisioned IOPS charges. | 高 | SI005, SI004 |
| CI010 | Advanced pricing also varies by region, cloud provider, and whether the Advanced security add-on is enabled, making the enterprise plan materially more configurable and more bespoke than Basic or Standard. | 高 | SI005, SI004 |
| CI011 | CockroachDB Cloud pricing changed for most customers after contract renewals that began after December 1, 2024, indicating the company has actively revised its monetization architecture rather than leaving older contracts untouched forever. | 中 | SI005 |
| CI012 | Public support pages show Cockroach Labs sells tiered subscriptions, with Enterprise support offering faster response times, a named customer success manager, dedicated Slack, technical advisory services, and root-cause analysis. | 中 | SI006 |
| CI013 | For CockroachDB Cloud customers with Enterprise Subscription support, Cockroach Labs states a 20 percent monthly fee applies based on actual Cloud Credit consumption. | 中 | SI006 |
| CI014 | Cockroach Labs explicitly markets migration assistance and performance tuning through support and professional services, implying a meaningful high-touch component in customer acquisition and expansion. | 高 | SI006, SI011, SI010 |
| CI015 | The MOLT toolkit covers schema conversion, initial data load, continuous replication, verification, and optional failback across PostgreSQL, MySQL, Oracle, and SQL Server migrations. | 高 | SI011, SI010 |
| CI016 | Because MOLT is designed for resilient, restartable, and minimal-downtime migrations, Cockroach Labs is optimized for consultative replacement projects rather than for purely impulsive self-serve adoption. | 中 | SI011, SI006, SI008 |
| CI017 | The public deployment-option documentation shows Cockroach Labs supports self-hosted, cloud, on-premises, and enterprise deployment patterns, which broadens its TAM but also broadens delivery complexity. | 高 | SI008, SI009, SI003 |
| CI018 | The 2021 Series F announcement said Cockroach Labs had tripled annual recurring revenue year over year and grown cloud revenue 500 percent in the previous quarter. | 高 | SI013, SI014 |
| CI019 | The same financing disclosure said Cockroach Labs had more than 200 customers and that more than 50 percent of customers were already on CockroachDB Dedicated. | 高 | SI013, SI014 |
| CI020 | GetLatka reports Cockroach Labs generated 128.3 million dollars of ARR or revenue in 2023, which remains the clearest public financial anchor even though it is not a fresh 2026 company disclosure. | 中 | SI016 |
| CI021 | GetLatka also estimates Cockroach Labs employed roughly 715 people by late 2025, giving a rough proxy for the company’s cost base even though the figure is third-party and unaudited. | 中 | SI016 |
| CI022 | The 2026 momentum release emphasizes performance, vector indexing, OEM and partner expansion, and leadership additions, but it does not publish current ARR, burn, or cash-flow metrics. | 中 | SI015 |
| CI023 | Cockroach Labs officially raised 278 million dollars in its Series F and said lifetime funding reached 633 million dollars at that point. | 高 | SI013, SI014 |
| CI024 | Secondary-market indicators in mid-2026 imply valuation around 6.9 billion dollars, but those signals do not add operating cash to the balance sheet the way a new primary round would. | 中 | SI018, SI020, SI019 |
| CI025 | Nasdaq Private Market and Notice together show that private-market liquidity exists for Cockroach Labs shares, which reduces distress concerns but does not answer runway, burn, or dilution questions. | 中 | SI020, SI019, SI021 |
| CI026 | Cloud bills can expand after adoption through storage, backup, changefeed, and cross-region data transfer charges, so the economic model is not only about provisioned compute. | 高 | SI005, SI012 |
| CI027 | On Standard, billing is based on capacity reserved rather than actual compute consumed, which can improve performance predictability but also creates slack cost for bursty customers. | 中 | SI005, SI012 |
| CI028 | Basic’s free and usage-based structure is a bottoms-up funnel, while Standard and Advanced are designed to monetize production reliability, multi-region scale, and compliance-sensitive workloads. | 中 | SI004, SI005, SI002, SI003 |
| CI029 | Enterprise support and migration assistance increase the lifetime value of larger customers, but they also imply service-delivery cost and some dependence on skilled solution engineering. | 中 | SI006, SI011, SI003 |
| CI030 | Compared with Aurora and Spanner, Cockroach monetizes a mix of provisioned infrastructure and usage-based line items that looks enterprise-oriented rather than radically cheaper on raw infrastructure alone. | 中 | SI005, SI022, SI023 |
| CI031 | Compared with Neon and PlanetScale, Cockroach’s production plans ask customers to reserve more explicit capacity up front, which likely raises ACV but also increases pilot friction. | 中 | SI004, SI025, SI024 |
| CI032 | The pricing documentation suggests Cockroach Labs can monetize customer growth not just through compute but through backup retention, CDC watched data, and cross-region traffic once workloads become mission critical. | 中 | SI005 |
| CI033 | Clusters running unsupported versions are not eligible for Cockroach Labs’ availability SLA, which ties enterprise supportability to customers staying on recent releases and therefore to an ongoing upgrade program. | 中 | SI007 |
