ClickHouse
开源分析数据库龙头有强产品拉力,但后期估值已提前计入繁重执行。
ClickHouse 有很强的产品市场拉力、可信的云商业化和头部客户采用,但私募市场估值已经提前计入大量未来执行,公开披露又不足以看清收入质量和利润率。
封面要素
公司概况
ClickHouse 是开源 ClickHouse 分析数据库背后的商业公司,把开发者驱动的采用飞轮,与面向实时分析、可观测性和新兴 AI 数据工作负载的托管云产品结合在一起。公司实体在 2021 年围绕 Yandex 内部起源、2016 年开源的技术成立。公开证据显示,公司以 San Francisco 为中心,工程团队全球分布,客户增长很快,融资画像在 2025 年明显扩张。
- 成立时间
- 2021-08-25
- 创始人
- Aaron Katz, Alexey Milovidov, Yury Izrailevsky
- 创立地点
- Portola Valley, California, United States
- 总部
- San Francisco, California, United States
- 产品
- ClickHouse 销售一套开源列式 OLAP 数据库,以及 ClickHouse Cloud;后者是全托管服务,负责 AWS、Azure 和 GCP 上的扩缩容、运维与基础设施。
- 客户
- 开发者、数据平台团队,以及运行实时分析、可观测性和 AI 相邻数据工作负载的企业。
- 商业模式
- 开源分发导入按用量计费的托管云收入,并通过专用部署和 BYOC 部署向企业扩张。
- 阶段
- Series C private company
- 融资情况
- 2025 年 5 月完成 $350M Series C 融资,估值约 $6.35B,披露融资总额超过 $650M。
执行摘要
主要优势
- 广泛采用的分析引擎和庞大贡献者社区撑起强开源分发,也让 ClickHouse 占住开发者心智。
- 多云可用、按用量计费,加上 ARR 快速扩张的证据,使云商业化更可信。
- 企业和互联网规模客户可被引用,包括 Cloudflare、Contentsquare,且已有具体性能和成本效果。
主要风险
- 公开披露仍缺少承销下行保护所需的 ARR 质量、毛利率、留存和现金流可见度。
- Series C 价格隐含 ARR 倍数大约在 30 多倍中段到 40 倍出头,高于许多公开和私有可比公司。
- DuckDB、StarRocks、Snowflake、Databricks 和超大规模云厂商带来的开源与捆绑式数据仓库竞争,可能压缩定价和扩张空间。
未决问题
- 队列层面的 ARR 质量、留存和毛利率数据仍未披露。
- 公开来源看不到客户集中度和头部账户暴露。
- 大规模开源使用能否转化为持久付费云收入,公开证据仍未证明其经济性。
目录
01公司概况
1.1 身份、起源与运营足迹
理解 ClickHouse,最简单的切入是把它看作两个相连的实体:一个 2009 年始于 Yandex 内部的分析数据库项目,以及一家 2021 年在 Delaware 注册、面向全球商业化该项目的公司。技术谱系很重要,因为它解释了 ClickHouse 为什么一进入市场就带着少见的性能可信度:在成为一家风险投资支持的创业公司之前,它已经为 Yandex.Metrica 级别的分析工作负载打磨多年。现在,这家商业公司把产品描述为面向实时分析的高速、开源、列式数据库,变现重心放在 ClickHouse Cloud,以及相邻的实时分析、可观测性和 AI/ML 工作负载上。 运营足迹同样是混合形态。官方历史把总部放在 San Francisco Bay Area,而当前第三方资料有的标注 San Francisco,有的标注 Portola Valley 或 Palo Alto。这些标签在街道地址层面并不一致,但在更大的判断上一致:ClickHouse 是一家 Bay Area 总部公司,在 Amsterdam 有重要办公室,并有意采用分布式团队。官方来源称员工分布在 10 多个国家,公开资料也支持到 2026 年员工规模超过 500 人。因此,在后续章节里,最稳妥的简称是:Bay Area 总部、Amsterdam 工程与欧洲枢纽、全球分布式运营。[CO001, CO003, CO004, CO008, CO009, CO010]
| 指标 | 数值 / 状态 | 日期 | 置信度 | 缺口 / 备注 |
|---|---|---|---|---|
| 项目启动 | 2009 年在 Yandex 内部 | 2009 | 高 | 项目起源,不是公司成立 |
| 商业公司成立 | Delaware 公司 | Aug 2021 | 高 | 与 2009 年项目和 2012 年生产部署不同 |
| 开源发布 | Apache 2.0 | 2016 | 高 | 开发者采用的基础 |
| 总部 / 办公室布局 | 湾区总部;Amsterdam 办公室 | 当前 | 中 | 公开来源中同时出现 Portola Valley、Palo Alto 和 San Francisco 标签 |
| 员工数 | 531-569 (500+) | Apr-May 2026 | 中 | PitchBook 和 Tracxn 数据不同,但都支持后期阶段规模 |
| 客户数 | >2,000 | 2025 | 中 | 公司披露;未经独立审计 |
| ARR / 收入标记 | ~$160M ARR;H1 2025 年化收入 ~$100M | 2025 | 低 | ARR 来自第三方估计,未经审计 |
| Series B | $250M,估值 $2B | Oct 2021 | 高 | 官方新闻稿和投资人数据库相互佐证 |
| Series C | $350M,估值约 $6.35B | May 2025 | 中 | 公司新闻稿确认融资;估值来自第三方报道 |
| Series C 后累计融资 | >$650M,另有 $100M 信贷额度 | May 2025 | 中 | 为保持章节一致性,未纳入 2025/2026 年后续活动 |
各行对齐项目时间线、公司成立和 2025 年规模标记;估值和 ARR 仍来自第三方,而非公司经审计披露。
[CO001, CO004, CO010, CO014, CO015, CO016]公司逻辑把开源起源、云变现、资本、全球团队和风险控制串起来。
[CO004, CO008, CO009, CO011, CO015, CO016]精选 KPI 概括成熟度、牵引力,以及从开源采用到后期规模的连续性。
1.2 创始人、领导层与治理
创始团队把原始技术作者身份、企业级商业化经验和产品规模化经验结合在一起。Alexey Milovidov 在 Yandex 内部创造了 ClickHouse,并继续以 CTO 身份扮演技术锚点。Aaron Katz 曾任 Salesforce 和 Elastic 高管,带来商业化打法并担任 CEO。Yury Izrailevsky 过往曾在 Netflix 和 Google 担任高级工程领导,现任总裁,负责把一个受开发者喜爱的开源项目,接到全球软件公司的产品与工程规模化轨道上。 治理更受投资人塑形,但公开透明度有限。Index 的长篇起源故事清楚显示,Index Ventures 的 Mike Volpi 和 Benchmark 的 Peter Fenton 不是被动出资人;他们与 Katz 一起设计了拆分独立结构,创立了一家独立、由多数股权控制的 Delaware 公司。其他当前资料显示,Volpi 和 Fenton 都在 ClickHouse 董事会任职。公司因此获得了经验丰富的基础设施软件监督,尽管更广泛的董事会和观察员名单并未完全公开。2025 年,ClickHouse 又任命 Kevin Egan 为 CRO、Mariah Nagy 为人力副总裁、Jimmy Sexton 为 CFO,领导层厚度进一步提升。这三人重要,因为他们表明公司正从创始人主导的组建期,转向覆盖收入、人才和财务的后期运营纪律。[CO005, CO006, CO007, CO021, CO022, CO033]
| 人物 | 角色 | 背景 | 覆盖范围 / 相关性 | 关键人物依赖 |
|---|---|---|---|---|
| Aaron Katz | 联合创始人兼 CEO | Salesforce 和 Elastic 前运营高管 | 商业化、融资和客户叙事 | 高 |
| Alexey Milovidov | 联合创始人兼 CTO | 在 Yandex 内部创建 ClickHouse | 核心架构和技术可信度 | 高 |
| Yury Izrailevsky | 联合创始人兼总裁 | Netflix 和 Google 前工程负责人 | 产品和工程扩张 | 高 |
| Kevin Egan | 首席营收官 | Atlassian、Slack、Dropbox、Salesforce 前高管 | 2025 年企业销售扩张 | 中 |
| Mariah Nagy | 人力副总裁 | Weights & Biases、Confluent、SurveyMonkey 前高管 | 分布式团队人才体系 | 低 |
| Jimmy Sexton | 首席财务官 | Snowflake 和 ServiceNow 前财务负责人 | 后期阶段财务纪律 | 中 |
| Mike Volpi | 董事会成员 / Index Ventures | Index 退休合伙人;长期投资开源 | 自公司成立起延续的投资人治理 | 中 |
| Peter Fenton | 董事会成员 / Benchmark | Benchmark GP,曾投出多个开源和基础设施成功案例 | 自公司成立起延续的投资人治理 | 中 |
这是对创始人、2025 年新增高管和公开可见投资人董事的部分但与决策相关的枚举;完整组织架构并未公开。
[CO005, CO006, CO007, CO021, CO022, CO033]1.3 资本基础、投资人延续性与商业规模
ClickHouse 的融资历史显示出少见的压缩式风险投资轨迹。2021 年 8 月公司拆分独立时,Index Ventures 和 Benchmark 领投 $50 million Series A。仅约两个月后,公司又以 $2 billion 估值完成 $250 million Series B,由 Coatue 和 Altimeter 领投,Benchmark、Lightspeed、Almaz 及其他成长型投资人参投。2021 年这轮密集融资为云产品建设提供资金,也把 ClickHouse 从一个受推崇的开源项目,推成一家有足够资本全球扩张的公司。 下一次重大重估发生在 2025 年 5 月。公司和投资人公告确认,ClickHouse 完成 $350 million Series C,由 Khosla Ventures 领投,新投资人 BOND、IVP、Battery Ventures 和 Bessemer 加入,跟投方包括 Benchmark、Coatue、Lightspeed、FirstMark、GIC 和 Nebius。Goodwin 与公司公告还确认了一笔由 Stifel 和 Goldman Sachs 牵头的 $100 million 信贷额度。第三方报道把该轮估值放在约 $6.35 billion 至 $6.4 billion,累计融资超过 $650 million。公开运营信号符合后期成长公司画像:2025 年客户数超过 2,000 家;公司称年增长超过 300%;2025 年中公开收入 / ARR 指标约为年化 $100 million,2025 年底 ARR 约 $160 million;到 2026 年员工数处于 500 人出头到中段区间。[CO017, CO018, CO019, CO020, CO023, CO024]
| 利益相关方 | 轮次 / 工具 | 角色 | 重要性 / 控制 | 尽调要求 |
|---|---|---|---|---|
| Index Ventures | Series A;后续轮次跟投 | 创始机构支持方 | 公司成立赞助方;Mike Volpi 任董事 | 确认当前持股和任何按比例跟投权 |
| Benchmark | Series A 轮、Series B 轮、Series C 轮 | 创始并持续支持的 VC | Peter Fenton 董事席位增加治理权重 | 确认持股、董事会委员会角色和储备资金 |
| Coatue | Series B 领投;Series C 跟投 | 成长投资人 | 验证 2021 年重估,并一直留到 2025 年 | 确认当前持仓规模和任何老股交易活动 |
| Altimeter | Series B 联合领投 | 2021 年成长投资人 | 帮助确立 $2B 估值台阶 | 厘清其在 2021 年后是否仍活跃 |
| Lightspeed | Series B;可见的持续投资人 | 成长 VC 与组合赞助方 | 公开组合页面把该机构直接绑定到 2021 年商业化扩张 | 确认持股和董事会观察员权利 |
| Almaz Capital | Series B 参与方 | 早期成长投资人 | 显示 2021 年轮次的跨境投资人组合 | 确认 Series C 后是否维持持股 |
| Khosla Ventures | Series C 领投 | 2025 年领投方 | 主导公司在 AI 时代的重大估值重置 | 审查条款、优先权和治理权利 |
| Stifel / Goldman Sachs | 2025 年 5 月信贷额度 | 债务提供方 | 给资本结构引入杠杆和契约考量 | 审查契约、提款条件和留置权 |
| Nebius / Yandex 历史遗留 | 2021 年贡献;后续认股权证 | 残余历史利益相关方 | 虽然 2025 年报道没有报告股权,但对地缘政治和股权结构解读很重要 | 确认认股权证机制和到期期限 |
该地图覆盖留存公开来源中明确可见的投资人和贷款方;它不是完整股权结构表,也不应被视作详尽所有权披露。
[CO017, CO018, CO019, CO020, CO023, CO024]1.4 里程碑、地缘政治背景与已披露风险
里程碑记录足够强,可以为报告其余部分建立可复用的事实底座。技术起源始于 2009 年,2012 年进入生产使用,2016 年开源。商业公司在 2021 年通过拆分独立成立,并连续完成 Series A 和 Series B 融资。下一阶段运营节点出现在 2022 年:ClickHouse 开设 Amsterdam 办公室,推出云产品早期访问,并公开说明俄罗斯入侵 Ukraine 后如何把工程人才迁出 Russia。这些动作不只是历史脚注;它们构成了治理和客户信任叙事的一部分,也让后来的企业客户增长成为可能。 因此,关键的反向视角不是典型的需求问题,而是信任与执行问题。ClickHouse 在进入西方企业销售时带着 Yandex 和 Russia 起源认知风险,也积累了公开披露的安全漏洞,包括多个内存安全问题和一个 2024 年查询缓存访问控制漏洞。公司的缓释叙事可信,但并非没有成本:它把法定住所与 Russia 切开,把工程师迁到 Amsterdam,强调西方投资人和董事,并持续交付产品里程碑,例如收购 HyperDX、发布 OpenHouse、完成 2025 年 Series C。对尽调来说,核心问题已经不再是 ClickHouse 是否摆脱了起源叙事,而是治理透明度、安全流程和后期运营控制,是否跟上了商业扩张速度。[CO002, CO010, CO012, CO028, CO029, CO032]
| 日期 | 事件 | 类型 | 金额 / 状态 | 参与方 | 含义 |
|---|---|---|---|---|---|
| 2009 | 实验性分析数据库项目在 Yandex 内部启动 | 创立 | Alexey Milovidov 和 Yandex 团队 | ClickHouse 的技术起点 | |
| 2012 | ClickHouse 在 Yandex.Metrica 投入生产 | 产品 | Yandex | 在公司成立前证明真实世界规模 | |
| 2016 | 按 Apache 2.0 开源发布 | 产品 | ClickHouse / Yandex | 开启外部开发者采用 | |
| Aug 2021 | ClickHouse, Inc. 注册成立并完成 Series A | 创立 | $50M | Aaron Katz、Alexey Milovidov、Yury Izrailevsky、Index、Benchmark 等创始人与投资方 | 创立独立、由风投支持的公司 |
| Oct 2021 | Series B 以 $2B 估值完成 | 融资 | $250M / $2B | Coatue、Altimeter、Benchmark、Lightspeed、Almaz 等 | 为云和全球 GTM 建设提供资金 |
| Mar 2022 | 公司发布 Ukraine 声明和迁址澄清 | 反向 | 迁址加速 | ClickHouse | 回应 Yandex / Russia 认知风险 |
| 2022 | Amsterdam 办公室开设,ClickHouse Cloud 进入早期访问 | 扩张 | 已上线 | ClickHouse | 建立欧洲枢纽并进入商业云阶段 |
| Mar 2025 | HyperDX 收购完成 | 产品 | 收购 | ClickHouse、HyperDX | 扩大可观测性版图 |
| May 2025 | OpenHouse 在 San Francisco 启动,Series C 宣布 | 融资 | $350M / ~$6.35B | Khosla Ventures 和广泛投资人联合体 | 重大估值重置和市场信号 |
| May 2025 | 宣布 $100M 信贷额度 | 融资 | $100M | Stifel、Goldman Sachs | 增加非稀释资本和债务复杂度 |
| Oct 2025 | 披露 Series C 延展轮和三位高级高管任命 | 治理 | 延展轮 + 高管任命 | Citi Ventures、Insight、Peak XV;Egan、Nagy、Sexton 等投资方与高管 | 加深管理层板凳并延长融资跑道 |
| Apr-May 2026 | 资料显示员工数为 531-569 人 | 扩张 | 500+ 名员工 | PitchBook、Tracxn | 确认后期阶段运营规模 |
时间线优先采用公司成立视角:2009 年和 2012 年指项目里程碑,2021 年标志法律意义上的创业公司成立。2025 年估值使用第三方报道,而非公司发布的定价。
[CO001, CO002, CO003, CO004, CO017, CO018]关键里程碑勾勒出 ClickHouse 从 Yandex 起源项目走向后期风险投资支持分析平台的路径。
02市场分析
2.1 市场边界、纳入支出与现状替代品
分析 ClickHouse 时,应把它放在分析型数据基础设施里,而不是泛泛的 数据库类别。ClickHouse 官方材料一直把产品定位为高速、列式 OLAP 数据库,用于实时分析、可观测性、数据仓库和 ML/GenAI 工作负载。 这个框架很重要,因为相关支出既不是“所有数据库”,也不是“所有商业 智能软件”。应纳入的预算池,是团队需要在大型事件、遥测或仓库数据集上 做高吞吐写入、低延迟 SQL 分析的基础设施层。 边界内有四个主要支出桶。第一是云数据仓库和 BI 加速,分析工程团队用 更快的分析存储替换或补足较慢的仓库层。第二是实时产品与事件分析,产品 或数据平台团队摄取流式事件,并服务仪表盘或面向用户的分析应用。第三是 可观测性和日志分析,SRE、平台和安全团队存储并查询日志、指标、trace 和高基数 OpenTelemetry 数据。第四是 AI 相邻分析基础设施,团队需要 向量感知检索、快速聚合,以及围绕 AI 系统的运营分析。 市场边界明确排除 OLTP 记录系统、作为独立类别的前端 BI 工具,以及本身 不能提供低延迟 SQL 服务的通用数据湖存储。现状替代品很强,而且分层明显: 对许多 Google Cloud 买家来说,BigQuery 是默认的无服务器云数据仓库; Datadog、Elastic 和 AWS OpenSearch 把日志、指标、trace、搜索,以及 越来越多的 AI 或向量工作流打包进各自托管平台;自托管开源基础设施仍是 偏重控制的团队用来避免托管服务锁定的替代方案。ClickHouse 的优势在于, 一个核心引擎可以横跨这些用例;但这套边界逻辑只有在本章把它视为重叠的 分析支出,而不是单一庞大的“数据库市场”时才成立。[CM001, CM002, CM003, CM004, CM017, CM032]
| 细分 / 品类 | 纳入支出 | 排除支出 | 买方 / 付款方 | 与 ClickHouse 的相关性 |
|---|---|---|---|---|
| 云数据仓库 / BI 加速 | 用于仪表盘和商业分析的托管及自托管分析存储与查询基础设施 | 前端 BI 许可证和通用 ETL-only 工具 | 分析工程或数据平台团队 / 集中式数据预算 | ClickHouse 数据仓库页面明确把产品定位为提升 BI 速度和并发的实时数仓 |