| CI034 | Basic and Standard are automatically upgraded on recent regular releases, while Advanced customers can choose innovation releases and manually drive major-version upgrades, reinforcing that Advanced is built for more controlled enterprise operations. | 中 | SI007 |
| CI035 | Public evidence does not support clean estimates for current gross margin, net retention, CAC, payback, cash balance, or runway, which leaves the true efficiency profile underdetermined even though public database-company filings show the level of disclosure a mature comp can provide. | 低 | SI015, SI016, SI017, SI027 |
| CI036 | That absence of burn and cash-balance disclosure means public evidence cannot prove a precise path to profitability, even though mission-critical infrastructure usually has attractive recurring-revenue characteristics. | 低 | SI016, SI006, SI029 |
| CI037 | The financial quality that is visible is encouraging: workloads are sticky, support is monetizable, and the company has already shown cloud-mix improvement and large production customers. | 中 | SI013, SI006, SI029 |
| CI038 | The financial verdict is therefore positive on revenue quality but cautious on underwriting: Cockroach Labs looks like a real late-stage infrastructure business with multiple monetization levers, yet public evidence is still too thin to underwrite margin durability or capital adequacy with conviction. | 中 | SI016, SI013, SI005, SI015, SI018 |
| CI039 | A third-party 2026 pricing review argues that migration work, professional services, and cross-region data transfer can push real-world Cockroach spend above the headline entry price, which is directionally consistent with Cockroach’s own multi-line-item billing documentation. | 低 | SI028, SI005, SI006 |
| CE001 | CockroachDB is positioned as a PostgreSQL-compatible distributed SQL database for always-on customer experiences rather than as a generic analytics or document platform. | 高 | SE001, SE002, SE029 |
| CE002 | Cockroach Labs publicly sells multiple deployment forms around that core engine, including managed cloud, self-hosted enterprise, migration tooling, and support. | 高 | SE003, SE004, SE012 |
| CE003 | The serverless or Basic motion is designed for fast starts and low-friction experimentation, while Standard and Advanced target progressively heavier production, security, and multi-region needs. | 中 | SE005, SE003, SE004, SE014 |
| CE004 | CockroachDB’s architecture documentation says clients can send SQL to any node, which is then translated into key-value operations over distributed ranges. | 高 | SE016, SE010 |
| CE005 | CockroachDB stores data in contiguous ranges of key-value pairs, automatically splits those ranges as they grow, and replicates them to at least three nodes by default. | 中 | SE016 |
| CE006 | Writes require quorum agreement through the Raft protocol, and the leaseholder or Raft leader coordinates consistent reads and writes for a range. | 中 | SE016 |
| CE007 | CockroachDB was designed to accept reads and writes on all nodes while remaining highly automated and deployable in any environment without platform lock-in. | 高 | SE016, SE001 |
| CE008 | Multi-region capabilities include primary regions, table localities such as global, regional by table, and regional by row, plus configurable survival goals. | 中 | SE015 |
| CE009 | Super regions and data-domiciling controls are explicitly documented as a way to keep replicas within selected regions for compliance-sensitive deployments. | 高 | SE015, SE018 |
| CE010 | Secondary regions can be configured to improve failover behavior by pre-positioning leaseholder candidates when a primary region fails. | 中 | SE015 |
| CE011 | Cluster virtualization separates a cluster’s control plane from its data plane and introduces a system virtual cluster plus user virtual clusters with separate administrative boundaries. | 中 | SE017 |
| CE012 | Cockroach Labs documents cluster virtualization as necessary only for physical cluster replication and currently limits a physical cluster to one system virtual cluster and one virtual cluster. | 中 | SE017 |
| CE013 | CockroachDB v25.2 claims roughly 50 percent higher throughput than 24.3 on average across nine workloads. | 中 | SE019 |
| CE014 | The same release describes preview buffered writes that improved performance by roughly 15 to 40 percent in cited tests by reducing round trips and redundant writes. | 中 | SE019 |
| CE015 | CockroachDB added preview vector indexing in 25.2 using Cockroach-SPANN, positioning the database for large-scale semantic search without moving data into a separate specialist store. | 高 | SE019, SE020 |
| CE016 | The C-SPANN architecture stores vector partitions as self-contained units in CockroachDB ranges so index data can split, merge, and rebalance like ordinary table data. | 高 | SE020, SE016 |
| CE017 | C-SPANN uses quantization based on RaBitQ and claims roughly a 94 percent reduction in vector size in common cases. | 中 | SE020 |
| CE018 | CockroachDB vector indexes support prefix columns so searches can be partitioned by user or region, aligning vector search with tenancy and data-locality controls. | 高 | SE020, SE015 |
| CE019 | CockroachDB v26.1 adds expanded HIPAA and PCI participation for Cloud Advanced on Azure, JWT and OpenID Connect role synchronization, automatic user provisioning, unified CMEK management, and native FIPS 140-3 support. | 高 | SE018, SE023 |
| CE020 | The AI-agent security post frames row-level security, strict authorization, and jurisdiction-based data placement as product controls for zero-trust AI access. | 高 | SE021, SE018, SE023 |
| CE021 | Trust and security materials also emphasize SOC 2, ISO, HIPAA readiness, PCI capability, and encryption controls as core elements of the product package. | 高 | SE007, SE006, SE018 |
| CE022 | Cockroach Labs publishes a responsible disclosure policy, technical advisories, and a public cloud incident history, which shows mature quality-control and incident-transparency processes. | 高 | SE022, SE008, SE009 |
| CE023 | Those same advisories show that the product surface is broad and correctness-sensitive enough to generate real security and data-integrity issues, which is the ordinary downside of a complex distributed database. | 中 | SE008, SE009 |
| CE024 | Migration tooling includes MOLT plus change-data-capture and replication workflows, reinforcing that CockroachDB is built to replace existing transactional systems rather than only to power greenfield applications. | 高 | SE013, SE012, SE002 |
| CE025 | CockroachDB’s roadmap velocity is visible in the short span between 25.2 performance and vector-indexing updates and 26.1 security and compliance updates. | 中 | SE019, SE018, SE018 |
| CE026 | The product is explicitly cloud-aware across AWS, GCP, and Azure while also supporting self-hosted and hybrid patterns, which is a meaningful differentiator versus one-cloud relational services. | 高 | SE001, SE003, SE004, SE018 |
| CE027 | PostgreSQL compatibility remains central to the product design, allowing use of familiar SQL tools while avoiding the organizational friction of a wholly new query model. | 高 | SE002, SE016, SE029 |
| CE028 | Compared with PostgreSQL alone, CockroachDB’s native range replication, multi-active topology, and locality controls are the hardest parts for customers to reproduce themselves. | 中 | SE016, SE015, SE002 |
| CE029 | Compared with Aurora and Spanner, CockroachDB’s main product-tech edge is portability across clouds and self-hosted environments rather than merely offering another managed relational endpoint. | 中 | SE027, SE028, SE001, SE003 |
| CE030 | Compared with Yugabyte and TiDB, CockroachDB leans harder into PostgreSQL familiarity, locality-aware SQL abstractions, and integrated vector or security messaging rather than broader compatibility with MySQL or multi-model workloads. | 中 | SE031, SE032, SE002, SE019, SE018 |
| CE031 | The product breadth is now wide enough to encompass core SQL, multi-region topology, migrations, CDC, compliance, vector indexing, and AI-agent governance in one platform story. | 中 | SE002, SE015, SE013, SE019, SE018, SE021 |
| CE032 | That breadth is a strength for enterprise buyers but also a complexity risk because upgrades, security settings, locality choices, and index behavior all need disciplined operation. | 中 | SE017, SE015, SE008, SE018 |
| CE033 | Developer-signal evidence from the GitHub repository and the PostgreSQL ecosystem helps explain why CockroachDB emphasizes compatibility and operational automation rather than asking users to abandon the relational toolchain. | 中 | SE010, SE029, SE030 |
| CE034 | The public product story increasingly targets regulated and global workloads where row-level security, CMEK, data sovereignty, and multi-region failover are buying criteria rather than nice-to-have features. | 中 | SE018, SE015, SE007 |
| CE035 | The AI-era expansion is not just marketing: CockroachDB now documents billions-scale vector indexing, owner-aware partitioning, and security controls for AI agents in the main product surface. | 高 | SE020, SE024, SE021, SE019 |
| CE036 | Public sources still cannot fully quantify how CockroachDB’s performance compares with direct rivals on customer-specific workloads, because the strongest performance claims remain vendor-authored rather than independently benchmarked. | 低 | SE019, SE001, SE031, SE032 |
| CE037 | The best product-tech verdict is that CockroachDB has evolved from a narrowly distributed-SQL proposition into a broad transactional platform with serious multi-region, compliance, and AI-era credibility. | 中 | SE002, SE015, SE019, SE018, SE020 |
| CE038 | Cockroach Labs continues to make resilience a first-class product theme through explicit performance-under-adversity positioning and outage-focused messaging, reinforcing that reliability is sold as a feature rather than only as an implementation detail. | 中 | SE025, SE026, SE011 |
| CU001 | Cockroach Labs publicly presents CockroachDB as trusted by enterprises across banking and fintech, retail and eCommerce, software and tech, media and streaming, gaming, manufacturing and logistics, and gambling. | 高 | SU002, SU001 |
| CU002 | The 2021 Series F materials said Cockroach Labs had more than 200 customers and tens of thousands of deployed clusters, establishing a meaningful commercial base well before the 2026 run date. | 高 | SU004, SU005 |
| CU003 | Those same Series F materials said more than 50 percent of customers were already on CockroachDB Dedicated, implying that managed-cloud adoption had become central to the customer mix early. | 高 | SU004, SU005 |