| 实时产品和事件分析 | 面向应用仪表盘和运营报表的事件摄取查询服务与低延迟 SQL 分析 | OLTP 记录系统,以及不带分析服务能力出售的流处理软件 | 产品工程或数据工程 / VP Engineering 或平台预算 | ClickHouse 官网和用例页面聚焦对连续摄取数据进行亚秒级分析查询 |
| 可观测性和日志分析 | 日志、指标、追踪、高基数 OpenTelemetry 数据留存与查询 | 不带分析存储的工单或事件管理 SaaS | SRE、平台工程或安全运营 / 基础设施或可观测性预算 | ClickStack 把 ClickHouse 直接放进 OTel 可观测性存储和查询层 |
| AI 和模型邻近分析 | 向量感知检索分析、智能体遥测,以及围绕 AI 系统的分析上下文 | 不带分析存储的核心模型训练或推理支出 | AI 平台团队 / ML 基础设施或创新预算 | ClickHouse 首页和相邻竞品页面显示,市场正转向 AI 驱动的分析和可观测性 |
| 自托管分析基础设施 | 客户在本地硬件、本地机房或 AWS、GCP、Azure 云 VM 上运营的集群 | 当不需要主权或完整运营控制时使用的托管 SaaS 控制平面 | 平台工程 / 基础设施预算 | 部署灵活性是核心差异化,因为 ClickHouse 能服务拒绝纯托管产品的买方 |
| 现状替代方案 | BigQuery、Datadog、Elastic、AWS OpenSearch,以及内部自管开源栈 | 不带分析数据平面的纯生产力软件品类 | 因工作负载和既有栈负责人而异 | 这些平台是 ClickHouse 最常替换或补足的对象,具体取决于用例 |
这里的边界是分析基础设施工作负载支出,而不是每一美元数据库支出。更适合理解 ClickHouse 为一套可复用引擎,跨数据仓库、事件分析和可观测性工作负载使用。
[CM001, CM002, CM003, CM013, CM032, CM034]2.2 TAM 下限、相邻市场视角与重叠纪律
ClickHouse 最干净的保守 TAM 下限,是云数据仓库类别。Mordor Intelligence 估算该市场 2026 年为 $14.94 billion,Research and Markets 估算为 $14.53 billion;两者都意味着,在加入任何可观测性或实时分析相邻支出前, 当前可服务类别已经超过 $10 billion。IndustryARC 对 2026 年的估算更乐观, 达到 $39.1 billion;这个数字在方向上有用,但应视为高端估计,因为它很可能 纳入了比 ClickHouse 直接捕获范围更广的 DWaaS 和数据存储工作负载。 更宽的上限视角来自流式分析。Grand View Research 估算该市场 2023 年为 $23.4 billion,2030 年达到 $128.4 billion,CAGR 为 28.3%,其中托管部署 已经贡献收入多数。ClickHouse 受益于同一条实时洞察需求转移,但这不是干净的 ClickHouse TAM,因为流式分析类别包含数据库层之上的流处理软件、服务和相邻工具。 可观测性提供第三个市场视角。Grand View Research、MarketsandMarkets 和 Mordor Intelligence 都把当前可观测性市场放在数十亿美元规模,云部署和大型 企业领跑采用。这与 ClickHouse 直接相关,因为 ClickStack 把 ClickHouse 定位为日志、指标、trace 和高基数 OpenTelemetry 数据的存储与查询层。重叠问题 很大:仓库、实时分析和可观测性预算不能直接相加,因为同一个买家可能用一个平台 承载多个工作负载。正确结论不是把它们求和,而是指出保守的当前 TAM 下限已经 超过 $10 billion,并且没有公开来源能干净隔离出更窄的“实时列式 OLAP 数据库” 细分市场;后者才更接近 ClickHouse 的真实 SAM。[CM019, CM020, CM021, CM022, CM023, CM024]
| 发布方 | 年份 | 地域 | 数值 | CAGR | 方法 | 置信度 | 局限 |
|---|---|---|---|---|---|---|---|
| Mordor Intelligence | 2026-2031 | 全球 | 2026 年 $14.94B -> 2031 年 $49.12B | 26.86% | 云数据仓库市场规模测算,厂商集合由 AWS、Google、Microsoft、Snowflake、Oracle 领衔 | 中 | 宽泛相邻品类;不是纯粹面向 ClickHouse 的 SAM |
| Research and Markets 报告 | 2026-2030 | 全球 | 2026 年 $14.53B -> 2030 年 $31.7B | 21.5% | 云数据仓库市场报告,明确列出 AI、计算存储分离和实时数据处理等趋势组合 | 中 | 仍是宽泛 DWaaS 品类,而不是仅限列式 OLAP 的一层 |
| MarketsandMarkets | 2026 | 全球 | 按应用、垂直行业、部署模式和类型拆分云数据仓库市场 | n/a | 市场分类佐证数据仓库需求会按客户分析、部署方式和组织规模拆分 | 低 | 免费抓取文本未清楚呈现核心市场规模数字 |
| IndustryARC | 2021-2026 | 全球 | 2026 年 $39.1B | 31.4% | 高端云数据仓库预测,强调 IoT、OLAP、MPP 和 DBaaS 需求 | 低 | 可能是本组里最宽的估计,或许高估了 ClickHouse 可触达支出 |
| Grand View Research | 2023 基准;2030 预测 | 全球 | 2023 年 $23.4B -> 2030 年 $128.4B | 28.3% | 流式分析市场,覆盖软件、服务、部署方式和终端应用拆分 | 中 | 可作有用的相邻上界,但不是干净的 ClickHouse TAM |
| Grand View Research | 2023 基准;2030 预测 | 全球 | 2023 年 $2.71B -> 2030 年 $5.40B | 10.7% | 可观测性工具与平台市场 | 中 | 窄于 ClickHouse 的完整覆盖范围,只以可观测性为中心 |
| MarketsandMarkets | 2023-2028 | 全球 | 2023 年 $2.4B -> 2028 年 $4.1B | 11.7% | 可观测性工具与平台市场,叙述框架包含衰退和远程访问 | 中 | 时间窗口更短,品类边界由厂商定义 |
| Mordor Intelligence | 2026-2031 | 全球 | 2026 年 $3.35B -> 2031 年 $6.93B | 15.62% | 可观测性市场,明确以 2026 年为基线 | 中 | 仅覆盖 ClickStack;不包含数据仓库或事件分析工作负载 |
表格有意采用多个视角。云数据仓库行给出保守的当前 TAM 下限;流式分析给出最宽泛的实时相邻市场;可观测性则刻画 ClickHouse 日益活跃的遥测专项切入点。
[CM019, CM020, CM021, CM022, CM023, CM024]分层查看市场:从宽口径流式分析 TAM,到已经与 ClickHouse 相关的保守云数据仓库底部。
[CM001, CM002, CM003, CM019, CM020, CM025]有来源支撑的高低区间,覆盖与 ClickHouse 相关的主要相邻品类。
云数据仓库行使用 Mordor Intelligence 和 Research and Markets 抓取到的两个明确 2026 年点估计。可观测性行用 Grand View Research 的 2024 年当前规模点估计和 Mordor 的 2026 年基准圈定区间。流式分析行围绕 Grand View Research 的 2023 年市场规模给出窄区间。最后一行保留 IndustryARC 更激进的 2026 年仓库估计,作为单独的高端口径,而不是混入基准区间。
[CM019, CM020, CM023, CM025, CM028, CM031]2.3 买方、用户、付款方与部署路径分层
ClickHouse 的买方地图按工作流拆分,而不是按行业口号拆分。第一类买方是数据 平台或产品分析团队,他们需要为仪表盘、面向用户的应用或运营报表提供实时事件 分析。这里的用户通常是数据工程师、后端工程师或分析工程师;预算负责人是 工程副总裁或数据平台负责人;采用触发点,是现有仓库的性能痛点,或 需要在持续摄取的事件流上服务交互式分析。 第二类买方是正在现代化仓库和客户分析工作负载的 BI 或分析工程组织。ClickHouse 官方仓库页面明确把产品定位为实时数据仓库,可用更低成本提升查询速度和并发。 在这一细分里,付款方往往是集中化的数据平台或 IT 预算,对比集合包括 BigQuery 和其他托管仓库平台。 第三类买方是处理日志、指标和 trace 规模的 SRE、可观测性或平台工程团队。 ClickStack 围绕 OpenTelemetry、亚秒级查询和高基数遥测的定位,直接指向这些 团队;Datadog、Elastic 和 AWS OpenSearch 的竞品页面也显示,同一买家已经在 采购统一的托管可观测性平台。对这些买家来说,付款方通常是基础设施、平台或安全 运营预算,关键触发点是当前日志留存和查询规模下的成本或性能痛点。 第四类、更早期的细分,是 AI/ML 平台团队,他们需要高速分析存储、向量感知检索, 以及围绕 AI 系统的可观测性。采用路径通常从自托管概念验证或既有开源使用开始, 当组织需要自动扩缩容、更简单的升级和更低运维负担时,再转向 ClickHouse Cloud 或托管 ClickStack。这条云 / 自托管双路径有战略意义,因为有些买家明确在优化 控制权和主权,另一些则在优化更快投产。[CM005, CM006, CM007, CM008, CM009, CM010]
| 细分市场 | 购买决策者 | 用户 | 付款方 | 工作流 | 预算负责人 | 采用触发因素 |
|---|---|---|---|---|---|---|
| 实时产品与事件分析 | 工程副总裁或数据平台负责人 | 数据工程师、后端工程师、分析工程师 | 产品或平台预算 | 摄取事件流,并提供亚秒级仪表盘或面向用户的分析 | 工程副总裁或平台负责人 | 数据仓库延迟受限,或需要对实时数据做交互式分析 |
| BI 和数据仓库现代化 | 分析工程负责人或数据总监 | 分析工程师、BI 工程师、数据架构师 | 中央数据平台或 IT 预算 | 替换或补足较慢的数据仓库层,提高并发并降低成本 | 首席数据官或数据平台负责人 | 加载转圈、查询延迟高,或数据仓库支出上升 |
| 可观测性与 OTel 数据 | SRE 负责人、平台工程经理或 SecOps 负责人 | SRE、可观测性工程师、平台工程师 | 基础设施、可观测性或安全运营预算 | 存储并查询日志、指标、追踪和高基数遥测数据 | 基础设施副总裁或 SRE 负责人 | 日志留存成本上升,或既有可观测性栈查询性能差 |
| AI 与模型相邻分析 | ML 平台负责人或 CTO | ML 工程师、数据工程师、平台工程师 | ML 基础设施或创新预算 | 为 AI 系统补上向量感知检索、分析上下文和遥测 | CTO 或 AI 平台负责人 | 需要把 AI 遥测、分析上下文和运营数据统一起来 |
| 重视控制权的自托管部署 | 平台架构师或重视合规的基础设施负责人 | 平台工程师、数据库工程师 | 基础设施预算 | 在本地或 AWS、GCP、Azure 上直接运行 ClickHouse,而不是使用只能全托管 SaaS 的平台 | 平台或基础设施负责人 | 主权、合规要求,或希望避开托管服务锁定 |
买方按工作流和运营需求拆分,而不是按行业口号拆分。一家公司内部可能有不止一个 ClickHouse 买方,只要数据仓库和可观测性预算落在不同组织。
[CM005, CM006, CM007, CM008, CM010, CM011]ClickHouse 用一套共享引擎服务多个分析基础设施买方。
[CM002, CM003, CM006, CM008, CM009, CM010]2.4 增长驱动、采用约束与时点含义
当前周期内,ClickHouse 最重要的需求向量有三条。第一是实时分析的结构性上升: Grand View Research 把流式分析增长归因于实时预测、数字化,以及大数据、IoT 和 AI 的更广泛采用。ClickHouse 官方实时分析页面与这一工作负载直接对应,强调 持续摄取、高查询并发,以及规模化交互式 SQL。第二是云数据仓库现代化。Research and Markets 把可扩展的存储 / 计算分离、实时数据处理、预测和运营分析列为云仓库 主要趋势;这些主题与 ClickHouse Cloud 自身的存算分离和自动扩缩容叙事匹配。 第三是可观测性数据增长。Grand View、Grafana、IBM 和 Elastic 都指向同一个 市场方向:云原生复杂度、OpenTelemetry 标准化和 AI 驱动的可观测性工作流,正在 提高高速、高效遥测存储的价值。 采用约束同样重要。现有厂商现在卖的是整合体验,而不只是原始存储引擎。BigQuery 把企业数据仓库、实时分析和 AI 结合在一起。Datadog 和 Elastic 把日志、指标、 trace 与 AI 辅助调查结合在一起。AWS OpenSearch 把搜索、可观测性、无服务器 部署和向量工作流结合在一起。这些产品制造了切换成本,因为买家往往比较的是完整 运营系统、定价模型和治理框架,而不只是基准测试数字。 成本纪律也有两面。ClickHouse 主打更低基础设施成本、更少副本开销和激进压缩; Grafana 和 IBM 则都认为,2026 年可观测性会从“全量采集”转向更高价值的遥测和 成本管理。也就是说,当买家想要更高效的存储 / 查询引擎时,ClickHouse 受益;但 如果团队选择减少数据采集量,或留在现有厂商的打包平台里,而不是再引入一层分析 层,ClickHouse 也可能输掉。就时点而言,近期最强顺风来自 AI 注入的分析、 OTel 原生可观测性,以及替换较慢的仓库或日志栈;主要约束是迁移工作量、现有平台 打包能力,以及需要跨多个预算负责人销售,而不是对应一条清晰的软件预算科目。[CM009, CM010, CM013, CM014, CM015, CM021]
| 驱动因素 / 约束 | 方向 | 时间 | 含义 | 尽调追问 |
|---|---|---|---|---|
| 实时预测、数字化、IoT 和 AI | 顺风因素 | 当前及未来多年 | 扩大时间敏感型分析工作负载规模,这类负载更偏好列式 OLAP 基础设施 | 衡量 ClickHouse 新工作负载中,事件分析与传统 BI 各占多少 |
| 云数据仓库现代化 | 顺风因素 | 当前 | 买方想要计算存储分离、实时处理和预测分析,但不想承受数据仓库延迟 | 追问 ClickHouse 作为替代品进入的频率,以及作为既有数据仓库旁边加速层进入的频率 |
| OpenTelemetry 优先的可观测性 | 顺风因素 | 当前,到 2026 年仍会增强 | 标准化遥测和高基数数据抬高了对高效日志、指标和追踪存储的需求 | 量化 ClickStack 或其他可观测性部署带来的收入和客户数 |
| AI 驱动的分析与 AI 可观测性 | 顺风因素 | 正在兴起,但已近在眼前 | AI 智能体、分析副驾和模型遥测创造了新的分析存储与检索需求 | 追问销售管线中有多少比例把 AI 或模型遥测列为主要购买触发因素 |
| 围绕统一可观测性的工具整合 | 混合顺风因素 | 当前 | ClickHouse 作为低成本核心引擎可以受益;但如果买方更愿意留在既有套件里,也可能输掉 | 索取整合项目对 ClickHouse 有利和不利场景的输赢分析 |
| 集成式既有平台 | 约束 | 当前且结构性 | BigQuery、Datadog、Elastic 和 AWS OpenSearch 把相邻工作流打包在一起,降低买方再加一个平台的意愿 | 按既有厂商和工作负载拿到胜率,识别 ClickHouse 实际替换打包替代方案的场景 |
| 迁移和运营变更成本 | 约束 | 当前 | 即便 ClickHouse 基准测试更好,团队仍要迁移 schema、查询、数据管道或遥测工作流 | 按细分市场询问部署时间中位数和专业服务负担 |
| 遥测成本约束与数据价值筛选 | 约束 | 当前 | 部分买方会减少数据量或留存,而不是更换存储引擎 | 索取留存分层使用模式和降本案例,证明数据价值优化之后 ClickHouse 仍能赢 |
顺风在工作负载增长和成本压力同时出现时最强。既有平台只要能打包足够多的相邻功能,切换就会显得有运营风险,约束随之上升。
[CM009, CM010, CM013, CM014, CM021, CM026]典型路径是从既有系统痛点出发,先做自主管理的概念验证,再进入托管或规模化生产部署。
[CM003, CM005, CM006, CM007, CM008, CM009]2.5 规模估算缺口、相互矛盾的估计与尽调问题
最大的分析缺口是类别重叠。云数据仓库、流式分析和可观测性都是观察 ClickHouse 的有效市场视角,但没有一个能完美代理 ClickHouse 实际变现的更窄层。把这些类别 相加会夸大 TAM,因为一个平台可以为同一客户服务多个工作负载;只用其中一个类别 又会低估公司的真实范围。因此,本章保留这些估计作为视角,而不是强行拼出一个 合成的 SAM 数字。 第二个缺口是 ClickHouse 自身缺少公开分层。可获得的抓取证据支持强社区规模、 广泛用例覆盖和跨垂直行业客户采用,但没有披露 ClickHouse 收入中多少来自可观测性 与仓库工作负载,多少客户选择云而非自托管,或 AWS、GCP 和 Azure 的组合如何随 地区和买家类型变化。没有这些披露,公开 SOM 估算更多是表演,而不是分析。 因此,最有用的尽调问题都在公司内部:按工作负载拆分的收入组合、按云厂商和细分 统计的云服务数量、按部署模型拆分的净留存、按客户队列统计的中位遥测或事件量, 以及对 BigQuery、Datadog、Elastic 和 AWS OpenSearch 的输赢数据。这些答案 才能把一个广泛且明显巨大的 TAM 故事,转换成更紧、更可投资的 SAM/SOM 图景。[CM017, CM022, CM030, CM041, CM042, CM043]
2.6 图表
03竞争格局
3.1 竞争版图与市场边界
ClickHouse 所处竞争场,比“数据仓库”这个标签暗示的更宽。公司 2021 年围绕一个已经在分析数据库里很有名的开源项目注册成立,如今商业化叙事横跨实时分析、数据仓库、可观测性和 AI 相关服务工作负载。这种广度拉来了两类直接对手。第一类是 Snowflake 和 Databricks 这类宽分析平台;当买家想要一个受治理的系统,统一承载工程、分析和 AI 时,它们可以拿下同一笔战略预算。第二类是 BigQuery、Redshift 和 Athena 这类超大规模云厂商既有服务;它们能解决足够多的同类任务,同时受益于既有云采购关系。替代品集合又不一样:DuckDB、Apache Druid、StarRocks 和 SingleStore 各自覆盖同一分析问题中更窄的切片,尤其是嵌入式分析、流式优先分析或低延迟服务。因此,实际现状不是一个替代方案,而是更宽套件、打包云服务,以及开源或自托管点解决方案的混合。[CP002, CP010, CP018, CP021, CP023, CP025]
| 竞争对手 | 类别 | 规模 / 融资 | 目标细分市场 | 部署 / 开源姿态 | 最适合的工作负载 | 定价 / 定位 |
|---|---|---|---|---|---|---|
| ClickHouse | 参照平台 | 2021 年成立的商业公司;2025 年 5 月完成 $350M Series C;2,000+ 客户 | 工程主导团队,建设面向用户的分析、可观测性和高速数据仓库工作负载 | 开源核心,自管服务器,并在主要云市场提供托管云 | 实时分析、可观测性、数据仓库、AI 相邻分析服务 | 按用量计费的云定价,计算和存储分开;公开定价理念比企业实际净价更清楚 |
| Snowflake | 直接既有厂商 | $9.77B RPO;790 个 Forbes Global 2000 客户;733 个 $1M+ 客户 | 企业分析、受治理数据共享、跨云 SQL 和 AI 买方 | 托管多云服务;自研平台,治理控制强 | SQL 分析、数据共享、受治理的 AI 数据平台 | 明确的计算、存储和数据传输定价,采用仓库 credit 和按秒计费 |
| Databricks | 直接宽平台对手 | 20,000+ 组织;70% 的 Fortune 500;1,200+ 合作伙伴 | 企业数据工程、lakehouse、治理和 AI 团队 | 商业 lakehouse 平台,姿态偏开放格式,而非开源核心 | 数据工程、lakehouse、治理、分析和 AI 工作流 | 公开标价和 SKU 组存在,但横向比较不如明确起步价那么简单 |
| BigQuery | 超大规模云既有厂商 | 背靠 Google Cloud 分发和免费层漏斗;母公司为上市公司,并披露投资者报告 | 以 GCP 为中心的分析、BI 和 AI 团队 | 托管 Google Cloud 服务;自研但高度无服务器化 | 无服务器数据仓库、AI 分析和 Google 原生数据应用 | 计算加存储定价,含免费层和 slot 预留 |
| Redshift / Athena | 超大规模云数据仓库替代品 | 背靠 AWS 分发和有年度报告规模的母公司 | 标准化使用 S3、SageMaker 和 zero-ETL 路径的 AWS 原生数据团队 | 托管 AWS 服务,而非开源产品 | S3 上的数据仓库、lakehouse、即席 SQL 和无服务器分析 | Redshift 和 Athena 都公布了具体入门价格,降低试点摩擦 |
| DuckDB | 嵌入式替代品 | 开源基金会项目;已审阅来源未显示大型企业现场销售动作 | 做本地、notebook、应用或嵌入式分析的开发者和分析师 | MIT 许可的嵌入式数据库,无服务器进程 | 本地分析、嵌入式分析、单节点分析处理 | 免费开源软件,而非公开企业 SaaS 价目表 |
| StarRocks | 相邻实时挑战者 | 已审阅公开材料显示其为较小独立厂商;主打企业级分析,但公开规模披露很少 | 希望在新鲜 lakehouse 或实时数据上跑低延迟 SQL 的团队 | 带开源色彩的分析数据库,有云业务野心,但公开商业细节有限 | 实时分析、lakehouse 查询、面向 AI 的 SQL 服务 | 公开商业定价不如主要超大规模云或 DBaaS 对手透明 |
| Apache Druid / Imply | 流式优先替代品 | Apache 项目加商业 Imply 发行版;Polaris 公开入门和标准层级 | 流式密集分析、广告技术、遥测和面向客户的实时仪表盘 | 开源 Druid 核心,Imply 提供商业云和企业封装 | 流式优先实时分析和高并发查询 | 开源核心,加上 $100/month 和 $600/month 起的 DBaaS 层级 |
| SingleStore | 相邻 HTAP 挑战者 | 私有分布式 SQL 厂商,有大型企业客户标识,但公开规模指标有限 | 把事务、分析和应用服务工作负载放在一起的团队 | 云 DBaaS 加自管部署,覆盖 VM、云主机、Docker 和 Kubernetes | 实时应用、HTAP 型工作负载和可用于 RAG 的运营分析 | 按用量消耗 credit 的定价,另有存储费用和承诺选项 |
规模和融资细节哪里有公开披露,就采用哪里;对于透明度有限的私有厂商,本行使用定性规模和销售路径信号,而不是杜撰收入数字。
[CP006, CP007, CP010, CP014, CP018, CP020]用序数分数比较 ClickHouse 与保留替代方案:一边是主权部署灵活性,另一边是产品广度与渠道力量。
坐标轴为分析师根据公开产品和部署材料综合出的序数评分;它们不是经审计的市场份额测量。
[CP003, CP004, CP014, CP018, CP021, CP025]3.2 厂商画像、工作负载适配与开源姿态
当工程主导的买家需要高并发分析服务、大数据集上的快速 SQL,或一个能支持可观测性和实时产品分析、又不强迫采购宽套件的引擎时,ClickHouse 最强。Snowflake 和 Databricks 是最接近的宽平台竞争者,但原因不同。Snowflake 拥有最成熟的公开规模,并以 AI Data Cloud 叙事强调治理和仓库经济性;Databricks 则拥有更宽的开放格式 lakehouse 叙事,以及大得多的合作伙伴和客户触达。BigQuery、Redshift 和 Athena 重要,不是因为它们在产品点对点上完全相同,而是因为它们让“足够好”的分析可以在 Google Cloud 或 AWS 内部采购。DuckDB 和 Druid 是更专门的替代品:DuckDB 适合嵌入式和本地分析,Druid 为流式较重的实时使用而生。StarRocks 和 SingleStore 位于中间地带,在低延迟分析服务上与 ClickHouse 重叠,同时更偏向 lakehouse 和 HTAP 式定位。开源故事在这里很重要:ClickHouse、DuckDB 和 Druid 明确保留项目级可信度;Databricks 更强调开放格式,大型现有厂商则强调托管服务。[CP001, CP011, CP014, CP015, CP017, CP018]
| 购买标准 | ClickHouse | Snowflake | Databricks | BigQuery | Redshift / Athena | DuckDB | StarRocks / Druid | SingleStore |
|---|---|---|---|---|---|---|---|---|
| 实时分析服务 | 强 | 中 | 中 | 中 | 中 | 弱 | 强 | 强 |
| 宽泛的受治理数据 + AI 套件 | 部分 | 强 | 强 | 强 | 部分 | 弱 | 弱 | 部分 |
| 开源核心或项目可信度 | 强 | 弱 | 弱 | 弱 | 弱 | 强 | 强 | 弱 |
| 自管或主权部署选择 | 强 | 弱 | 部分 | 弱 | 弱 | 强 | 强 | 强 |
| 超大规模云打包 / 采购力 | 中 | 中 | 中 | 强 | 强 | 弱 | 弱 | 弱 |
| 嵌入式或应用原生分析适配度 | 中 | 弱 | 弱 | 弱 | 弱 | 强 | 中 | 强 |
单元格是基于已审阅官方产品和文档页面综合出的序数判断;它们表示姿态和适配度,不代表每种工作负载上都经过审计的基准优势。
[CP001, CP002, CP011, CP015, CP019, CP021]可视化概括 ClickHouse 领先之处、套件领先之处,以及更窄替代方案仍可信的领域。
热力图标签是有证据支撑的序数判断,概括产品姿态,而不是一对一基准测试分数。
[CP001, CP011, CP015, CP019, CP021, CP025]3.3 定价、部署模型与销售能力
定价和部署模型,是 ClickHouse 与竞争场区分最清楚的地方。它的公开定价叙事强调存储和计算独立扩缩、自动扩缩容,以及 scale-to-zero 经济性;文档则把自托管和托管消费放在同一个底层引擎上。Snowflake 在计费机制上比 ClickHouse 更明确:计算仓库消耗 credits,规格公开,平台分离计算、存储和数据传输成本。Databricks 公开但较难做基准测试对比,它发布的是标价和 SKU 组,而不是一条简单可比的列表费率。BigQuery、Redshift、Athena、Imply 和 SingleStore 都比 ClickHouse 暴露更清晰的公开起点,这有助于买家建模试点和间歇性工作负载。部署模型也影响信任和监管:Snowflake、BigQuery 和 Athena 主要是托管服务选择;SingleStore 仍支持自托管部署;DuckDB 则天然是本地的。这种组合会驱动销售结果。超大规模云厂商自有产品可以借助广泛采购杠杆,Databricks 可以依靠大得多的装机基础和伙伴集合。因此,ClickHouse 最自然的胜场,是技术差异化足以抵消较小外勤销售动作的场景。[CP003, CP004, CP005, CP012, CP013, CP016]
| 平台 | 价格 / 单位 / 合同模式 | 包含能力 | 折扣 / 未知项 | 竞争含义 |
|---|---|---|---|---|
| ClickHouse | 按用量计费的云定价,计算和存储分开计费,支持自动扩缩和缩容至零 | 面向实时分析、数据仓库、可观测性和 AI 服务场景的托管云 | 公开定价思路清楚,但企业实际折扣区间未披露 | 工程买家若看重突发负载效率,ClickHouse 很有吸引力;CFO 想要一条公开标价时,卖点较弱 |
| Snowflake | 计算额度加存储和数据传输费用;仓库按秒计费,最低 60 秒 | 托管多云分析与 AI 数据平台,提供受治理的数据仓库 | 净价随版本、云厂商和谈判后的额度经济性变化 | 计费机制很容易做基准,拿 Snowflake 对标 ClickHouse 建模不难 |
| Databricks | 各云厂商的未折扣标价和 SKU 组 | 覆盖数据工程、治理、分析和 AI 的湖仓平台 | 公开标价存在,但实际比较取决于 SKU 组合和谈判条款 | 平台范围广是优势,但标价复杂,难以直接做苹果对苹果比较 |
| BigQuery | 无服务器计算加存储计费,提供免费层和槽位预留 | Google Cloud 内的数据仓库和 AI 平台服务 | 实际成本取决于扫描量、槽位承诺和周边 Google 服务 | 试点路径很顺;对以 GCP 为中心的团队,无服务器模式足够简洁 |
| Redshift | 预置型起价 $0.543/小时,无服务器起价 $1.50/小时,并有预留折扣 | AWS 原生数据仓库和湖仓服务 | 总成本仍取决于数据、并发和 AWS 资产环境 | 入门价格清楚,加上 AWS 采购杠杆,Redshift 是现实的既有替代方案 |
| Athena | 按处理数据量或使用计算量付费,无需管理基础设施 | 直接在 S3 和其他来源上做临时 SQL 与 Spark 分析 | 查询模式和数据布局保持克制时才高效 | 间歇性 AWS 原生分析很适合;作为持久在线、高并发服务层,差异化较弱 |
| DuckDB | 免费开源软件 | 嵌入式本地分析引擎 | 企业支持和托管控制平面经济性不是产品重点 | 本地分析的强替代品,但不是云平台的一一替代 |
| StarRocks | 已审来源中的公开商业定价有限 | 实时、湖仓和 AI 分析引擎 | 客户可能需要直接接触才能确认价格 | 透明度较低,即使技术定位有吸引力,也拖慢快速采购 |
| Imply Polaris | Starter 起价 $100/月,Standard 起价 $600/月,Custom 按询价 | 基于 Druid 技术脉络的实时分析 DBaaS | 高规模或定制环境仍需企业级沟通 | 对早期测试和较小负载透明度高 |
| SingleStore | 按用量额度加存储费用,并有承诺用量定价 | 面向实时交易加分析的云 DBaaS | TCO 取决于负载形态和所选版本 | 初始建模比 ClickHouse 或 Databricks 更清楚,但优化对象是略有不同的混合负载买家 |
本表比较公开包装和计费机制,而非谈判后的净价或完整工作负载 TCO。
[CP005, CP012, CP013, CP016, CP020, CP022]3.4 护城河耐久性、切换成本与竞争风险
ClickHouse 的护城河真实存在,但有条件。最强的耐久优势,是开源可信度、部署主权,以及围绕速度敏感型分析工作负载建立的产品口碑。这种组合在规模化厂商中并不常见:Snowflake 更托管、更套件驱动;Databricks 更宽、更以工作流为中心;超大规模云厂商服务受采购驱动多于社区驱动。但同一结构也限制了硬锁定。买家可以多栖,因为类别按工作负载分割,开放或自托管替代方案仍然可信。Snowflake 和 Databricks 从上方威胁 ClickHouse,靠的是向受治理数据和 AI 套件扩展。BigQuery、Redshift 和 Athena 从侧面威胁它,靠的是云账户控制和打包相邻功能。DuckDB、Druid、StarRocks 和 SingleStore 从下方或侧面威胁特定切片,提供嵌入式、流式、lakehouse 或 HTAP 替代方案。因此,耐久承销问题不是 ClickHouse 技术是否强,而是这项技术优势转化成赢单的速度,能否快过更宽打包产品和更窄替代品对溢价的侵蚀。[CP006, CP007, CP008, CP034, CP035, CP042]
| 护城河主张 | 威胁 | 严重程度 | 缓释措施 / 尽调问题 |
|---|---|---|---|
| 开源可信度和开发者采用降低采用摩擦 | DuckDB、Druid、StarRocks 等开放替代品同样面向工程师,也降低锁定效应 | 中 | 向管理层询问开源用户转为付费云和企业支持的转化率 |
| 灵活部署支撑数据主权和受监管买家的叙事 | 采购便利性盖过主权收益时,超大规模云服务仍会胜出 | 高 | 索取按自托管、主权和托管云部署划分的 ARR 结构 |
| 速度敏感的服务层和可观测性是 ClickHouse 有吸引力的切入点 | Snowflake、Databricks、Redshift 都在持续营销性能提升和 AI 周边分析广度 | 高 | 查看可观测性和面向用户分析评估中的工作负载级赢单 / 输单数据 |
| 公开定价思路显示突发负载经济性高 | AWS、BigQuery、Athena、Imply、SingleStore 等既有厂商往往发布更简单的公开起价 | 中 | 索取企业实际折扣区间和从试点到生产的成本曲线 |
| 资本和客户势能让 ClickHouse 在企业交易中具备可信度 | 商业触达仍落后于超大规模云厂商,也可能落后于 Snowflake 和 Databricks 的一线销售覆盖 | 高 | 检验伙伴来源销售管线、渠道杠杆和区域销售覆盖,相对直接对手的差距 |
| 工作负载专精避免沦为同质化套件 | 即便 ClickHouse 赢在速度,更广的受治理数据与 AI 套件仍能吃下更多预算 | 高 | 询问 ClickHouse 是赢下主平台交易,还是主要作为专用工作负载引擎落地 |
关键问题在于,ClickHouse 能否把技术和社区强项转化为相对于更广套件和更窄替代品的持续份额提升。
[CP035, CP039, CP042, CP043, CP044, CP045]序数评分卡聚焦最可能决定 ClickHouse 能否继续赢过更宽套件和采购驱动既有厂商的维度。
评分是分析师基于已审阅公开证据得出的序数判断;它们不是公司披露 KPI。
[CP005, CP034, CP035, CP043, CP044, CP046]3.5 图表
04财务情况
4.1 变现与定价架构
ClickHouse 的商业故事比财务披露更清楚。纵观云、定价和用例页面,公司把用户从免费 / 开源采用引向 ClickHouse Cloud;这是一项托管服务,围绕按用量付费、计算与存储分离扩缩,以及空闲资源自动降配来定位。2022 年的发布节奏有财务意义,因为它显示公司在迭代一种开发者能理解的变现模型:AWS beta 之后,2022 年 12 月 6 日 GA 版本把试用期延长到 30 天,推出较低支出的 Development Services,并改进计算计量。TechCrunch 和 m3ter 补充了一个有用的商业细节:管理层有意推进产品驱动动作,提供透明的消费计费,并在 GA 前把 beta 定价从读 / 写单元简化为存储加计算。公开页面还显示,变现不止一个 OLAP SKU。ClickHouse 现在把平台卖向实时分析、数据仓库、通过 ClickStack 实现的可观测性、AI 原生工作负载,以及通过 ClickPipes 实现的托管摄取。仍不透明的是实际组合。没有已审阅的公开来源披露多少收入来自自助云、专用企业部署、BYOC,或更新的可观测性和 AI 附加项。[CI013, CI014, CI015, CI016, CI017, CI018]
| 收入来源 | 机制 | 单位 / 定价逻辑 | 当前公开状态 | 收入质量视角 | 尽调问题 |
|---|---|---|---|---|---|
| ClickHouse Cloud 多租户服务 | 基于共享云基础设施,为分析、AI 和可观测性负载提供托管服务。 | 按计算加存储用量计费。 | 明确在售,并定位为商业化核心。 | 对存在性信心高;对按客户队列划分的实际结构信心低。 | 索取自助客户与企业客户的云 ARR 拆分。 |
| 专属云 / 隔离部署 | 面向大型或监管更重客户的更高控制力云环境。 | 定制企业合同,ACV 可能高于自助模式。 | 公开有描述,但未公开定价。 | 对战略重要性信心中等;对实际贡献信心低。 | 索取专属部署的 ACV 区间和赢单率。 |
| 自带云(BYOC) | ClickHouse 在客户环境内管理控制平面。 | 谈判式企业定价,可能是服务费加支持费经济性。 | Sacra 公开描述为已上线的部署模式。 | 对商业化路径信心中等;对体量信心低。 | 索取 BYOC 客户数和合同 ARR。 |
| 可观测性 / 云上 ClickStack | 基于 ClickHouse Cloud 的托管可观测性栈,覆盖日志、链路追踪、指标和回放。 | 云用量加留存 / 摄取经济性。 | 已公开营销,并与成本效率主张绑定。 | 对产品贴合度信心中等;对当前收入占比信心低。 | 索取可观测性专用客户数、ARR 和留存。 |
| 实时分析工作负载 | 在 ClickHouse Cloud 上承载面向用户的仪表盘、反欺诈、营销和运营分析。 | 消费随摄取量、并发和存储增长。 | 核心营销工作负载类别。 | 它推动付费用量的信心高,但具体行业结构仍是私有信息。 | 索取头部垂直行业和按工作负载家族划分的毛利率。 |
| 数据仓库与 BI | 面向 BI 和高并发负载的现代数据仓库定位。 | 按消费计价,并有企业扩张潜力。 | 核心营销工作负载类别。 | 对需求面信心高;对纯数据仓库收入结构信心低。 | 索取数据仓库 ARR、典型数据占用规模和竞争赢单 / 输单明细。 |
| 托管摄取 / ClickPipes 与集成 | 托管连接器和摄取工具,简化上线与扩张。 | 可能通过更高云用量、高级功能或附加率变现。 | 公开可见,但未单独定价。 | 对其推动扩张信心中等;对直接变现信心低。 | 索取 ClickPipes 和合作伙伴集成的附加率与增购数据。 |
本表部分列举截至 2026-05-27 公开可见的收入面;已审来源未披露实际收入结构或实际定价。
[CI013, CI020, CI021, CI022, CI030, CI031]| 商业化面 | 公开定价信号 | 实际披露内容 | 财务含义 | 具体尽调问题 |
|---|---|---|---|---|
| 开源核心 | 免费 | 核心数据库仍开源且免费使用。 | 带来采用杠杆,但没有单独披露的许可证收入线。 | 索取从社区到云的付费转化漏斗。 |
| 云试用 / 免费增值入口 | 30 天试用和 $300 额度 | 官方云页面仍提供免费试用额度,供用户开始使用。 | 支撑低摩擦 PLG 获客,但试用到付费转化未披露。 | 索取按细分市场和队列划分的试用转化率。 |
| 标准云消费 | 按用量计费的计算 + 存储 | 官方页面称计算和存储可分开扩缩,客户只按实际使用付费。 | 适合突发负载;实际成交价仍不透明。 | 索取折扣后每计算单元和每 TB 存储的实际价格。 |
| Development Services | 低月度花费入门层 | 2022 年 GA 说明推出了面向入门用户的 Development Services 方案。 | 改善开发者上手,但毛利和增购收益未披露。 | 索取当前套餐限制、单位经济性和升级转化。 |
| 专属 / BYOC 企业 | 谈判定价 | 公开来源称企业客户可选择专属集群或 BYOC。 | 可能提升 ACV 和留存,但已审来源均未披露合同结构。 | 索取企业平均 ACV、合同期限,以及相对自助模式的毛利差异。 |
| 集成 / ClickPipes / 可观测性附加项 | 已审来源未列出标价 | 公开页面营销集成、可观测性和托管摄取,但未拆出定价。 | 可能提高钱包份额和留存,但不会显示为单独收入线。 | 索取附加率、打包规则和交叉销售收入贡献。 |
公开证据确认按用量计费和免费试用驱动的定价机制,但未确认实际定价、折扣或混合客户经济性。
[CI016, CI017, CI018, CI019, CI022, CI024]公开证据指向云驱动模型:把免费 / 开源采用转化为按量云收入,并推动更广工作负载扩张。
仅为定性桥接。公开来源确认了模型组件,但未披露转化率、混合价格实现或每个阶段的云毛利。
[CI013, CI017, CI020, CI021, CI022, CI029]4.2 增长信号与公开估计区间
对一家私有基础设施公司来说,增长证据异常强,但它们以碎片形式出现,而不是经审计财务报表。公司关联的 Series C 材料称,ClickHouse 在此前一年增长超过 300%,截至 2025 年 5 月客户超过 2,000 家。后续 2025 年延展材料称,ARR 在此前一年多翻了四倍以上;TechCrunch 则在 2026 年 1 月报道,云 ARR 仍同比增长超过 250%。最干净的第三方数字估计来自 Sacra,其估算 2025 年年化收入约 $160 million,较 2024 年底 $45 million 的退出运行率增长 256%。合在一起,这些数据点支持一个谨慎的公开估计区间:2025 年 ARR 或年化收入约 $150 million 至 $200 million。这个区间与用户给出的指引一致,也避免假装一家私营公司已经发布经审计收入。销售含义偏正面但不完整。ClickHouse 似乎正在把开源采用、免费试用转化和 AI 重度账户里的企业扩张叠加复利;但公开来源仍未披露 CAC、回本周期、ACV 分层或净留存。[CI006, CI007, CI008, CI009, CI010, CI011]
| 指标 | 公开数值 / 区间 | 信心 | 为何重要 | 尽调问题 |
|---|---|---|---|---|