| CU004 | Apps Run The World tracks CockroachDB deployments across multiple organizations and geographies, supporting the view that adoption is not limited to a handful of logos even though its counts should be treated as directional. | 中 | SU011, SU002 |
| CU005 | ReadyContacts advertises a 226-company Cockroach Labs customer list and emphasizes vertical and geography slicing, which is useful only as directional third-party evidence rather than as a definitive paying-customer count. | 低 | SU012, SU004 |
| CU006 | Cockroach Labs’ visible customer proof is weighted toward mission-critical operational workloads rather than casual departmental apps. | 高 | SU002, SU003, SU004 |
| CU007 | Booking.com uses CockroachDB for its Order Platform, a highly available order-management system supporting the company’s Connected Trip initiative. | 中 | SU013 |
| CU008 | Booking.com migrated that platform from Cassandra after needing ACID guarantees, CDC, secondary indexes, and lower operational overhead. | 中 | SU013 |
| CU009 | Booking.com says its migrated Order Platform now holds around 20 TB of data on CockroachDB Cloud across multiple European regions. | 中 | SU013 |
| CU010 | Booking.com cites a 99.99998 percent SLI for writes as acceptable for its reservation use case, showing that CockroachDB is used in a high-availability consumer booking workflow. | 中 | SU013 |
| CU011 | Shipt built a distributed payment system on CockroachDB to prioritize correctness, concurrency control, and multi-region availability. | 中 | SU015 |
| CU012 | Shipt publicly describes a deployment with 12 nodes across four regions, about 1 to 2 million payment transactions per day, and a 99.999 percent availability target. | 中 | SU015 |
| CU013 | Shipt explicitly preferred CockroachDB over Spanner because it wanted the option to deploy across multiple clouds rather than accept single-cloud lock-in. | 中 | SU015 |
| CU014 | Form3 uses CockroachDB as the backbone of a managed payments platform serving banks and fintechs across the UK, EU, and US. | 中 | SU016 |
| CU015 | Form3 runs CockroachDB across AWS, GCP, and Azure with a replication factor of three, presenting one of the clearest public examples of true multi-cloud production use. | 中 | SU016 |
| CU016 | Form3 reports strict 100 and 200 millisecond P99 SLAs, about 700 TPS, and customer-local topology placement, indicating that CockroachDB supports latency-sensitive regulated payment flows. | 中 | SU016 |
| CU017 | Form3 says customers such as Lloyds Bank and Nationwide Building Society rely on its platform, indirectly showing that CockroachDB sits beneath major financial-institution workloads. | 中 | SU016, SU023 |
| CU018 | Riskified says it chose CockroachDB to remove a PostgreSQL single-writer bottleneck while preserving PostgreSQL compatibility for a client-facing fraud platform. | 高 | SU019, SU024 |
| CU019 | Riskified reports 99+ percent customer retention, over 10,000 online transactions per second, and a zero-downtime migration, making it one of the strongest customer-quality proof points in the public set. | 高 | SU019, SU024 |
| CU020 | Riskified’s story reinforces that security, data integrity, and elasticity matter to customers whose own end users never directly see the database layer. | 中 | SU019 |
| CU021 | Route uses CockroachDB to power always-on data for more than 1 billion orders and says it serves more than 13,000 brands and millions of active app users. | 中 | SU020 |
| CU022 | Route reports about 52 TB of storage and multiple billion-plus-record tables on CockroachDB, supporting the claim that the product can handle large relational operational datasets. | 中 | SU020 |
| CU023 | Route also says paid support delivered unusually strong ROI, a useful signal that support quality can reinforce retention and expansion for sophisticated customers. | 中 | SU020 |
| CU024 | Netflix now operates more than 380 CockroachDB clusters, including over 160 production clusters and more than 60 multi-region clusters, after first offering CockroachDB-as-a-Service internally in 2020. | 中 | SU017 |
| CU025 | Netflix describes CockroachDB as its database solution for applications needing consistency, high availability, and region failover, showing internal standardization beyond a single isolated use case. | 中 | SU017 |
| CU026 | Netflix’s gaming platform uses a 48-node cluster across four regions, illustrating that customer proof extends into newer multi-region entertainment workloads as well as traditional media systems. | 中 | SU017 |
| CU027 | DoorDash publicly operates CockroachDB as a fully abstracted internal database service, with about 2,300 nodes across 300+ clusters, roughly 1.2 million peak QPS, 1.9 petabytes on disk, and close to 900 changefeeds. | 中 | SU021 |
| CU028 | DoorDash’s migration story shows CockroachDB winning a hard production replacement against Aurora Postgres because outages and scaling bottlenecks had become unacceptable. | 高 | SU022, SU021 |
| CU029 | Hard Rock Digital uses CockroachDB for a multi-region sportsbook that must satisfy state-by-state data residency and availability requirements. | 中 | SU018 |
| CU030 | Hard Rock reports running about 100 database nodes with 32 vCPUs each during NFL peak season, then scaling down to roughly one-third of that footprint afterward without downtime. | 中 | SU018 |