| 2025 年年化收入 / ARR | $150M-$200M 公开估计区间(中点约 $160M) | 中 | 为估值和规模讨论提供锚点,但不假装有经审计收入。 | 索取月度经常性收入桥和经审计 FY2025 收入。 |
| 2024 年末年化收入 | $45M(Sacra 估计) | 中 | 为评估 2025 年增长主张提供基准。 | 索取 2024 和 2025 年月度收入历史。 |
| 进入 2026 年的云 ARR 增长 | >250% 同比 | 中 | 显示云业务强劲扩张,并支撑融资叙事。 | 索取按季度和部署模式划分的云 ARR。 |
| C 轮前公司披露的增长 | 前一年 >300% | 中 | 重要的势能主张,但来自公司口径而非审计数据。 | 索取增长指标定义,并与 GAAP 收入调节。 |
| 客户数 | 到 2025 年 >2,000 家客户 | 中 | 有用的广度信号,但客户数几乎不能说明集中度。 | 索取按客户队列划分的 ARR 集中度和 ACV 分布。 |
| 毛利率代理指标 | 计算-存储分离和自动扩缩可能支撑有吸引力的增量毛利 | 中 | 解释投资者为何可能在公开财务披露有限的情况下仍承销该模式。 | 索取实际云毛利率、基础设施 COGS 和支持成本负担。 |
| 销售效率代理指标 | PLG 免费试用、100+ 早期云客户、开源社区漏斗 | 低-中 | 暗示获客效率高,但不能替代 CAC 或回本周期。 | 索取自助与企业客户各自的 CAC、销售爬坡和回本周期。 |
| 净留存 / 流失 | 未公开披露 | 低 | 判断收入质量和耐久性的关键。 | 索取按队列划分的 NRR、GRR、客户数流失和扩张贡献。 |
本表混合披露事实、明确估计和定性代理;没有已审公开来源提供经审计的 ClickHouse 单位经济性表。
[CI010, CI011, CI012, CI023, CI024, CI025]区间图比较公开 2025 年收入估计,以及已披露资本池和估值锚。
收入数字是估计,不是经审计结果。融资总额合并公司公告的股权轮次和公开披露的 $100M 信贷额度。
[CI004, CI005, CI006, CI011, CI012, CI039]4.3 资本充足性与不透明成本结构
资本可得性看起来很强,但流动性可见度很差。ClickHouse 2021 年 8 月从约 $50 million Series A 资本起步,2021 年 10 月又以 $2 billion 估值完成 $250 million Series B 扩大资本基础,随后在 2025 年 5 月完成由 Khosla Ventures 领投的 $350 million Series C。2025 年材料还披露,由 Stifel 和 Goldman Sachs 牵头的 $100 million 信贷额度;2026 年 1 月报道又把一轮新的 $400 million 融资与 $15 billion 估值联系起来。这些都是强信号,说明投资人仍相信托管云分析可以快速复利。但公开记录仍缺少承销核心:账上现金、月度烧钱速度、基础设施 COGS、毛利率、净留存、客户集中度,以及信贷额度是否已提取或受 covenant 约束的任何细节。上市数据平台同行继续提交最新年报和风险因素,让外部人士能对标利润率结构和竞争。ClickHouse 没有。因此,融资栈可见,但现金流桥和下行情形保护仍是私有信息。[CI001, CI002, CI003, CI004, CI005, CI006]
| 项目 | 公开数值 | 日期 | 为何现在重要 | 剩余缺口 |
|---|---|---|---|---|
| 早期融资基础 | 约 $50M A 轮 | Aug-2021 | 显示公司以相对克制的资本基础进入 B 轮。 | 没有从 A 轮到 B 轮的公开烧钱桥。 |
| B 轮 | $250M,估值 $2B | Oct-2021 | 确立公司为资本充足的独立商业实体。 | 没有公开追踪募集资金用途与实际支出。 |
| C 轮 | $350M,由 Khosla Ventures 领投 | May-2025 | AI 时代规模主张和产品扩张背后的重大再融资。 | 融资后没有公开现金余额或现金跑道披露。 |
| 信贷额度 | $100M,由 Stifel 和 Goldman Sachs 牵头 | May-2025 | 增加非股权流动性能力。 | 已提取金额、契约、利率和到期日未公开。 |
| C 轮延伸轮 | 来自 Citi Ventures、Insight Partners、Peak XV 等的追加资本 | Oct-2025 | 显示初始 C 轮后仍能继续获得融资。 | 已审来源未披露延伸轮的确切规模。 |
| 最新估值信号 | 据报道,Jan-2026 的 $400M 融资估值 $15B | Jan-2026 | 确认投资者胃口强,资产负债表可选项充足。 | 仍未披露当前现金、烧钱速度或稀释结构。 |
| 当前流动性 | 未公开披露 | 截至 2026-05-27 | 这是判断真实现金跑道的核心阻塞点。 | 需要管理层提供现金、预测烧钱速度和契约余量。 |
本表聚焦当前资本充足性信号,而非重复完整公司历史时间线;流动性和债务细节仍为私有信息。
[CI001, CI002, CI003, CI004, CI005, CI006]| 缺失指标 | 为何重要 | 当前公开状态 | 具体尽调路径 |
|---|---|---|---|
| 按产品 / 部署模式划分的收入结构 | 需要用来判断收入质量,以及耐久企业云的占比。 | 未披露。 | 索取按自助、专属、BYOC、可观测性和支持划分的云 ARR。 |
| 实际成交价和折扣 | 需要区分公开标价思路和实际变现。 | 未披露。 | 索取客户合同、折扣表和续约定价队列。 |
| 毛利率和基础设施 COGS | 需要检验计算-存储分离是否真能带来有吸引力的云单位经济性。 | 未披露。 | 索取毛利率桥、托管支出、支持成本和按部署类型划分的贡献毛利。 |
| NRR、GRR 和流失 | 需要检验云扩张和多工作负载整合的耐久性。 | 未披露。 | 索取按队列划分的留存、降级和扩张贡献。 |
| 现金、烧钱速度和现金跑道 | 需要判断融资依赖和下行保护。 | 未披露。 | 索取资产负债表、现金流量表、月度烧钱和基准 / 下行情景现金跑道。 |
| 债务条款和授信提取 | 需要检验契约风险和增量杠杆。 | 信贷额度存在,但条款未披露。 | 索取信贷协议、已提余额、定价网格、抵押品和契约余量。 |
| 客户集中度 / 头部客户敞口 | 需要评估 AI 带来的赢单是分散还是集中。 | 未披露。 | 索取前 10 大客户 ARR 占比、队列毛利率和续约时间表。 |
这些是阻止仅凭公开信息完成承销判断的主要障碍,尽管增长和融资信号异常强。
[CI024, CI025, CI026, CI027, CI035, CI036]纸面上成本结构方向有利,但公开记录仍未验证毛利和基础设施负担。
仅为定性。来源支持经济机制,但不支持实际毛利率或贡献利润率数值。
[CI023, CI024, CI025, CI037, CI038]融资信号很强,但每条路径最后都落到流动性信息缺口。
仅为信号流。公开融资和估值事件已知,但它们没有披露当前现金桥或债务使用情况。
[CI001, CI003, CI005, CI006, CI027, CI039]4.4 承销视角
因此,财务结论是:营收动能强,披露质量弱。ClickHouse 有可信证据支持云主导商业模式、免费增值 / 开源漏斗、多工作负载变现,以及作为私有数据平台少见的高速增长。2025 年和 2026 年融资轨迹也说明,投资人愿意在明显更高估值上继续承销这个故事。但公开数据仍不足以对收入质量做硬承销。自助与企业云之间的收入组合没有披露,毛利率或基础设施成本负担没有披露,净留存没有披露;尽管有股权和信贷组合,流动性桥也没有披露。正确解读不是 ClickHouse 缺少财务质量,而是外部人目前能更清楚看到增长情景,而非下行情景。任何投资备忘录都应把 2025 年 ARR 或年化收入约 $150 million 至 $200 million 的估计,以及利润率上行逻辑视为可信但仍需私有资料室证据支撑,之后才值得建立确信。[CI011, CI012, CI024, CI025, CI026, CI027]
4.5 图表
05产品与技术
5.1 产品范围与交付模型
与其说 ClickHouse 是单一 SKU,不如说它是一组围绕同一分析引擎锚定的产品家族。公开记录支持三个具体交付表面。第一,是开源列式 OLAP 数据库本身。第二,是 ClickHouse Cloud;它被明确定位为全托管服务,并可在主要云市场 / 云厂商上使用。第三,是围绕该引擎的托管工作流表面,最显眼的是用于摄取的 ClickPipes,以及 SQL console 和 clickhousectl 这类云运维表面。也就是说,客户决策重点不在于 ClickHouse 是“数据库还是平台”,而在于他们希望 ClickHouse 帮自己运营技术栈的哪一层。 从客户工作流看,产品最强的场景,是团队需要高速分析存储加运营便利,而不是开箱即用的完整语义 BI 栈。ClickHouse Cloud 的卖点是减轻运维:无服务器运维、自动扩缩容、备份、复制和云厂商选择。ClickPipes 降低团队必须维护的定制摄取管道量。Microsoft 连接器和 Azure 集成证据显示,工作流正在延伸到 BI 和事件管道;模块图也显示,与原始自托管部署相比,云控制平面会实质改变设置、扩缩容和集成工作。主要尽调细节在于,许多周边工作流仍依赖第三方或伙伴工具,而不是单一纵向整合的 ClickHouse 原生应用层。[CE001, CE002, CE003, CE004, CE005, CE006]
| 模块 / 资产 | 主要用户 | 状态 / 成熟度 | 差异化 | 尽调缺口 |
|---|---|---|---|---|
| 开源 ClickHouse 服务器 | 数据平台和基础设施团队 | 生产成熟的核心产品,OSS 采用广泛 | 采用 MergeTree 存储模型的向量化列式 OLAP 引擎 | 需要按客户验证从既有数据仓库或搜索栈迁移的投入 |
| ClickHouse Cloud 托管服务 | 希望托管分析基础设施的团队 | 已在主要云市场 / 云厂商 GA,2026 年仍在推进区域扩展 | 托管运维、自动扩缩、共享存储设计、并行副本、计算分离 | 需要更清晰的公开 SLO 和按工作负载划分的基准测试方法 |
| ClickPipes 托管摄取 | 把事件或 CDC 数据导入 ClickHouse 的平台团队 | 已上市,2026 年按区域和连接器继续扩展 | 云原生摄取,不必自建 ETL 或管理消费者集群 | 连接器深度和积压处理因数据源而异,需要用真实 schema 测试 |
| SQL 控制台和 clickhousectl | 运维、分析师和开发者 | 已上市 | 查询和管理云服务的原生操作界面 | 公开反馈仍要求更深入的查询计划和角色治理体验 |
| 官方客户端和连接器生态 | 应用开发者和分析工程师 | 2026 年维护活跃度仍有强信号 | Python、JavaScript、Docker、Power BI/Fabric、dbt 和 ODBC 路径已有文档 | 连接器覆盖面广,但支持责任分散在核心、合作伙伴和社区层级 |
本矩阵区分核心引擎、托管云、摄取平面、管理界面和生态包;它不是收入拆分。
[CE001, CE002, CE003, CE005, CE019, CE024]团队如何从源系统进入 ClickHouse 上的托管分析工作流。
该工作流抽象了多种部署选择。不是每个客户都会用到每个节点,但抓取材料明确证明了每个节点。
[CE019, CE020, CE021, CE022, CE025, CE026]5.2 引擎、存储与云架构
核心技术差异化仍来自引擎设计。ClickHouse 架构材料明确描述了带可选代码编译的向量化执行模型;更底层的架构页面则解释,数据按列块处理,大多数操作分派到数组而非标量值上。这与产品目标一致:它面向宽数据集上的大型分析扫描、聚合、过滤和 join,而不是逐行交易工作负载。存储层围绕 MergeTree 家族构建,插入会生成不可变 part,并在后台合并。这些 part 按主键排序,索引是稀疏的:它为 granule 记录 mark,而不是为每一行建索引,因此即便数据集很大,索引也小到足以常驻内存。 这套架构重要,因为它同时解释了性能上行和运营权衡。稀疏主索引与有序存储改善了裁剪和压缩,但同一文档指出,稀疏读取仍可能每个 block 拉取额外行,因此性能取决于 schema 设计和 key 选择。在云部署中,ClickHouse 又增加一层差异化:计算与存储分离、对象存储支撑的并行副本、Shared Catalog 协调,以及用于读写工作负载隔离的 compute-compute 分离。Shared database engine 文档清楚表明,SharedMergeTree 式无状态计算不只是托管 VM 外面包一层;它是一种不同的运营模型,面向本地磁盘不应持有持久状态的动态计算环境。[CE009, CE010, CE011, CE012, CE013, CE014]
| 层 / 组件 | 作用 | 依赖 | 风险 |
|---|---|---|---|
| 查询处理层 | 解析、规划并运行分析型 SQL,采用向量化执行,可选代码编译 | 核心引擎内部机制和列块处理模型 | 性能优势取决于工作负载匹配和 schema 设计,不能只看查询速度说法 |
| MergeTree 存储部件 | 存储不可变的有序部件,并在后台合并,服务高摄取分析表 | 主键设计、合并设置、后台资源 | 主键设计不佳或小部件过多,会侵蚀摄取 / 查询效率 |
| 稀疏主索引和数据颗粒 | 在支持数据跳读的同时压住索引内存占用 | 排序键选择和 index_granularity | 稀疏读取仍可能拉取额外行,因此剪枝效果受工作负载影响 |
| 共享目录 / 共享数据库引擎 | 协调无状态云计算与 SharedMergeTree 风格表,不再绑定本地磁盘所有权 | 中央目录状态和 Keeper 支撑的协调 | 控制平面或元数据协调现在比本地磁盘耐久性更关键 |
| 云计算池和对象存储支撑的并行副本 | 靠计算分离和复制后的对象存储访问扩展读写 | 云编排,加底层对象存储和云厂商原语 | 云厂商区域成熟度和预览功能会影响功能可用性 |
| 集成层和表引擎 | 将 ClickHouse 接到消息代理、数据库、湖格式和对象存储 | 连接器质量,加外部系统 API 和 schema | 真实客户体验取决于连接器归属和集成支持层级 |
本架构表结合了稳定的引擎设计与云运行模型细节。它是有证据支撑的产品架构图谱,不是完整源代码或控制平面图。
[CE009, CE010, CE011, CE012, CE013, CE014]引擎层和云运营层合在一起,解释 ClickHouse 对分析负载的产品差异化。
这张图是根据文档和产品页面综合出的产品架构图,不是完整源代码或基础设施拓扑图。
[CE009, CE010, CE011, CE012, CE013, CE014]5.3 集成、生态与工作流适配
ClickHouse 的工作流故事宽而务实,不是纯原生。集成索引明确区分核心、伙伴和社区集成,这是一个有用信号:ClickHouse 支持广泛生态,但并非每个连接器都按同一标准维护,也不一定有同一支持模型。在摄取上,最强公开证据是 Kafka engine 与 ClickPipes 的组合。Kafka engine 提供低层 table-engine 路径,带 consumer-group、安全和 materialized-view 控制;ClickPipes 则是面向 Kafka、S3、Postgres、MongoDB、GCS、MySQL 及其他来源的托管云路径。Cloudflare Logpush 指南尤其有价值,因为它展示了一个具体生产模式——用 S3 作为持久缓冲,并具备 exactly-once 和 replay 语义——而不是泛泛的集成营销。 对基础设施公司来说,开发者表面也异常强。GitHub 活动、包分发和容器采用都指向成熟开源生态:主仓库有大量 star / fork 以及频繁发布,Python 与 JavaScript 客户端持续维护,Docker 镜像 pull 量很高。独立来源进一步说明,产品不限于传统仪表盘;采用者和公司故事横跨可观测性、云平台、SEO、区块链和面向客户的分析。需要注意的不是生态,而是架构。HypeQuery 的分析认为,一旦部署规模化,团队往往会在 ClickHouse 之上搭建语义、转换或自助层,让分析师和业务用户不必直接理解高度优化的 schema。换句话说,ClickHouse 非常适合作为工作流的性能核心,但受治理自助服务的最后一公里,往往要靠相邻工具拼出来。[CE019, CE020, CE021, CE022, CE023, CE024]
| 用户任务 | 当前工作流 | ClickHouse 方案 | 可衡量收益 | 限制 |
|---|---|---|---|---|
| 摄取流式事件和日志 | 自建或运维消费者、落地区和重试机制 | Kafka 引擎或 ClickPipes 托管摄取,将数据导入 ClickHouse Cloud | 托管摄取减少自定义 ETL,并可借助缓冲对象存储路径加入重放、exactly-once 式语义 | 运行表现仍取决于源系统、schema 漂移和连接器成熟度 |
| 建模数据仓库转换 | 在数据库外运行 SQL 模型和 CI/CD | dbt-clickhouse 适配器,支持增量、MV、分布式和测试 | 让分析工程师围绕 ClickHouse 统一转换和部署流程 | 对部分分布式和超大型模型模式,适配器仍有限制 |
| 连接 BI 和语义工具 | 将数据导出或复制到独立 BI 存储 | 借助 ODBC/Power Query 直连 BI,并接入更广泛的核心 / 合作伙伴 / 社区集成 | 支持 DirectQuery/导入,让分析更贴近运营数据 | 部分云服务场景仍需要 ODBC 驱动和网关桥接 |
| 运维面向客户或可观测性分析 | 在大规模事件流上扩展仪表盘和查询 | ClickHouse 核心叠加云管理、缓存和扩展能力 | 独立证据和官方证据都指向大规模下的亚秒级或实时分析工作流 | 团队可能仍需自定义抽象层,才能支撑广泛自助服务 |
| 普及自助分析 | 分析师需要平台团队协助,在优化后的 schema 上查询 | 把 ClickHouse 放在语义层或转换层下面,作为性能核心 | 保住引擎性能,同时扩大内部对治理后指标的访问 | 这层抽象通常由生态构建,而不是 ClickHouse 原生掌握 |
本表收益是有来源支撑的工作流结果,不是经审计的 ROI 声称。最后一公里自助服务层往往位于 ClickHouse 原生产品之外。
[CE019, CE020, CE021, CE022, CE023, CE024]外部系统和支撑层塑造核心数据库引擎之外的完整 ClickHouse 工作流。
依赖图混合了原生依赖和生态依赖,因为评估完整工作流采用的买方会同时关心两者。
[CE003, CE010, CE024, CE027, CE046, CE047]5.4 信任、路线图与产品风险
在信任和运营成熟度上,公开表面对基础设施产品而言高于平均,但对硬核企业尽调仍不完整。ClickHouse 公开记录了一组有意义的控制项——SSO、MFA、RBAC、私有连接选项、IP 过滤、CMEK,以及包含 GDPR、HIPAA、ISO 27001、PCI DSS 和 SOC 2 的合规清单。Azure GA 公告还补充了围绕网络隔离、流量加密和多可用区复制的具体平台级说法。这些都是有用信号,说明托管服务面向生产敏感工作负载设计,而不只是实验。 剩余风险不是没有控制项,而是公开深度不均。TrustRadius 反馈仍指向角色粒度和身份提供商限制;更广泛的工作流证据也显示,自助部署可能需要额外的语义或转换层,而 ClickHouse 并未完全拥有该层。2026 年更新日志显示平台路线图活跃——自动扩缩容调整、支出控制、索引分片预览、AWS/GCP/Azure 区域工作和 BYOC 扩展——但预览功能还不等于成熟、广泛部署的默认能力。对买家来说,含义很直接:ClickHouse 看起来在高性能分析工作负载上技术很强,尤其是在云托管时;但企业尽调仍应追问基准测试方法、身份治理深度,以及团队为了安全地民主化访问,必须自建多少周边平台代码。[CE028, CE029, CE030, CE031, CE032, CE033]
| 控制项 / 质量信号 | 状态 | 范围 | 含义 | 缺口 |
|---|---|---|---|---|
| 合规基线 | 公开列示 | GDPR、HIPAA、ISO 27001、PCI DSS、SOC 2 及相关项 | 说明云服务面向企业,而不是业余托管服务 | 公开清单不等于买方所需控制集已有范围匹配的审计证据 |
| 身份与访问控制 | 公开列示 | SSO、MFA、RBAC、IP 过滤、CMEK、私有连接 | 说明 ClickHouse 理解买方对企业云控制的预期 | 关于 SCIM、角色粒度和 IdP 兼容性的公开细节仍偏少 |
| 网络与可用性控制 | Azure GA 中已有公开描述 | 网络隔离、流量加密、多 AZ 复制 | 支撑需要韧性和受保护流量路径的生产分析工作负载 | 需要按云厂商和套餐层级披露恢复目标细节 |