| CU031 | Hard Rock and Form3 both show that regulatory or jurisdictional constraints are recurring purchase drivers, not edge cases, for the CockroachDB customer base. | 高 | SU018, SU016, SU023 |
| CU032 | The financial-services use-case page says Cockroach Labs has worked with dozens of companies including Fortune 50 banks, supporting the idea that enterprise demand reaches beyond the handful of named case studies. | 中 | SU023 |
| CU033 | SumUp’s story shows CockroachDB supporting a global payments migration where downstream analytics and back-office tools had to remain consistent during cutover. | 中 | SU014 |
| CU034 | Superbet and Hard Rock together indicate that online betting and gaming are repeatable customer segments, not just one-off anecdotes. | 中 | SU025, SU018 |
| CU035 | Across Booking.com, Riskified, DoorDash, and Route, the recurring pattern is migration off legacy relational or adjacent systems into CockroachDB for scale, resilience, and less application-layer complexity. | 高 | SU013, SU019, SU022, SU020 |
| CU036 | Across Shipt, Form3, Hard Rock, and Netflix, multi-region survivability is sold and used as an operating requirement rather than as optional architecture polish. | 高 | SU015, SU016, SU018, SU017 |
| CU037 | Multiple customer stories mention direct help from Cockroach Labs account teams, support teams, or professional services, implying that post-sale delivery is part of customer success for complex deployments. | 中 | SU020, SU018, SU019 |
| CU038 | The roster of named customers skews toward sophisticated engineering organizations with platform teams, which likely supports larger contract values but may narrow the accessible buyer pool. | 中 | SU017, SU021, SU016, SU013, SU026 |
| CU039 | Public customer evidence is unusually rich in depth but still mostly vendor-authored, which means the quality of individual stories is stronger than the independence of the overall proof set. | 中 | SU002, SU013, SU016, SU017 |
| CU040 | Public sources do not pin down an exact 2026 paying-customer count, because official figures are stale and third-party databases are inconsistent in method and precision. | 中 | SU004, SU011, SU012, SU007 |
| CU041 | The customer base appears to offer strong expansion potential because many visible workloads are central transaction systems, control planes, or regulated platforms with high switching costs once deployed. | 中 | SU020, SU016, SU017, SU018 |
| CU042 | GetLatka’s late-2025 headcount estimate around 715 employees is consistent with a company large enough to support a substantial enterprise customer base, although it does not itself prove customer count or retention. | 中 | SU007, SU008 |
| CU043 | GitHub and Stack Overflow signals suggest developer familiarity with PostgreSQL and open-source ecosystems remains high, which likely helps Cockroach land customers that want relational tools without abandoning modern distributed architecture. | 中 | SU027, SU028, SU001 |
| CU044 | The best customer verdict is that Cockroach Labs has high-quality reference customers and real production depth across several demanding verticals, but public evidence still leaves exact customer count and independent retention quality less certain than the logo roster implies. | 高 | SU002, SU004, SU019, SU017, SU020, SU016 |
| CR001 | The 2026 technical-advisory record shows that CockroachDB’s risk surface is not theoretical: public advisories cover privilege escalation, index corruption, live-data deletion, silent import data loss, and backup incompleteness. | 高 | SR006, SR014 |
| CR002 | A May 2026 advisory disclosed two privilege-escalation vulnerabilities that could let authenticated users gain elevated privileges beyond those assigned to their account. | 中 | SR006 |
| CR003 | An April 2026 advisory disclosed a partial-index corruption bug during backfill on multi-column-family tables under concurrent updates. | 中 | SR006 |
| CR004 | A February 2026 advisory disclosed a race condition between MVCC garbage collection and range splits that could delete live data under rare conditions. | 中 | SR006 |
| CR005 | Another February 2026 advisory said object-storage read failures during AVRO imports could lead to silent data loss without an error being reported to the user. | 中 | SR006 |
| CR006 | The advisory log also includes earlier backup, restore, changefeed, and encryption-related defects, reinforcing that distributed-database correctness is an ongoing operational discipline rather than a solved problem. | 中 | SR006 |
| CR007 | The production checklist says fault tolerance depends on explicit topology choices such as at least three nodes or three regions, uniform locality labeling, and avoiding multiple nodes on one machine, meaning resilience is partly purchased through careful design rather than granted automatically. | 高 | SR010, SR011 |
| CR008 | The same checklist recommends minimum CPU, RAM, IOPS, and hardware-uniformity practices, implying that under-provisioning or heterogeneous fleets can directly degrade stability and performance. | 高 | SR010, SR012 |
| CR009 | CockroachDB documents hotspots as bottlenecks that scaling out alone may not solve, including hot rows, index tails, queueing hotspots, and lookback hotspots. | 中 | SR013 |
| CR010 | Hotspot documentation explicitly says some write patterns can limit a distributed cluster to the performance of a single node, which is a meaningful risk for teams expecting effortless horizontal scale. | 中 | SR013 |
| CR011 | Upgrade documentation requires healthy replication, load-balancing, backup validation, review of skipped-release notes, and sometimes manual finalization, showing that version changes carry real operational risk. | 高 | SR009, SR008 |
| CR012 | Once a major-version upgrade is finalized, the cluster cannot be rolled back to the prior major version. | 中 | SR009 |
| CR013 | Innovation releases have shorter support windows and no Assistance Support phase, which increases the penalty for version drift or poor upgrade timing. | 高 | SR008, SR009 |
| CR014 | CockroachDB v26.1 is an Innovation release that reaches end of support in 2026, underscoring how quickly customers can fall onto unsupported paths if they standardize on the wrong release. | 中 | SR008 |
| CR015 | The status-history page and advisories together show transparency, but they also confirm the normal reality of a complex database platform: bugs and service incidents are not hypothetical edge cases. | 中 | SR007, SR006 |
| CR016 | Customer stories repeatedly describe direct reliance on Cockroach Labs support or account teams for setup, migration, tuning, or expansion, implying some deployments remain operationally non-trivial even after purchase. | 高 | SR030, SR033, SR034, SR031 |
| CR017 | Hard Rock says Cockroach Labs helped set up a complicated state-by-state topology, while Route says account-team familiarity improved problem resolution, illustrating real services dependency in sophisticated accounts. | 高 | SR030, SR034 |
| CR018 | The financial-sector push raises regulatory stakes because DORA imposes ICT risk management, incident reporting, resilience testing, and third-party-risk obligations on financial entities and their providers. | 高 | SR019, SR023 |
| CR019 | Because Form3 and other financial customers use CockroachDB in payment-critical flows, Cockroach Labs is exposed to more stringent diligence expectations than a generic developer-tool vendor. | 中 | SR031, SR019, SR023 |
| CR020 | Gaming and sportsbook deployments face legal-topology complexity because the Wire Act and state-locality requirements can force bespoke placement and data-residency designs. | 高 | SR020, SR030 |
| CR021 | Hard Rock’s case study shows that those gaming requirements are not academic: the company needed database gateway nodes in each state and a mix of AWS Regions, Local Zones, and Outposts. | 高 | SR030, SR020 |
| CR022 | The privacy policy says Cockroach Labs logs device, access, and interaction information and may transfer personal information across borders, which increases data-governance scrutiny for regulated or privacy-sensitive buyers. | 中 | SR018, SR004 |
| CR023 | Cloud terms say customers remain responsible for properly configuring and securing their accounts and associated credentials, leaving meaningful shared-responsibility risk with the customer. | 中 | SR016 |
| CR024 | Cloud terms also say beta services and free offerings can be terminated or suspended at any time, may be incomplete, and receive no indemnification or support. | 中 | SR016 |
| CR025 | Cockroach Labs explicitly disclaims uninterrupted or error-free service in both website and SaaS legal materials, which is standard legally but still matters because the company sells always-on infrastructure. | 高 | SR015, SR016 |
| CR026 | The SaaS terms allow suspension for nonpayment, suspected risky use, or bankruptcy-like events and allow termination for convenience with notice, giving Cockroach Labs wide contractual control over service continuity. | 中 | SR016 |
| CR027 | The acceptable-use policy restricts penetration testing without prior written consent and sets conditions around public benchmarking, which can slow independent validation and create go-to-market friction in competitive bake-offs. | 中 | SR017 |
| CR028 | PostgreSQL remains a free, mature default with deep ecosystem familiarity, so Cockroach must continually justify complexity, migration cost, and pricing against a strong status-quo alternative. | 高 | SR026, SR036, SR033 |
| CR029 | Aurora and Spanner create a different kind of competitive risk: for cloud-aligned buyers they may look operationally simpler or more naturally bundled, even if they reduce portability. | 中 | SR024, SR025, SR031, SR025 |
| CR030 | Because many visible customers are sophisticated platform or regulated teams, Cockroach Labs may face TAM and sales-cycle risk if the product remains too complex for mainstream relational workloads. | 中 | SR034, SR032, SR036, SR021 |
| CR031 | Migration-led wins at DoorDash, Riskified, Booking.com, and Form3 show strong product value, but they also imply longer implementation cycles and a heavier need for solution engineering. | 高 | SR036, SR033, SR035, SR031 |
| CR032 | The combination of support-heavy accounts, regulated workloads, and complex topologies can pressure gross margin if service intensity rises faster than automation. | 中 | SR030, SR034, SR033, SR027 |
| CR033 | Cockroach markets portability, but many public stories still rely on major cloud primitives or managed Kubernetes offerings, so multi-cloud value does not eliminate dependency on hyperscaler economics and outages. | 中 | SR031, SR030, SR011 |