| 运维 UX 反馈 | 独立反馈好坏参半 | SQL 控制台、角色和 SSO 体验 | 独立评论证据有助于区分产品控制声明和操作者体验 | 评论证据偏个案,需要用参考客户验证 |
| 连接器依赖 | 已知要求 | 部分 Microsoft 云流程需要 ODBC 驱动和网关 | BI 连接今天已经可用,并非只是愿景 | 相比真正浏览器原生的 SaaS 连接器,额外网关 / 驱动步骤增加落地摩擦 |
本表只记录已抓取来源中可见的控制项和质量信号。不应把它当作面向客户的安全、隐私或架构审查包替代品。
[CE028, CE029, CE030, CE042, CE050]| 日期 / 阶段 | 功能 / 里程碑 | 状态 | 含义 | 来源 |
|---|---|---|---|---|
| 2026-04 | 双窗口垂直自动扩缩 | 滚动发布 / 已上线 | 云服务在调优变动工作负载下的成本响应,而不是固定使用单一回看窗口 | 2026 云服务变更日志 |
| 2026-04 | AWS、Azure 和 GCP 间的云市场订阅共享 | 已上线 | 证实其多云商业机制真实存在,而不只是营销层面的可用 | 2026 云服务变更日志 |
| 2026-04 | 索引分片 | 私有预览 | 说明团队在主动降低内存压力,并提升向量、全文搜索等重索引工作负载的性能 | 2026 云服务变更日志 |
| 2026-04 to 2026-05 | GCP 上的 BYOC,以及 AWS/Azure/GCP 区域扩展 | GA / 已上线 | 增强平台覆盖,服务数据驻留和云厂商选择需求 | 2026 云服务变更日志 |
| 2026-05 | 组织支出提醒和主服务闲置 | GA | 说明云 FinOps 和工作负载隔离能力在成熟 | 2026 云服务变更日志 |
| 2026-05-26 | clickhouse-connect 1.1.0 包发布 | 已发布 | 客户端生态在本次调研窗口内仍在发布有意义更新 | PyPI clickhouse-connect 包 |
这是公开路线图和发布可见性表。它反映运行日期前的公开信息,不是内部产品管理待办。
[CE005, CE031, CE032, CE033, CE036]基于公开材料审视核心引擎、托管云、摄取、生态和面向买方的信任信号成熟度。
这些成熟度标签是基于公开证据质量和截至运行日期的功能推出状态得出的分析判断。
[CE005, CE031, CE032, CE033, CE042, CE046]5.5 图表
06客户情况
6.1 客户基础与细分组合
ClickHouse 可见客户基础并不是泛 SMB 意义上的广,而是在少数技术买家群体里很深;这些群体都需要在超大数据集上做低延迟分析。最主要、反复出现的细分是可观测性和遥测:Cloudflare、OpenAI、Anthropic、Tesla、Qonto、Langfuse 和 Lyft 都用 ClickHouse 存储或查询日志、trace、指标或内部分析事件流,在这些场景中,速度和高基数探索比传统 BI 打磨更重要。第二个主要细分是面向客户的产品和使用分析,可见于 Microsoft Clarity、Replo、Mintlify、Padlet、Ramp、Buildkite 和 Polymarket;ClickHouse 位于仪表盘、报表、预算、实验或排行榜产品背后,外部客户会直接触达这些产品。共同买方通常是工程、平台或数据团队;最终用户则是更广的一组分析师、运营人员、财务团队、教师、开发者或 SaaS 客户,他们被有意隔离在原始 ClickHouse 复杂度之外。这种分层是真优势,因为它集中在 ClickHouse 性能优势可被直接感知的工作负载上;但也意味着,评估客户集中风险时,应像看客户品牌数一样看工作负载家族:公开证明在可观测性和实时分析上密度高得多,在传统企业仓库用例上则少得多。可获得的公开记录也按证据质量清晰分层。近期案例研究客户提供高置信度生产证明,带具名发言人和量化结果;eBay、Spotify、Uber 和 ByteDance 这类老牌大客户主要出现在采用者列表或幻灯片引用里,除非尽调拿到更新的部署细节,否则应视为较弱证据。 [CU001, CU002, CU003, CU004, CU035, CU036]
| 细分市场 | 买方 / 用户 / 付费方 | 主要用例 | 代表性规模信号 | 收入 / 战略价值 | 关键缺口 |
|---|---|---|---|---|---|
| 可观测性和遥测平台 | 买方 = 平台 / SRE / 数据工程;用户 = 内部工程师和运维;付费方 = 基础设施 / 平台预算负责人 | 日志、trace、指标、事件分析、计费分析 | Cloudflare 千万亿行级分析;OpenAI 每日 PB 级日志;Tesla 千万亿行级 Comet | ClickHouse 性能优势最清晰、最站得住脚的锚定工作负载 | 没有按可观测性客户群披露 ARR 构成或续约率 |
| 产品和客户分析 SaaS | 买方 = 产品 / 数据工程;用户 = 产品团队和最终客户;付费方 = 产品或分析预算 | 嵌入式仪表盘、行为分析、实验、归因、报表 | Microsoft Clarity 数百万项目;Replo 100B+ 事件;Padlet 4000 万月活用户;Buildkite 每月 12B 次测试 | 支撑面向客户的分析 SKU 和更高 ARPU 的企业报表界面 | 公开证据偏向成功迁移,可能遗漏失败试点 |
| AI 和 LLM 原生软件 | 买方 = 平台 / ML 基础设施;用户 = 模型、基础设施和可观测性团队;付费方 = 核心平台预算 | LLM 可观测性、智能体追踪、安全遥测、模型运维 | Anthropic Claude 4 可观测性;Langfuse 十亿级智能体追踪;Mintlify 智能体与人类文档分析 | 增长最快的证据组,且 2025–2026 年新鲜度强 | 尚不清楚使用会停留在部门级,还是扩展为大型平台合同 |
| 金融科技和市场平台运营 | 买方 = 工程 / 数据团队;用户 = 财务运营、市场平台运营、风控团队;付费方 = 产品 / 平台预算 | 支出分析、预测、预算、欺诈 / 风险、排行榜和市场统计 | Ramp 50,000+ 客户;Qonto 600,000+ SMB;Polymarket 数百 rps 排行榜 API | 证明可变现分析可以嵌入运营产品 | 没有公开合同期限、增购率或多产品渗透数据 |
| 开发者工具和 CI/CD 分析 | 买方 = 工程平台;用户 = 开发者、发布经理、QA 团队;付费方 = 工程生产力预算 | 测试分析、功能分析、发布可观测性 | Buildkite 70B 条记录和 25k eps 峰值显示其适配开发者工具分析 | 适合高基数、自助式内部分析,并具备先落地再扩张潜力 | 支出可能受开发者工具整合周期影响 |
| 工业、教育和企业运营 | 买方 = 运营 / 数据负责人;用户 = 业务运营、教师、分析师;付费方 = 业务线负责人 | 工厂智能、课堂参与度分析、运营 BI | Contentsquare 迁移后 13 个月留存;Padlet 课堂指标覆盖 246 个国家中的 242 个 | 说明 ClickHouse 不止服务纯开发者工具,也能进入嵌入式运营分析 | 公开案例未披露账户规模或续约深度 |
细分市场基于公开案例研究和采用者引用分组,而不是来自已披露客户名单。 类似 null 的缺口反映公开商业细节缺失,不表示该细分市场不重要。
[CU001, CU002, CU003, CU004, CU021, CU022]| 指标 | 数值 | 日期 / 时期 | 来源 | 置信度 | 含义 | 缺失分母 |
|---|---|---|---|---|---|---|
| Cloudflare 查询规模 | 96T 事件 / 小时;每日 1.61 千万亿事件,<2 秒完成 | 2025 年 meetup / 2026 年案例 | ClickHouse Cloudflare 案例 | 高 | 最强公开旗舰证据,证明真实生产规模 | 未披露 Cloudflare 支出、集群成本或合同金额 |
| Cloudflare 部署年限 | 2016 年末起投产;1,000+ 活跃副本;每秒插入数亿行 | 2016–2023 部署历史 | ClickHouse meetup 报告 | 高 | 显示生产使用周期异常持久 | 未揭示与 ClickHouse Inc. 的商业关系 |
| OpenAI 摄取增长 | PB/日,月增长 >20%,90 个分片 x2 副本 | 2025 | ClickHouse OpenAI 案例研究 | 高 | 证实在需求病毒式增长下也适配超大规模可观测性 | 未披露支出、留存或 SaaS 合同条款 |
| OpenAI 激增韧性 | 夜间日志量激增 50%;索引修复后 CPU 降低 40% | 2025 年 3 月 | ClickHouse OpenAI 案例研究 | 高 | 体现运行弹性和上游协作价值 | 单一账户事件,不是客户组合层面的性能数据 |
| Tesla 负载测试 | 连续 11 天 1B 行 / 秒;摄取 >1 千万亿行 | 2025 | ClickHouse Tesla 案例研究 | 高 | 证实极端遥测和 PromQL 兼容用例 | 未披露内部部署经济性 |
| Microsoft Clarity 覆盖规模 | 数百万项目;数百万亿事件;数百 PB | 2020–2026 当前架构描述 | Microsoft Clarity 工程博客 | 高 | 来自大型客户自有界面的强嵌入式分析证明 | 公开文章未说明 ClickHouse 商业模式 |
| Buildkite 使用增长 | 六个月内每月测试执行从 3B 增至 12B;年初至今 70B 条记录;峰值 25k eps | 2025 | ClickHouse Buildkite 案例研究 | 高 | 初始部署后采用范围明确扩大 | 未披露使用 Test Engine 分析的客户数 |
| Padlet 实时管道 | 每月 8B 事件;14 rps;中位数 45 ms;p99 690 ms | 2025-2026 | ClickHouse Padlet 案例研究 | 高 | 显示面向大众市场规模的嵌入式课堂分析 | 无变现或合同规模细节 |
| Qonto 可观测性压缩 | 231 TB 未压缩属性存入 376 GB(压缩率 99.84%) | 2026 | ClickHouse Qonto 案例研究 | 高 | 金融服务运营中成本 / 规模优势的有力证明 | 节省未直接转化为 ClickHouse 合同扩张 |
| Lyft 分析吞吐 | >450 TB 每日读取;~4 TB 每日写入;平均数百 qps,峰值数千 | 2025-2026 | ClickHouse Lyft 案例研究 | 高 | 证明大型企业内部采用广度 | 仅内部分析不自动等于大额外部收入 |
本表混合当前和历史证明点,因为 ClickHouse 不发布标准化客户 KPI 集。 缺失分母主要反映合同、席位或 ARR 上下文缺失,而不是技术规模弱。
[CU005, CU007, CU009, CU010, CU012, CU013]最强公开客户案例所呈现的典型买方路径:从发现问题,到生产部署,再到内部扩张。
[CU003, CU021, CU022, CU023, CU024, CU025]6.2 具名生产用户与部署模式
最强的生产证据从 Cloudflare 开始:它自 2016 年起就在生产环境使用 ClickHouse,如今 HTTP 和 DNS 分析、日志分析、Workers 运行时分析、内部分析、客户仪表盘、Firewall Analytics、Radar 和用量计费都靠它支撑。公开披露横跨 Cloudflare 早年自写的工程博客、ClickHouse 主办的 meetup 报告,以及 2026 年新发布的 ClickHouse 客户故事,几乎连续覆盖近十年的部署历史,这种连续性并不常见。Cloudflare 之外,ClickHouse 目前最强的参考案例多是其官网 2025-2026 年客户故事:OpenAI 谈到 90 个分片的 PB 级可观测性,以及因流量激增触发的优化工作;Anthropic 描述了用于上线 Claude 4 的定制隔离、类 Cloud 架构;Tesla 详述了万万亿行级 Comet 指标平台;Microsoft Clarity 说明 ClickHouse 如何在数百万亿事件规模下,把 30 分钟的热力图生成变成瞬时任务;Contentsquare 则展示了多租户 SaaS 分析从 Elasticsearch 迁移到 ClickHouse 的路径,基础设施成本下降 11 倍,p99 查询提升 10 倍。同样的模式也出现在 Replo、Mintlify、Padlet、Buildkite、Ramp、Qonto、Lyft、Polymarket 等更新的云原生 SaaS 案例中:团队通常先用 ClickHouse 解决一个产品界面里的延迟、基数或成本瓶颈,生产环境跑稳后,再扩展到邻近用例。这支撑了账户内先落地、再扩张的可信故事,但也意味着许多参考材料是厂商撰写的案例,而不是客户自己写的工程复盘;因此,结果性主张需要按支持来源的独立性和新鲜度打折。 [CU005, CU006, CU007, CU008, CU009, CU010]
| 客户 | 细分市场 | 部署 / 用例 | 生产 / 试点 | 结果 | 限制 |
|---|---|---|---|---|---|
| Cloudflare | 云基础设施 / 可观测性 | HTTP 与 DNS 分析、日志分析、Workers 运行时分析、Radar、计费分析 | 生产,长期运行(自 2016 年起) | 96T 事件 / 小时、每日 1.61 千万亿事件且 <2s;1,000+ 活跃副本;计费任务每日数百万次调用 | 同类最佳证据,但与 ClickHouse 的商业条款未披露 |
| Contentsquare | 数字体验分析 SaaS | 主 SaaS 分析从 Elasticsearch 迁移到 ClickHouse | 生产,按客户完成全量迁移 | 基础设施成本降低 11x,p99 改善 10x,留存 13 个月,迁移期间零回归 | 客座文章证据强,但仍发布在 ClickHouse 自有博客 |
| OpenAI | AI / LLM 可观测性 | 面向研究、ChatGPT 和企业 API 的 PB 级可观测性 | 生产 | 每日 PB 级日志,90 个分片 x2 副本;通过架构调整应对 50% 激增,并优化 40% CPU | 技术证明强,但没有合同或续约细节 |
| Anthropic | AI / LLM 可观测性 | 安全、隔离的可观测性栈,用于支撑 Claude 4 开发和监控 | 生产 | 在 Anthropic 安全环境内定制类 Cloud 部署;具名运维者称 ClickHouse 对 Claude 4 发布至关重要 | 结果具备战略意义,但数字不如 Cloudflare 或 Tesla 明确 |
| Tesla | 工业 / 车队可观测性 | 面向海量指标分析、兼容 PromQL 的 Comet 平台 | 生产环境 | 实时摄取数千万行 / 秒;测试连续 11 天达到 1B 行 / 秒;累计摄取超过 1 千万亿行 | 案例研究由供应商撰写,未披露支出或席位扩张 |
| Microsoft Clarity | 产品分析 | 免费分析、热图、仪表盘和可视化报告 | 生产环境 | 数百万个项目;数百万亿次事件;热图从约 30 分钟缩短到即时生成 | 客户自有工程证明很强,但未披露 ClickHouse 商业封装细节 |
| Replo | 商户分析 SaaS | 面向 Shopify 商户的产品内分析,跟踪会话、购买、A/B 测试和 AOV | 生产环境 | 4,000+ 商户、100B+ 事件、3,000-5,000 事件 / 秒;仪表盘在约 1 分钟延迟下保持响应 | 近期证明扎实,但需注意这是 Replo,不是请求核验的 Reprise 标志客户 |
| Buildkite / Ramp / Qonto / Lyft / Polymarket 客群 | 开发者工具 / 金融科技 / 出行运营 | 实时分析、预算、CI 仪表盘、可观测性、预测和排行榜 API | 生产环境 | 各自披露了具体使用量或节省指标,说明标杆客户之外也能重复落地并扩张 | 多数证据仍是供应商撰写的案例研究,而非客户自发博客 |
本表优先列出有可用生产细节的具名部署。eBay、Spotify、Uber 和 ByteDance 仍是公开采用者引用, 但本表不把它们列为高细节行;Discord 和 Reprise 未出现在已获取的佐证来源中,因此作为证据缺口跟踪, 而不是当作已验证部署处理。
[CU006, CU007, CU010, CU011, CU013, CU014]| 客户 / Logo | 主要公开证明 | 获取到的最新证据 | 证据质量 | 量化结果可见度 | 尽调含义 |
|---|---|---|---|---|---|
| Cloudflare | ClickHouse 案例研究及 Cloudflare 工程博客 | 2026 年 5 月 | 高 | 高 | 作为核心背书客户看待 |
| OpenAI | ClickHouse 客户故事,附具名工程发言人 | 2025-2026 年仍当前 | 高 | 高 | 对规模证明很强;仍需索取合同和扩张细节 |
| Anthropic | ClickHouse 客户故事,附具名技术负责人 | 2025 | 高 | 中 | 战略证明强,数字细节中等 |
| Tesla | ClickHouse 客户故事,附具名高级工程师和压力测试指标 | 2025 | 高 | 高 | 很适合作为极限规模可观测性的销售背书 |
| Microsoft Clarity | 客户自有工程博客 | 2026 | 高 | 高 | 证据不由 ClickHouse 托管,因此价值更高 |
| Contentsquare | 客座迁移博客及外部架构综述 | 2022-2026 | 中高 | 高 | 迁移 / 成本证明强,但仍有一部分托管在供应商渠道 |
| Uber | 外部架构摘要及采用者列表中的旧幻灯片引用 | 2020-2026 | 中 | 中 | 作为旗舰证明前,需要直接当前客户背书 |
| Spotify / eBay | 官方采用者列表条目及较旧幻灯片 / 网站引用 | 2018-2020 年历史证据 | 中低 | 低 | logo 信号好,当前生产细节弱 |
| ByteDance | 外部采用者摘要 / 社区清单级证明 | 2026 年外部清单 | 低 | 低 | 没有一手证据前,不要用该 logo 支撑承销判断 |
证据质量基于来源独立性、新鲜度、具名操作者细节和量化结果具体度。Discord 和 Reprise 未在已获取佐证中出现, 因此不进入打分行。
[CU004, CU011, CU013, CU014, CU020, CU021]ClickHouse 客户案例里可见的常见部署路径,展示产品通常如何从评估推进到更广泛的组织采用。
[CU011, CU012, CU021, CU022, CU023, CU024]公开客户证据按独立性、量化结果具体度、新鲜度和生产成熟度排序。
[CU004, CU005, CU006, CU011, CU013, CU014]6.3 留存代理指标、采购路径与扩张动作
ClickHouse 不公开披露客户 NRR、GRR、Logo 留存、续约率或头部账户扩张队列,因此耐久性只能从较弱的代理指标推断。公开可用的最佳代理指标,是产品评价和客户故事的形态。积极一面,PeerSpot 用户给 ClickHouse 8.6/10 的评分,并反复提到性能、压缩、可扩展性,以及自托管开源版仍然可用带来的较低厂商锁定。TrustRadius 评论者也把 ClickHouse 描述为主要的实时数仓,并称赞 MergeTree 性能和数据跳读。消极一面,同一组评论反复抱怨云端 RBAC 粒度、SSO 缺口、文档、UI 成熟度、部署复杂度和云成本管理。这些问题并不致命,但说明留存依赖技术能力足够强、能消化产品粗糙边缘的客户。公开客户故事也显示出一致的采购和扩张路径。团队往往从一个狭窄但高价值的工作负载切入,证明延迟或成本收益,再扩张到相邻产品界面。开源可试用降低了评估摩擦;当工作负载上升到多团队或企业级重要性后,ClickHouse Cloud 的存储与计算分离、自动扩缩容、缩至零和运维卸载又支持继续扩张。这个采购形态有利,但没有续约和扩张率数据,它仍是投资逻辑,不是已验证的客户耐久性指标。 [CU015, CU016, CU022, CU023, CU024, CU025]
| 指标 | 数值 / 状态 | 分群 | 置信度 | 尽调追问 |
|---|---|---|---|---|
| 净收入留存率(NRR) | 未公开披露 | 全组合 | null | 要求按云、自管支持和企业支持客群披露 NRR |
| 总收入留存率(GRR) | 未公开披露 | 全组合 | null | 要求按头部客户客群和工作负载族披露 GRR 与 logo 总留存 |
| 续约 / 合同期限 | 获取的公开材料未披露 | 企业客户 | 低 | 要求披露合同期限中位数、续约率和早期扩张节奏 |