| CR034 | Independent critique emphasizes hidden implementation and optimization costs, supporting the idea that total cost of ownership can be harder than headline pricing suggests. | 中 | SR021, SR003 |
| CR035 | TrustRadius provides only thin independent review depth versus the richness of vendor case studies, leaving external validation of pain points and renewal sentiment underdeveloped. | 低 | SR022, SR002 |
| CR036 | Public sources do not disclose current NRR, GRR, customer concentration, burn, or runway, which limits precise risk weighting across commercial, financial, and support dimensions. | 中 | SR027, SR028, SR029 |
| CR037 | Private-company opacity is especially important in this case because many visible strengths—high-quality customers, portability, and resilience—could still mask uneven margin, concentration, or support-cost realities. | 中 | SR027, SR034, SR030, SR031 |
| CR038 | The strongest mitigations are visible and practical: use supported releases, patch aggressively, design topology carefully, validate backups, and test workloads for hotspots before scale forces a redesign. | 高 | SR008, SR009, SR010, SR013 |
| CR039 | The clearest kill criteria would be evidence of weak renewal economics, major unresolved customer concentration, or a pattern of severe correctness incidents in supported production releases. | 中 | SR006, SR027, SR028 |
| CR040 | Risk transmission is fast in this business: a serious data-integrity event can hit trust, renewals, support load, and valuation simultaneously because the database sits directly inside customer revenue workflows. | 高 | SR006, SR033, SR035, SR034 |
| CR041 | Competition, complexity, and support intensity are linked risks rather than separate ones: if Cockroach is only clearly superior in the hardest workloads, customer acquisition may remain selective and expensive. | 中 | SR026, SR024, SR021, SR036 |
| CR042 | The best overall risk verdict is that Cockroach Labs’ principal threats are execution and operational-discipline risks rather than existential product disbelief: the product is credible, but the company must keep complex deployments safe, supportable, and economically attractive against simpler alternatives. | 高 | SR006, SR008, SR016, SR026, SR024, SR030 |
| CV001 | Cockroach Labs raised $278 million in Series F financing in December 2021 at a $5 billion valuation. | 高 | SV003, SV004 |
| CV002 | The 2021 financing disclosure said lifetime funding had reached $633 million at that point. | 高 | SV003, SV004 |
| CV003 | Third-party sources in mid-2026 imply secondary-market valuation around $6.9 billion, roughly 38 percent above the last primary round. | 高 | SV008, SV009, SV010 |
| CV004 | Secondary-market price discovery can validate investor interest, but unlike a new primary round it does not inject fresh operating capital onto the balance sheet. | 高 | SV008, SV009, SV010 |
| CV005 | GetLatka’s 2023 $128.3 million revenue or ARR estimate remains the clearest public revenue anchor for Cockroach Labs. | 中 | SV006 |
| CV006 | GetLatka’s late-2025 headcount estimate around 715 employees supports the view that Cockroach is operating at meaningful late-stage scale with a substantial cost base. | 中 | SV006, SV007 |
| CV007 | The 2026 momentum release highlights AI-scale resilience, performance, and partner momentum but does not disclose current revenue, margin, or retention metrics. | 中 | SV005 |
| CV008 | At the 2021 $5 billion primary valuation and the 2023 $128.3 million revenue anchor, Cockroach Labs would have traded at about 39x revenue. | 中 | SV003, SV006 |
| CV009 | At the 2026 $6.9 billion secondary marker and the same $128.3 million revenue anchor, Cockroach Labs would imply about 53.8x revenue. | 中 | SV008, SV006 |
| CV010 | Even if Cockroach had reached $170 million of annualized revenue by 2025-2026, the 2021 primary price would still imply roughly 29.4x revenue. | 中 | SV003, SV006 |
| CV011 | At that same $170 million revenue test point, the 2026 $6.9 billion secondary valuation would still imply roughly 40.6x revenue. | 中 | SV008, SV006 |
| CV012 | Even at a more generous $200 million revenue test point, the 2026 secondary marker would still imply about 34.5x revenue. | 中 | SV008, SV006 |
| CV013 | MongoDB’s July 2026 market cap and TTM revenue imply about a 10.2x revenue multiple. | 中 | SV012, SV013 |
| CV014 | Confluent’s July 2026 market cap and TTM revenue imply about a 9.6x revenue multiple. | 中 | SV014, SV015 |
| CV015 | Snowflake’s July 2026 market cap and TTM revenue imply about a 19.9x revenue multiple, representing a premium upper bound for public data-infrastructure software. | 中 | SV016, SV017 |
| CV016 | Elastic’s July 2026 market cap and TTM revenue imply about a 3.9x revenue multiple, illustrating how sharply mature infrastructure multiples can compress. | 中 | SV018, SV019 |
| CV017 | The public comp range from roughly 3.9x to 19.9x suggests that Cockroach’s implied private-market multiple is rich versus current public software infrastructure benchmarks. | 中 | SV012, SV013, SV014, SV015, SV016, SV017, SV018, SV019 |
| CV018 | For a $6.9 billion valuation to look merely Snowflake-like on a 19.9x multiple, Cockroach would need roughly $347 million of annual revenue. | 中 | SV016, SV017, SV008 |
| CV019 | For a $6.9 billion valuation to look MongoDB-like on a 10.2x multiple, Cockroach would need roughly $677 million of annual revenue. | 中 | SV012, SV013, SV008 |