| PeerSpot 产品评分 | 8.6 / 10 | 独立评论者 / 评估者 | 中 | 核验评分背后的评论数量趋势,以及企业版与开源用户构成 |
| TrustRadius 定性口碑 | 性能和数据跳读反馈正面;SQL 控制台、云 RBAC 和 SSO 缺口反馈负面 | 从业者 / 数据工程师 | 中 | 将评论主题与支持工单量和流失原因对照 |
| 供应商锁定代理指标 | 开源被反复提及,可降低锁定并简化试用采购 | 技术买家 | 中 | 衡量云端成交中有多少比例来自自管或 OSS 使用起点 |
| 支持 / 运营代理指标 | 云用户提到运营负担被卸下;评论者仍指出文档和部署复杂 | 云和自托管用户 | 中 | 要求披露支持 SLA 达成率、价值兑现时间和实施成功率 |
| 产品粗糙度代理指标 | 抱怨集中在 UI、安全、管理成熟度和成本监控,而非核心性能 | 企业评估者 | 中 | 要求披露企业客户主要异议、丢单原因和售后升级类别 |
这些是公开留存代理指标,不是真正的商业留存指标。评测网站口碑和运营反馈能指示买方满意度与摩擦, 但不能替代续约、收缩或扩张数据。
[CU031, CU032, CU033, CU034, CU038]6.4 集中度风险与证据质量评估
公开记录同时支持两个相反结论。第一,从表层看,ClickHouse 并没有危险地集中在任何单一具名客户或行业:抓取到的案例覆盖云基础设施、前沿 AI、开发者工具、金融科技、教育、数字银行、消费互联网和工业分析。这种广度降低了显而易见的单一垂直行业集中度。第二,证据基础明显集中在工作负载家族。可观测性、遥测、面向客户的分析和高基数运营报表,相比其他数据平台用例被严重高估。如果这组买方走弱,如果超大规模云厂商原生方案缩小性能差距,或新进入者把可观测性后端商品化,公开参考案例的密度就可能制造虚假的多元化感。不同客户名下的证据质量也差异很大。Cloudflare 是金标准参考,因为部署在多年里同时得到 ClickHouse 和 Cloudflare 工程渠道佐证。Contentsquare 也很强,因为它结合了客户署名的技术迁移文章和独立架构汇总。OpenAI、Anthropic、Tesla、Microsoft Clarity、Mintlify、Padlet、Buildkite、Ramp、Qonto、Lyft 和 Polymarket 是当前性很强的厂商撰写案例,带有具名操作者和指标,但主要仍出自 ClickHouse 控制的发布渠道。相比之下,在尽调核实当前范围、合同价值和部署状态前,eBay、Spotify、Uber 和 ByteDance 最好只视为中低置信度的 Logo 证据。Discord 和 Reprise 没有出现在抓取到的官方或公开佐证集合中;没有直接客户或合同证据时,不应把它们计入生产证据。剩余最大的盲点,是头部客户 ARR 占比、续约经济性,以及最显眼的公开 Logo 究竟会转化为持久商业集中度,还是只是门面型营销参考。 [CU004, CU014, CU020, CU035, CU036, CU037]
| 扩张驱动因素 / 风险 | 类型 | 影响 | 尽调路径 |
|---|---|---|---|
| 开源到云转化 | 扩张驱动因素 | 降低初始采购摩擦,并为之后进入托管云或企业支持创造变现路径 | 要求按分群和获客批次披露从 OSS 评估者到付费云转化的漏斗 |
| 账户内多工作负载扩张 | 扩张驱动因素 | 客户常从一个分析工作负载切入,再叠加可观测性、计费、发布或 AI 等相邻用例 | 要求披露账户级模块扩张时间线,以及按首个工作负载拆分的净扩张 |
| 托管服务卸下运维负担 | 扩张驱动因素 | ClickHouse Cloud 为小型平台团队减少补丁、分片和容量规划负担 | 要求披露企业交易中云相对自管的胜率差 |
| 工作负载族集中 | 集中度风险 | 公开证明高度集中在可观测性和实时客户分析,意味着 GTM 按用例存在集中度 | 要求按可观测性、产品分析、数仓和 AI/LLM 客群拆分 ARR 结构 |
| 标杆 logo 证据脆弱性 | 集中度风险 | eBay、Spotify、Uber 和 ByteDance 多数仍停留在清单级证明,因此市场感知的 logo 深度可能高于已验证的当前使用深度 | 获取直接客户背书、当前支出区间和近期使用证明,再评估标杆名称 |
| 头部客户收入不透明 | 集中度风险 | 已获取公开来源均未披露前 10 大客户收入占比、最大客户 ARR 或续约集中度 | 要求披露前 10 大 ARR 集中度、最大 logo 占比,以及过去 8 个季度的集中度趋势 |
| Discord / Reprise 证明缺口 | 集中度风险 | 请求核验的名称未在已获取官方 / 公开来源中得到佐证,给 logo 准确性带来尽调噪音 | 将这些 logo 纳入牛市情形前,要求管理层安排直接客户访谈或提供当前合同证据 |
扩张驱动因素由反复出现的迁移和第二工作负载案例支撑;集中度风险来自商业披露缺失, 以及不同 logo 客群之间证据深度不均。
[CU030, CU031, CU035, CU036, CU037, CU038]6.5 证据摘录
07风险
7.1 严重性排序与承销框架
ClickHouse 的风险画像并不寻常:公司同时拥有明显的产品动能,也还缺少足够扎实的承销记录。公开材料中最强的正面因素真实存在:公司在 2025 年 5 月以 $6.35 billion 估值融资 $350 million,称同比增长超过 300%,并披露客户超过 2,000 家,覆盖 AI、可观测性、实时分析和数仓工作负载。但估值论证更多由使用动能和品类定位支撑,而不是由已披露的经济性支撑。融资材料没有公布 ARR、收入、毛利率或盈利能力,因此从开源采用和免费试用转化,到具备公开市场质量的持久盈利,这条路径仍不透明。最重要的风险由此集中在五个主题:开源商品化和治理张力;Snowflake 与 Databricks 对大型云数据预算的竞争;StarRocks 在重 Join 工作负载、DuckDB 在本地开发者工作流上形成的新替代压力;云端正常运行时间、安全和补丁执行风险;以及客户质量风险,因为公开材料描述了规模,却没有展示集中度、留存或队列经济性。单看任何一项风险都不致命,但合在一起,它们把融资后的执行门槛抬得很高。[CR001, CR002, CR003, CR034, CR035, CR037]
| 信号 | 证据 | 反向原因 | 抵消因素 | 下一步监控 |
|---|---|---|---|---|
| 没有公开经济性支撑的估值 | 2025 年融资披露了客户增长,但没有披露 ARR、利润率或盈利能力。 | 公开单位经济性尚未可见,执行风险已经被资本化。 | 公司确实有产品动能和蓝筹客户。 | 收入质量披露和队列经济性 |
| 生态内部人士对开源核心模式的担忧 | Altinity 称,重要功能正变成云端专属,治理需要更清晰的隔离。 | 即使产品需求仍强,社区信任也可能被侵蚀。 | Apache 2.0 核心、大社区和多种部署模式仍降低了即时锁定效应。 | 路线图透明度、贡献者情绪,以及任何基金会式治理动作 |
| 独立基准测试对连接和并发的保留意见 | Exasol 报告分布式完整性较弱,且高并发下运行时间恶化 1.39x。 | 工作负载结构不匹配的买家,可能无法兑现亮眼基准测试叙事。 | ClickHouse 仍在许多以聚合为中心和性价比叙事中占优。 | 规范化、连接密集型和高并发用例的参考客户 |
| 企业版计划锁住合规功能 | HIPAA 和 PCI 是企业版功能,SLA 覆盖部分承诺消费合同。 | 变现可能比原始用户增长暗示的更依赖一组更窄的高端账户。 | 如果这些买家扩张,增购潜力是真实的。 | 按计划层级拆分的 ARR 结构和企业总留存率 |
| 可见云端可用率不错但并非顶尖 | 公开状态页显示,2026 年 2–5 月汇总可用率为 98.62%。 | 云产品在公开层面仍背负服务质量执行风险。 | AWS 专项可用率报告为 100%,且可用率披露透明。 | 按服务层级拆分的事故分布和客户级 SLO 达成情况 |
这些不是失败判断,而是目前公开信息中最有支撑的反向信号;如果管理层执行滑坡,它们可能扩大成打破投资逻辑的因素。
[CR009, CR020, CR025, CR034, CR038, CR041]剩余风险集中在变现质量、竞争和 OSS 治理,而不是某个未解的法律事件。
[CR015, CR020, CR023, CR028, CR035, CR037]7.2 竞争、替代与开源变现风险
最清晰的结构性风险,是 ClickHouse 必须同时打赢两场仗。面对 Snowflake 和 Databricks,它必须证明,一个扎根开源的引擎能拿下高端云分析预算,而不只是成为特定工作负载的性能补充或迁移目标。独立排名仍显示心智份额差距很大:Snowflake 和 Databricks 位居云数据栈前列,ClickHouse 虽然动能上升,排名仍更靠后。与此同时,市场低端正在碎片化。DuckDB 是本地、嵌入式和开发者优先分析场景中的可信替代,用户可以推迟甚至完全绕开托管服务。StarRocks 在运营分析里是更尖锐的直接威胁,因为已发布的基准材料持续把 ClickHouse 描述为在复杂多表和分布式 Join 场景下更弱。变现叠加层更关键:Altinity 的批评显示,ClickHouse 仅限 Cloud 的功能可以提升付费产品差异化,但也会加深社区不信任、增加分叉维护负担,并让 OSS 路线图显得从属于云端打包。这是可行策略,但执行风险随之上升,原因恰恰在于 ClickHouse 的开发者社区大到足以察觉并作出反应。[CR020, CR021, CR023, CR024, CR027, CR028]
| 威胁 | 重要性 | 公开证据 | 可能性 | 严重性 | 当前抵消因素 |
|---|---|---|---|---|---|
| Snowflake 和 Databricks 的预算引力 | 它们仍是大型企业预算中的标杆云数据平台,独立心智也强得多。 | DB-Engines 在 2026 年 5 月将 Snowflake 排第 6、Databricks 排第 7,而 ClickHouse 排第 26;TechCrunch 仍称 ClickHouse 是挑战者。 | 高 | 高 | ClickHouse 针对特定工作负载性能叙事很强,也有迁移故事 |
| StarRocks 在重 JOIN 场景的势头 | 基准测试讨论持续把 StarRocks 定位为宽表和多表场景更强,而这些场景对运营分析很关键。 | StarRocks 赞助的 Habr 基准测试声称 SSB 性能高 2.2x,并称 ClickHouse 未能完成其 TPC-H 测试集。 | 中高 | 高 | ClickHouse 对外强调 JOIN 支持改进,且仍赢下许多以聚合为中心的工作负载 |
| DuckDB 本地 / 嵌入式替代 | 开发者无需采用托管云服务,就能解决许多探索式和嵌入式分析问题。 | Exasol 基准测试称,在运营简单性重要的场景,DuckDB 仍有吸引力;DB-Engines 显示 DuckDB 心智仍在上升。 | 中 | 中高 | ClickHouse 用本地 CLI、clickhouse-local 和 chDB 反制,把用户留在其生态内 |
| OSS 足够好型竞争 | 如果自管 ClickHouse、fork 或托管替代品能解决足够多用例,云变现可能落后于广泛采用。 | Altinity 警告仅云端功能和开源核心漂移会改变社区激励;Tinybird 则主打托管 ClickHouse 和另一种开发者体验。 | 中 | 高 | 大社区和官方云工具仍给 ClickHouse 留下强升级路径 |
| 性能叙事在客户工作负载上反转 | 如果并发或分布式 JOIN 在真实买家工作负载下表现不佳,独立基准测试可被用来反驳 ClickHouse。 | Exasol 报告 TPC-H 风格测试中并发退化 1.39x、分布式完成度更弱。 | 中 | 中高 | ClickHouse 在基准测试中心和产品页面仍掌握强成本 / 性能叙事 |
基准测试证据混有独立材料和竞争对手撰写材料;方向性比精确倍数更可靠。
[CR023, CR024, CR025, CR026, CR027, CR028]竞争、治理张力和可靠性问题,都会传导到同一组估值支撑变量:增长质量、留存和云毛利。
边是定性方向,不加权;该图意在展示风险如何通过不同通道传导到估值支撑。
[CR020, CR023, CR025, CR027, CR036, CR037]7.3 云端执行、安全与法律合规风险
ClickHouse 在合规和云运维上做了不少实质工作,但公开记录仍支持一份严肃的执行风险清单。积极一面,公司记录了 SOC 2 Type II、ISO 27001、U.S. DPF、HIPAA、PCI、GDPR 和 CCPA 工作流,这对受监管买方应当重要。但同一批材料也显示,并非所有保护都是普遍可用:部分控制和认证属于企业版功能,SLA 仅限部分承诺消费合同,正常运行时间仍是一个需要持续观察的实时指标,而不是可以安静假设的事实。公开状态页显示,2026 年 2 月至 5 月的汇总正常运行时间只有 98.62%。对一家快速增长的基础设施公司而言,这个数字方向上尚可,但还不足以消除云可靠性尽调。安全态势带来第二重张力。ClickHouse Cloud 未受 2025 年标志性 RCE 影响,这是正面的运营信号;但 OSS 记录中仍有 RCE、查询缓存和崩溃类漏洞,要求客户和分叉维护者持续跟进补丁。对于一家卖进企业可观测性、数据仓库和 AI 应用的公司,可靠性和补丁纪律是产品价值的一部分,不是后台卫生。[CR008, CR009, CR010, CR011, CR012, CR013]
| 风险 | 司法辖区或暴露面 | 状态 / 证据 | 可能性 | 严重性 | 缓释措施 / 当前姿态 | 剩余暴露 |
|---|---|---|---|---|---|---|
| 隐私与跨境传输合规漂移 | 欧盟 / 英国 / 美国隐私制度 | ClickHouse 引用 U.S. DPF 以及内部 GDPR 和 CCPA 计划;隐私义务仍在演进。 | 中 | 高 | SOC 2、ISO 27001、DPF、隐私政策和企业合规流程已就位。 | 高 —— 受监管买家仍需证据证明这些控制按地区和层级有效运行 |
| 企业合规集中 | HIPAA / PCI 客户 | HIPAA 和 PCI 是企业版功能,意味着受监管高价客群。 | 中 | 中高 | 面向高价值买家的追加销售路径和合规控制已经存在。 | 中高 —— 受监管账户流失或放缓,会不成比例地伤害变现质量 |
| OSS 漏洞信任冲击 | 自管 ClickHouse 部署 | 2025 年 RCE 及更早的崩溃 / ACL 问题仍留在公开安全变更日志中。 | 中 | 高 | 云服务未受 CVE-2025-1385 影响,项目也发布修复和公告。 | 中高 —— 自管事件仍可能伤害整个平台的品牌信任 |
| 查询缓存授权缺陷 | 基于角色的访问和行策略 | GitHub 公告记录了:启用查询缓存且同一用户切换角色时,可能绕过 RBAC。 | 中低 | 高 | 变通办法是在多角色模式中避免查询缓存,或按角色拆分用户。 | 中 —— 问题已修补,但显示安全设计复杂度不低 |
| 开源核心治理反弹 | OSS 路线图 / 贡献者治理 | Altinity 称重要能力仅在云端提供,并要求更清晰的 OSS 治理。 | 中 | 中高 | Apache 2.0 核心、大社区和多种部署模式仍能降低即时锁定。 | 中高 —— 路线图张力可能消耗社区好感,并推高 fork 话术 |
| 安全事件后的 fork 维护负担 | 下游 fork 和私有变体 | CVE-2025-1385 公告要求 fork 维护者自行移植修复。 | 中 | 中 | 上游为受支持版本发布补丁。 | 中 —— 正经 fork 会继承补丁负担,可能撕裂生态信任 |
覆盖范围是局部的,仅限截至 2026-05-27 在公开合规、公告和治理材料中可见的风险。
[CR010, CR011, CR012, CR013, CR014, CR015]| 风险 | 机制 | 证据 | 可能性 | 严重性 | 缓释成熟度 | 剩余暴露 |
|---|---|---|---|---|---|---|
| 云正常运行时间不达标或事件历史嘈杂 | 可用性滑坡会削弱可观测性、AI 和数仓工作负载的信任;这些场景预期交互式性能。 | 公开状态页显示 2026 年 2-5 月聚合正常运行时间为 98.62%。 | 中 | 高 | 中等 —— 状态页和 SLA 流程已存在 | 中高 |
| 安全补丁成为产品交付的一部分 | 反复出现的 OSS CVE 倒逼严格升级;如果修复落后于客户集群,会拖累企业信任。 | 安全变更日志列出近期版本中的 RCE、缓存和崩溃类问题。 | 中 | 高 | 中等 —— 公告和修复公开 | 中高 |
| 仅云端功能变成支持和迁移负担 | 托管便利可能加快变现,但也会增加自管用户希望复刻的运维面。 | 云服务页面把 ClickPipes、编排、备份、扩缩容和补丁作为差异点。 | 中 | 中高 | 中等 | 中高 |
| 分布式 JOIN 可靠性仍取决于工作负载 | 如果买家带来更规范化或 JOIN 更重的工作负载,基准测试弱点可能变成云成本或性能事故。 | Exasol 基准测试称,TPC-H 风格 JOIN 中节点增加时查询完成率下降。 | 中 | 中高 | 低至中等 —— ClickHouse 对外强调 JOIN 改进,但独立批评仍在 | 中高 |
| 对超大云厂商的执行依赖 | ClickHouse Cloud 的分发和可靠性取决于 AWS、GCP 和 Azure 覆盖范围上的成功执行。 | 云服务页面称服务已上架三大云市场。 | 中 | 中 | 中等 | 中 |
| 规模化后运营复杂度迁移给客户 | 即便有自动化,承诺消费用户仍需要围绕 SLO、租户和支持质量的保证。 | SLA 措辞仅限部分承诺消费合同,而不是覆盖全漏斗。 | 中 | 中高 | 中等 | 中高 |
可能性和严重性为定性判断,直接对应公开正常运行时间、公告和云运营披露。
[CR008, CR009, CR014, CR015, CR022, CR025]7.4 客户集中度、GTM 质量与监控触发器
最后一类风险不是 ClickHouse 有没有需求,而是需求能否转化为持久、多元、可盈利的云收入。公开证据清楚支持一个开发者驱动漏斗:免费额度、本地模式、嵌入式选项、OSS 社区规模和容易上手的试验,都是为了最大化采用率而设计。这是产品分发优势,但如果相对少数企业客户贡献了过高比例的付费云支出,它也可能掩盖较弱的队列经济性。公开材料描述了 2,000 多家客户,并列出几个门面账户,却没有展示头部客户占比、NRR、毛利率,或云端相对自托管的转化。这个缺口很重要,因为 HIPAA、PCI 和 SLA 承诺等高端功能似乎主要落在企业买方身上。因此,投资人需要监控的不只是采用速度,还有组合质量:大型账户是否在扩张;随着更多关键工作负载落地,正常运行时间能否守住;仅限云端的功能能否在不疏远 OSS 基础的情况下完成转化;来自 Snowflake、Databricks、StarRocks 或 DuckDB 的竞争,是否压缩定价,或把 ClickHouse 推向比当前估值假设更窄的用例集合。[CR004, CR005, CR006, CR007, CR018, CR019]
| 风险 | 交易对手或依赖方 | 机制 | 证据 | 严重性 | 缓释措施 / 尽调要求 |
|---|---|---|---|---|---|
| 客户集中度披露不足 | 头部云客户及队列 | 披露客户数看起来很广,但少数大额承诺消费客户可能主导付费云业务经济性。 | 公开披露显示客户 >2,000,并提到承诺消费合同的 SLA,但没有 NRR 或头部账户结构。 | 高 | 索取按 ARR 口径的前 10 大客户贡献、NRR、总留存率和行业集中度 |
| 开发者驱动漏斗变现不均 | 自助与 OSS 用户 | 试用量和本地使用很高,但未必能顺畅转化为持久云端扩张。 | 30 天免费试用、$300 额度、OSS 核心、本地模式,以及社区优先叙事。 | 中高 | 索取各队列转化率、按获客渠道拆分的回本周期,以及自托管用户的云端附加率 |
| 依赖超大规模云厂商销售路径 | AWS、GCP、Azure 市场 | 云市场扩大分发,但也带来平台、利润率和运营依赖。 | 云产品页面称 ClickHouse Cloud 已上架三大云市场。 | 中 | 核查扣除云市场费用后的净收入,以及按云厂商拆分的集中度 |
| 高端合规收入集中 | 受监管企业买家 | HIPAA、PCI,以及可能更深的支持承诺都放在企业版计划里,抬高对大客户的依赖。 | 合规文档显示,HIPAA 和 PCI 绑定 Enterprise 计划,SLA 绑定承诺消费合同。 | 中高 | 索取按计划层级拆分的 ARR 结构、受监管垂直行业敞口,以及企业队列流失历史 |
| 知名客户光环夸大多元化 | Anthropic、Tesla、Meta、Sony、Instacart 等 | 知名客户能证明相关性,但也可能遮住狭窄的收入基础。 | 新闻稿列出大客户,但没有按账户、细分市场或地域披露收入拆分。 | 中高 | 索取最大客户占比、前 20 大客户占比,以及 AI 与非 AI 收入结构 |
这张风险清单有意聚焦变现质量,而不是原始客户标识数量或下载速度。
[CR003, CR004, CR006, CR008, CR012, CR019]| 风险 | 可监控触发项 | 阈值 / 事件 | 行动含义 |
|---|---|---|---|
| 2025 年估值下的盈利路径 | ARR、毛利率、烧钱速度或经营杠杆披露 | 到下一次重大融资或公开市场事件时,仍没有可信的单位经济性披露 | 将估值视为叙事驱动,并要求更大的安全边际 |
| 替代 Snowflake / Databricks | 相对既有云数据仓库的大型企业胜单 | ClickHouse 持续只作为旁路组件或狭窄工作负载引擎被采用,而不是主平台 | 下调终局份额假设和云扩张预期 |
| StarRocks 和 DuckDB 替代 | 客户工作负载结构与基准测试讨论 | 连接密集型工作负载转向 StarRocks,或嵌入式分析留在 DuckDB,未升级到云端 | 降低对广泛工作负载扩张和云端附加率假设的信心 |
| 云执行质量 | 公开可用率、事故频率和安全公告 | 状态可靠性进一步走弱,或补丁相关问题在可见客户环境中反复出现 | 提高执行折价,并要求更强的运营证据 |
| OSS 治理稳定性 | 路线图清晰度和社区语气 | 云端专属差异扩大,同时贡献者不满或分叉言论加速 | 纳入更高生态风险和更慢的社区驱动采用 |
| 客户集中度 | 头部账户结构与队列留存 | 少数企业账户贡献大部分 ARR,或试用转化后留存恶化 | 下调收入质量假设,并重新评估估值支撑 |
每个触发项都设计成可通过管理层披露或直接尽调来监控,而不是只靠产品叙事。
[CR001, CR023, CR024, CR034, CR035, CR037]ClickHouse 商业模式依赖一条升级路径:把开源和本地模式用户带进高级云层级,同时不让用户流向替代品。
该图展示商业依赖,不代表法律控制;替代路径既可能导流,也可能削弱云升级漏斗。
[CR006, CR017, CR018, CR022, CR030, CR036]7.5 证据摘录
08估值
8.1 建议:观察——产品和开源漏斗真实存在,但 Series C 已经把顶级执行力计入价格
ClickHouse 具备数据基础设施投资人想要的许多要素:强大的开源引擎,AI 和可观测性工作负载中的真实客户拉力,以及一个可随工作负载扩大而复合增长的用量计费云产品。问题在价格,不在质量。独立报道把 2025 年 5 月 Series C 估值放在约 $6.35 billion,而第三方对该业务 ARR 的估计集中在约 $150 million 到 $185 million,Sacra 的中枢估计为 $160 million。也就是说,这轮融资大约按尾随 ARR 的 35 倍到 42 倍定价。这个溢价显著高于公开市场上的 Snowflake 和 MongoDB;即便与 Databricks 相比也仍更贵,尽管 Databricks 收入基数大得多、平台宽度更广,并披露了留存情况。正确判断是以中等置信度观察,并标记高估值风险:保持跟踪,但在承销这个头部估值前,需要看到新证据,证明 ClickHouse 能迅速把开源需求转化为持久的云端经济性。[CV004, CV006, CV009, CV012, CV014, CV029]
| 决策字段 | 当前观点 | 决策含义 |
|---|---|---|
| 建议 | 观察 | 继续跟进,但仅凭公开证据,不应把 2025 年 5 月价格视为明显便宜。 |
| 信心 | 中 | 增长和采用有充分支撑;ARR 质量、留存和利润率尚未被公开验证。 |
| 风险评级 | 高 | 估值风险高,因为 Series C 大约按过去 ARR 的 35x-42x 定价。 |
| 估值立场 | 偏贵,但只有按前瞻视角才说得通 | 如果 ClickHouse 能快速把 ARR 复合增长到 $200M 中段至 $300M 区间,这个估值可以成立。 |
| 承保依据 | 前瞻 ARR,而非过去 ARR | 采用里程碑式承保,不要只依赖 2025 年名义估值。 |
| 上调触发项 | 证据到位后才买入 | 从观察上调到买入前,要求披露留存、利润率和云转化。 |
这张表反映的是投资判断,不是通用产品质量评分。核心争论在于:Series C 估值中已经计入了多少执行。
[CV009, CV012, CV014, CV029, CV043, CV044]| 论点 | 当前证据 | 什么会改变观点 |
|---|---|---|
| 开源采用创造低摩擦漏斗 | Sacra 提到 ~46K GitHub stars、开发者采用强劲,以及 250%+ 云 ARR 增长。 | 展示从开源到付费云的队列转化,以及大型企业账户内的持续扩张。 |
| AI 和实时分析是真实顺风 | ClickHouse 赢下需要低延迟分析查询的 AI 原生和可观测性客户。 | 持久预算归属和可复制的大客户部署证据会增强信心。 |
| 反方:公开可比公司差距很大 | 按过去口径看,ClickHouse 的倍数高于 Snowflake、MongoDB、SingleStore,甚至高于 Databricks。 | 只有 ARR 足够快地放大,让当前估值变成前瞻高十几倍或低二十几倍,这个差距才会收窄。 |
| 反方:平台宽度落后于 Databricks | Databricks 披露 20,000+ 家组织和 >140% 留存,ClickHouse 尚未公开给出同等数据。 | 更完整的企业功能集和更充分披露,才足以支撑相对 Databricks 更小的折价。 |
| 反方:披露偏薄 | ARR 为估计值;在已保留的公开来源中,毛利率、NRR 和融资条款仍未披露。 | 经审计财务数据或贷款人级别 KPI 披露,会实质性提高承保信心。 |
把投资逻辑放在分发和品类契合上时,它最强;把反向逻辑放在叙事质量与公开披露质量的差距上时,它最强。
[CV010, CV011, CV019, CV020, CV029, CV034]投资建议由一组张力驱动:一边是异常强的开源分发和云增长,另一边是与同行倍数的差距以及披露缺口。
该图表达承销逻辑,不是产品内部流程。
[CV011, CV012, CV014, CV029, CV037, CV043]8.2 Series C 定价背景:$6.35B 是前瞻承销价格,不是公开可比公司价格
2025 年 5 月这轮融资本身毫无疑问是真实的:ClickHouse 融资 $350 million,引入了广泛的一线投资人联合体,并叠加 $100 million 信贷额度。公开证据更难证明的是,公司是否已经凭当时规模赚到了中个位数十亿美元估值。ClickHouse 披露增长超过 300%、客户超过 2,000 家;Sacra 估计 2025 年年化收入约 $160 million,进入 2026 年时云 ARR 增速高于 250%。这些都是很优秀的经营信号。即便如此,隐含尾随倍数仍接近 40 倍;哪怕给予公司更宽的 $150 million 到 $185 million ARR 区间,市场要求的承销区间仍落在 30 倍中段到 40 倍低段。也就是说,投资人实质上是在今天为明天的 ARR 付款,而不是为一个成熟、由监管文件支撑的收入基数付款。[CV001, CV002, CV003, CV004, CV009, CV010]
按后视口径,ClickHouse 的 Series C 标记远高于上市公司和开源可比公司,也仍高于 Databricks 更富溢价的私募倍数。
倍数是用公开估值和收入数据简化桥接出的 EV 或市值 / 收入;私营公司企业价值可能不同于头条股权估值。
[CV013, CV014, CV017, CV019, CV024, CV027]8.3 可比公司组:Databricks 是愿景目标,Snowflake 和 MongoDB 是公开市场现实,SingleStore 是更接近的私营下限
可比公司组把估值张力说得很清楚。Snowflake FY2026 收入为 $4.68 billion,2026 年 5 月下旬交易倍数约为 13 倍收入。MongoDB 是最有用的公开开源基准,交易倍数更接近 10 倍。Databricks 仍享有约 24.8 倍收入的估值,但它建立在已披露的 $5.4 billion 收入运行率、平台上超过 20,000 家组织,以及高于 140% 的净留存之上。相比之下,SingleStore 披露 ARR 高于 $123 million,现金流接近盈亏平衡,但其最后一次已知披露估值只有约 $1 billion,按这个简化桥接大约是 8 倍 ARR。因此,ClickHouse 自己约 40 倍的尾随倍数高于所有具名同行。愿意支付这个溢价的唯一理性理由,是相信开源采用和 AI 时代工作负载增长能让 ClickHouse 快速迈向类似 Databricks 的前瞻规模,而不是落入类似 Snowflake、MongoDB 或 SingleStore 的定价区间。[CV015, CV017, CV018, CV019, CV020, CV021]
| 可比对象 | 2026 年规模指标 | 估值 / 倍数 | 相关性 | 局限 |
|---|---|---|---|---|
| ClickHouse Series C | ~$150M-$185M ARR 区间;2,000+ 客户 | $6.35B;~35x-42x ARR | 这正是待回答的承保问题。 | 收入区间是估计值,并非经审计。 |
| Databricks | $5.4B 收入运行率;>65% 增长 | $134B;~24.8x 收入 | 最好的高增长私有数据和 AI 平台可比对象。 | 产品面更宽,客户基础大得多。 |
| Snowflake | $4.68B FY2026 收入 | ~$61.55B 市值;~13.1x 收入 | 更纯粹的公开云数据仓库可比对象。 | 成熟上市公司,增长更低,治理约束不同。 |
| MongoDB | $2.46B FY2026 收入 | ~$24.74B 市值;~10.0x 收入 | 最好的公开开源溢价基准。 | 工作负载结构不同,在可观测性和分析上的直接重叠较少。 |
| SingleStore | >$123M ARR;现金流接近盈亏平衡 | ~$1B 最近已知估值;~8.1x ARR | 更接近的实时数据库可比对象,也是有用的私有市场底部参照。 | 估值时点更旧,流动性也低于实时融资定价。 |
这是本章明确选取的完整可比集合:ClickHouse 自身融资锚、一个规模化私有平台可比对象、两个公开可比对象,以及一个更接近实时数据库的私有可比对象。
[CV012, CV014, CV017, CV019, CV024, CV027]8.4 情景分析:只有 ClickHouse 能迅速从约 40 倍尾随 ARR 降到更低的前瞻倍数,这个估值才成立
关键估值问题不是 ClickHouse 好不好,而是公司能否以足够快的复合增长,让 $6.35 billion 的入场价事后看起来合理。按同行倍数倒推,所需规模跳升很大。按 Snowflake 的公开倍数,ClickHouse 需要约 $485 million ARR 或收入。按 MongoDB 的倍数,需要约 $632 million。即便给一个类似 Databricks 的溢价,也仍要求约 $256 million。因此,基准情景应放在前瞻 ARR 上,而不是尾随 ARR。若 ClickHouse 能把开源漏斗转化为企业云 ARR,并在未来 12 到 18 个月跨过约 $280 million 到 $320 million,当前估值开始显得可以辩护。若增长停滞、企业就绪度滞后,或公开可比公司进一步降估值,下行会很剧烈,因为当前倍数和公开交易区间之间有太多空气。[CV010, CV013, CV017, CV019, CV027, CV030]
| 情景 | 核心假设 | 估值逻辑 | 价值区间(十亿美元) | 概率信号 |
|---|---|---|---|---|
| 悲观 | 增长放缓并接近公开可比公司轨迹,企业功能缺口持续存在,公开数据基础设施倍数进一步走软。 | 更接近公开开源和数据仓库同业的十几倍出头可比框架。 | 3.0-4.0 | 如果 ClickHouse 证明留存和利润率质量之前 ARR 增长就降速,该情景最可能发生。 |
| 基准 | ClickHouse 将开源需求转化为企业云 ARR,并在 12-18 个月内达到大约 $280M-$320M ARR。 | 高十几倍至 ~20x 前瞻 ARR 的过桥估值。 | 5.0-6.5 | 需要持续超高增长,同时拿出 ARR 质量和企业就绪度的可信证据。 |
| 乐观 | ARR 很快达到大约 $375M-$430M,AI / 开源溢价守住,企业功能缺口收窄。 | 突围型基础设施赢家按 20x-22x 前瞻 ARR 定价。 | 7.5-9.5 | 需要顶级执行,以及市场对 AI 数据基础设施持续支付溢价的胃口。 |
区间是基于情景的估值桥,不是 DCF 输出。基准情景把 Series C 视为按前瞻承保的价格,而不是按当前规模定价。
[CV010, CV031, CV032, CV033, CV038, CV039]| 触发项 | 阈值或事件 | 对投资逻辑的传导 | 行动含义 |
|---|---|---|---|
| 增长证据失效 | 管理层数据不再支持 12-18 个月内通向 ~$300M ARR 的路径。 | 当前估值失去前瞻 ARR 支撑。 | 按公开开源和数据仓库倍数重新承保。 |
| 留存或利润率证据偏弱 | NRR 或毛利率披露显示,经济性明显弱于高溢价同业。 | 高溢价增长质量被高估。 | 不为 2025 年名义估值溢价买单;要求更低进入价,否则回避。 |
| 企业功能缺口持续存在 | 面向大型受监管工作负载,安全、治理或合规能力仍未成规模。 | 开源采用未能转化为大额、持久的云合同。 | 削减溢价,并延长产品与客户尽调。 |
| 公开可比公司进一步降估值 | Snowflake 或 MongoDB 式倍数大幅跌破当前 10x-13x 水平。 | 即使 ClickHouse 运营执行良好,退出测算也会被压缩。 | 下调目标进入价,并假设更保守的退出结果。 |
| 治理或股权结构意外 | 优先权、债务契约或其他高级索取权实质扭曲普通股等价价值。 | 名义估值夸大可投资经济性。 | 文件审阅前,将本轮视为不如名义估值显示得有吸引力。 |
这些触发项都可监控。每一项都会直接传导为倍数压缩、ARR 增长放慢或退出经济性恶化。
[CV031, CV032, CV033, CV039, CV044, CV045]当前标记接近基准情景顶部,需要很强的前瞻 ARR 增长,才不至于滑入悲观区间。
数值以百万美元计,代表基于情景的估值结果,不是贴现现金流估计。
[CV031, CV032, CV033, CV038, CV039, CV045]ClickHouse 在市场和产品因素上得分很高,但披露质量和当前进入估值明显拖后腿。
分数是 0-10 的承销判断,由公开证据综合而来,用于内部投委会讨论。
[CV009, CV011, CV029, CV037, CV043, CV044]8.5 退出就绪度与尽调要求:缺失指标恰恰是证明溢价所需的指标
证据缺口不在产品相关性,而在经济性证明。公开来源没有提供 ClickHouse 经审计财务、净留存、毛利率,也没有说明开源用户转化为付费云队列的质量。留存下来的公开来源同样没有披露清算优先权或股权结构细节,而这些都可能改变 Series C 头部价格的经济含义。这让这轮融资看起来更像一笔押注品类领导地位的信念交易,而不是一个由投资人级完整披露支撑的价格。因此,本章尽调姿态应继续聚焦那些能够改变建议的指标。要给出买入判断,需要看到经审计或管理层背书的 ARR 质量、留存、利润率结构和企业转化披露。没有这些,更强的论点是密切观察公司、保持价格纪律,并等业务证明自己能长进这个倍数后再承销,而不是只凭叙事走向这个倍数。[CV037, CV043, CV044, CV045]
| 主题 | 缺失证据 | 重要性 | 负责人或尽调路径 |
|---|---|---|---|
| ARR 质量 | 云端与自托管结构、客户队列和头部账户集中度。 | 检验开源漏斗是否高效转化为付费云 ARR。 | 向管理层索取 KPI 包和收入桥。 |
| 留存和利润率 | 按主要客户队列拆分的 NRR、毛利率、贡献利润率和回本周期。 | 判断 ClickHouse 是否配得上相对公开同业的持久溢价。 | 索取贷款人级别运营指标或审计支持。 |
| 股权结构表和条款 | 与 Series C 或后续融资相关的清算优先权、参与权和债务契约。 | 名义估值可能高估普通股等价价值。 | 审阅融资文件和瀑布模型。 |
| 企业就绪度 | 安全、治理、合规以及大型受监管客户胜单证据。 | 这是把开源热度转化为持久企业云支出的前提。 | 开展产品尽调和客户访谈。 |
| 退出路径 | 当前 IPO 与战略出售计划,以及对公开可比公司波动的敏感性。 | 风险投资回报取决于未来倍数底部,不能只看当前叙事。 | 向管理层和投行询问当前退出框架和情景敏感性。 |
缺失证据集中在经济性、治理和退出测算,而不是产品相关性。正因如此,建议仍停留在观察。
[CV037, CV043, CV044, CV045]8.6 证据摘录
免责声明
本报告是基于公开证据的尽调快照,不构成投资建议。重要财务、法律、技术和合同事实仍未公开;作出任何投资决策前,应直接向管理层和一手文件核验。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | Work on ClickHouse began inside Yandex in 2009 as an effort to run analytical queries on real-time, non-aggregated data. | 高 | SO001, SO021 |
| CO002 | ClickHouse entered production in 2012 to power Yandex.Metrica. | 高 | SO001, SO021 |
| CO003 | ClickHouse was released as open-source software under the Apache 2.0 license in 2016. | 高 | SO001, SO021 |
| CO004 | ClickHouse, Inc. was formed in August 2021 as a Delaware corporation separate from the earlier Yandex project. | 高 | SO021, SO022 |
| CO005 | Aaron Katz is a co-founder and the CEO of ClickHouse. | 中 | SO001, SO013 |
| CO006 | Alexey Milovidov is a co-founder, CTO, and the original creator of ClickHouse. | 高 | SO001, SO021 |
| CO007 | Yury Izrailevsky is a co-founder and president of ClickHouse. | 中 | SO001, SO020 |
| CO008 | ClickHouse is a fast, open-source, column-oriented database management system built for real-time analytics. | 高 | SO003, SO005 |
| CO009 | The company commercializes the open-source core primarily through ClickHouse Cloud and related managed real-time analytics services. | 中 | SO003, SO011, SO019 |
| CO010 | ClickHouse Cloud entered early access in 2022. | 中 | SO001, SO011 |
| CO011 | ClickHouse says it has employees in over 10 countries and operates with a distributed-team model. | 高 | SO001, SO002 |
| CO012 | ClickHouse opened European offices in Amsterdam in 2022. | 中 | SO001, SO023 |
| CO013 | PitchBook currently labels ClickHouse as headquartered in San Francisco, CA, confirming a San Francisco Bay Area headquarters identity. | 中 | SO024, SO001 |
| CO014 | Public location records differ on exact Bay Area labeling: Colorado registration lists Portola Valley as the principal address while Craft lists Palo Alto and Amsterdam office locations. | 中 | SO022, SO023, SO024 |