| CV020 | Those thresholds are materially above the last public Cockroach revenue anchor, which is why the current valuation debate turns more on hidden growth and retention quality than on visible numbers. | 中 | SV006, SV008, SV016, SV012 |
| CV021 | Premium customer proof from DoorDash, Netflix, Route, Riskified, Form3, and Hard Rock supports paying more than a generic database-software multiple. | 高 | SV024, SV025, SV026, SV027, SV028, SV029 |
| CV022 | The product and customer evidence together support a premium to lower-quality infrastructure names because Cockroach sits in mission-critical, high-switching-cost operational systems. | 中 | SV024, SV003, SV025, SV026 |
| CV023 | The main anti-thesis is not lack of product credibility but valuation fullness: public data does not yet prove enough revenue, margin, or retention quality to justify a 30x-plus revenue profile with confidence. | 中 | SV006, SV008, SV020 |
| CV024 | Independent market data frames the distributed-database segment at roughly $4.48 billion in 2026, which makes a $6.9 billion private valuation look aggressive if the company is underwritten only against a narrow category framing. | 低 | SV021 |
| CV025 | Broader cloud-database and distributed-SQL market reports indicate that category framing matters enormously; a narrow distributed-database lens is less forgiving than a broad cloud-data-infrastructure lens. | 中 | SV022, SV023, SV021 |
| CV026 | AI-era positioning and partner momentum can help preserve valuation narrative, but without updated financials they do not by themselves de-risk the entry price. | 中 | SV005, SV008 |
| CV027 | Historical growth quality did support a premium earlier: the Series F materials said ARR had tripled year over year and cloud revenue had risen 500 percent in the prior quarter. | 高 | SV003, SV004 |
| CV028 | The same materials said Cockroach had more than 200 customers and that more than half were already on Dedicated, which supports the idea of improving revenue quality entering 2022. | 高 | SV003, SV004 |
| CV029 | MongoDB’s 10-K illustrates the level of disclosure a mature public database company provides around revenue, risk, and operations—detail that Cockroach does not yet provide publicly. | 中 | SV020, SV006, SV005 |
| CV030 | Because Cockroach remains private and opaque, the valuation case depends unusually heavily on inference from customer quality, category positioning, and secondary price signals. | 中 | SV006, SV008, SV024 |
| CV031 | Support intensity, migration-heavy sales, and operational complexity deserve some discount versus cleaner self-serve software stories, even if the absolute product quality is high. | 中 | SV025, SV026, SV029, SV030 |
| CV032 | Competition from free PostgreSQL and managed cloud databases also limits how much multiple expansion an investor should assume without evidence of exceptional net retention or margin structure. | 中 | SV031, SV002, SV020 |
| CV033 | A defensible bull case requires Cockroach to show materially higher current revenue than the 2023 public anchor, continued premium customer wins, and evidence that support intensity does not overwhelm economics. | 中 | SV006, SV024, SV025, SV026 |
| CV034 | A defensible bear case is that Cockroach remains a high-quality product with a rich secondary price but insufficient public proof of scale to justify paying above high-teens infrastructure multiples. | 中 | SV008, SV006, SV018, SV019 |
| CV035 | The base case is neither collapse nor euphoria: a high-quality infrastructure company that likely deserves a premium to average software names, but not a blind premium to the best public franchises. | 中 | SV012, SV016, SV024, SV006 |
| CV036 | A reasonable low/base/high valuation frame today is roughly $3.0-4.5B bear, $4.8-6.2B base, and $6.5-8.0B bull, with the current secondary marker sitting in the upper end of what can be defended publicly. | 中 | SV008, SV006, SV016, SV018 |
| CV037 | That range means the $5B primary round can still be defended as strategic late-2021 pricing, but the 2026 secondary marker already prices in meaningful positive hidden information. | 中 | SV003, SV008, SV006 |
| CV038 | The recommendation implied by public evidence is not “avoid the company”; it is “continue diligence, but treat current pricing as full unless private data materially improves the revenue and efficiency picture.” | 高 | SV024, SV006, SV008, SV020 |
| CV039 | The stance becomes more bullish if diligence confirms revenue well above $200 million, strong net retention, credible margin structure, and limited concentration. | 中 | SV006, SV020, SV008 |
| CV040 | The thesis breaks if hidden data instead shows weak renewal economics, customer concentration, support-heavy gross margin drag, or slower-than-assumed revenue scaling. | 中 | SV006, SV020, SV030 |
| CV041 | The most important diligence asks are current ARR or revenue, NRR and GRR, gross margin, support-cost intensity, cash runway, and customer concentration. | 中 | SV006, SV005, SV020 |
| CV042 | The best valuation stance is that Cockroach Labs is a high-quality late-stage infrastructure asset priced at a demanding level; the company is attractive, but the current public evidence does not make the current secondary price look clearly cheap. | 高 | SV008, SV006, SV024, SV012, SV016 |