| CO015 | PitchBook lists ClickHouse with 531 total employees. | 中 | SO024 |
| CO016 | Tracxn reports ClickHouse had 569 employees as of April 2026, supporting a public 500-plus headcount range. | 中 | SO025 |
| CO017 | ClickHouse raised a $50 million Series A in August 2021 led by Index Ventures and Benchmark. | 高 | SO021, SO018 |
| CO018 | ClickHouse raised a $250 million Series B at a $2 billion valuation in October 2021. | 高 | SO004, SO018 |
| CO019 | Series B participants included Index Ventures, Benchmark, Lightspeed, Redpoint, Almaz Capital, Yandex, FirstMark, and Lead Edge alongside lead investors Coatue and Altimeter. | 高 | SO004, SO018 |
| CO020 | Lightspeed identifies ClickHouse as a 2021 Series B portfolio investment. | 中 | SO020, SO018 |
| CO021 | Mike Volpi currently serves on ClickHouse's board. | 中 | SO027 |
| CO022 | Peter Fenton currently serves on ClickHouse's board. | 中 | SO028 |
| CO023 | ClickHouse raised a $350 million Series C in May 2025 led by Khosla Ventures. | 高 | SO005, SO026 |
| CO024 | Series C participants included BOND, IVP, Battery Ventures, Bessemer Venture Partners, Benchmark, Coatue, Lightspeed, FirstMark, GIC, and Nebius. | 高 | SO005, SO006, SO018 |
| CO025 | ClickHouse also secured a $100 million credit facility led by Stifel and Goldman Sachs alongside the Series C. | 高 | SO005, SO008, SO026 |
| CO026 | Third-party coverage after the May 2025 round placed ClickHouse's valuation at roughly $6.35 billion to $6.4 billion. | 中 | SO026, SO012 |
| CO027 | Public financing releases put ClickHouse's total funding above $650 million after the May 2025 Series C close. | 高 | SO005, SO026 |
| CO028 | ClickHouse disclosed more than 2,000 customers by May 2025. | 中 | SO005, SO010 |
| CO029 | Series C communications said the company grew over 300% during the prior year. | 中 | SO005, SO009 |
| CO030 | Sacra estimated ClickHouse reached about $160 million in ARR by the end of 2025, growing 256% year over year. | 中 | SO011 |
| CO031 | PitchBook News reported ClickHouse hit $100 million in annualized revenue in the first half of 2025 and that recurring revenue had roughly doubled over the preceding six months. | 中 | SO019 |
| CO032 | Public company and media sources place Anthropic, Tesla, Cisco, Sony, and other large enterprises among the visible customer and use-case footprint around ClickHouse. | 中 | SO005, SO012 |
| CO033 | ClickHouse added Kevin Egan as CRO, Mariah Nagy as VP People, and Jimmy Sexton as CFO in 2025. | 中 | SO010 |
| CO034 | In March 2022 ClickHouse said it had no operations in Russia, no Russian investors, and no Russian members of its board. | 中 | SO002 |
| CO035 | ClickHouse said it accelerated relocation of the original Russian engineering team to Amsterdam after the invasion of Ukraine, and PitchBook later reported Nebius retained warrants but no equity stake. | 中 | SO002, SO019 |
| CO036 | JFrog disclosed seven RCE and DoS vulnerabilities in ClickHouse DBMS. | 高 | SO015, SO016 |
| CO037 | Ubuntu's July 2024 security notice said older ClickHouse builds had heap overflows and possible arbitrary code execution risks including CVE-2021-43305. | 高 | SO016, SO015 |
| CO038 | A 2024 bug, CVE-2024-22412, allowed query cache to bypass role-based access controls until patched. | 中 | SO017 |
| CO039 | Public sources show open-source adoption scaling from more than 20,000 GitHub stars in 2021 to about 46,000 stars by early 2026. | 中 | SO004, SO011 |
| CO040 | Major scale milestones included the 2022 Amsterdam-office and cloud launch phase, the March 2025 HyperDX acquisition, and the May 2025 OpenHouse conference plus Series C. | 中 | SO001, SO012, SO025 |
| CO041 | Public directories sometimes date ClickHouse to 2009 or 2012 because they track the project origin, but official corporate chronology distinguishes the 2021 company formation from the earlier Yandex project. | 中 | SO001, SO022, SO024, SO025 |
| CO042 | Index says Mike Volpi and Peter Fenton joined Aaron Katz in negotiating a majority-controlled independent spin-out with Alexey Milovidov's team relocating to Amsterdam. | 中 | SO021 |
| CO043 | The investor map spans founding investors Index and Benchmark, 2021 growth investors such as Coatue, Altimeter, Lightspeed, and Almaz, 2025 growth investors led by Khosla, and debt providers Stifel and Goldman Sachs. | 中 | SO004, SO005, SO018, SO020 |
| CM001 | ClickHouse describes itself as a fast open-source column-oriented database management system for real-time analytical reports using SQL queries. | 中 | SM001 |
| CM002 | DB-Engines describes ClickHouse as a high-performance column-oriented SQL DBMS for OLAP that is available as both open-source software and a cloud offering. | 中 | SM006 |
| CM003 | ClickHouse's official product positioning spans real-time analytics, observability, data warehousing, and ML or GenAI workloads rather than one narrow software category. | 高 | SM001, SM003 |
| CM004 | The public GitHub repository describes ClickHouse as a real-time analytics database management system, reinforcing the project's developer-facing identity. | 中 | SM005 |
| CM005 | ClickHouse documentation lists five deployment modes: ClickHouse Server, ClickHouse Cloud, ClickHouse CLI, clickhouse-local, and chDB. | 中 | SM002 |
| CM006 | ClickHouse Server can be deployed locally, on-premises, or on major cloud providers including AWS, GCP, and Azure. | 中 | SM002 |
| CM007 | ClickHouse Cloud is the fully managed ClickHouse deployment mode that removes operational tasks such as updates, backups, scaling, and security patching. | 高 | SM002, SM003 |
| CM008 | ClickHouse Cloud is available on all three major cloud marketplaces, giving buyers a managed service option across AWS, GCP, and Azure. | 高 | SM002, SM003 |
| CM009 | ClickHouse Cloud markets compute-storage separation, pay-for-use compute, and lower replica overhead as core cost-efficiency features. | 中 | SM003 |
| CM010 | ClickHouse pricing emphasizes automatic scaling up and down, scaling unused resources to zero, and separate storage and compute billing. | 中 | SM009 |
| CM011 | ClickHouse's real-time analytics page emphasizes best-in-class query performance, continuous ingest, high query concurrency, and integration with common analytics tools. | 中 | SM010 |
| CM012 | ClickHouse's data warehousing page positions the product as a real-time data warehouse for business intelligence with faster queries at a fraction of the cost of traditional warehouse stacks. | 中 | SM011 |
| CM013 | ClickStack positions ClickHouse as an OpenTelemetry-native observability stack for logs, metrics, traces, session replays, and errors. | 中 | SM004 |
| CM014 | ClickStack claims 10-100x cost savings and sub-second queries on high-cardinality telemetry, directly targeting observability storage-cost pain. | 中 | SM004 |
| CM015 | ClickStack also offers a managed deployment path on ClickHouse Cloud for buyers who want observability without self-managing infrastructure. | 中 | SM004, SM003 |
| CM016 | ClickHouse's community page reports 12k+ Slack members, 2.9k+ contributors, 29k+ pull requests, 796 releases, and 47.6k+ GitHub stars. | 高 | SM001, SM007 |
| CM017 | ClickHouse's official adopters page documents a broad set of companies using ClickHouse and publishing success stories, supporting cross-vertical market applicability. | 中 | SM008 |
| CM018 | Altinity describes itself as the second-largest contributor to ClickHouse and highlights ecosystem tools such as the Kubernetes operator, clickhouse-backup, and a Grafana plugin, indicating non-vendor ecosystem depth around the project. | 中 | SM027 |
| CM019 | Mordor Intelligence estimates the cloud data warehouse market at USD 14.94 billion in 2026 and USD 49.12 billion by 2031 at a 26.86% CAGR. | 中 | SM012 |
| CM020 | Research and Markets values the cloud data warehouse market at USD 14.53 billion in 2026 and USD 31.7 billion by 2030 at a 21.5% CAGR. | 中 | SM013 |
| CM021 | Research and Markets identifies artificial intelligence, compute-storage separation, real-time data processing, and predictive or operational analytics as major cloud data warehouse trends. | 中 | SM013 |
| CM022 | MarketsandMarkets frames the cloud data warehouse market by application, vertical, deployment model, and type, corroborating that buyer budgets are segmented rather than monolithic. | 低 | SM014 |
| CM023 | IndustryARC projects the cloud data warehouse market to reach $39.1 billion by 2026 at a 31.4% CAGR, a much more aggressive estimate than Mordor or Research and Markets. | 低 | SM015 |
| CM024 | IndustryARC says large enterprises are the largest current cloud data warehouse buyers, IT and telecom is the highest-growth application segment, and North America holds a 41.5% share. | 中 | SM015 |
| CM025 | Grand View Research sizes the streaming analytics market at USD 23.4 billion in 2023 and USD 128.4 billion by 2030 at a 28.3% CAGR. | 中 | SM016 |
| CM026 | Grand View Research attributes streaming analytics growth to real-time forecasting, digitalization, and the spread of big data, IoT, and AI. | 中 | SM016 |
| CM027 | Grand View Research reports that hosted deployment held 51.7% of streaming analytics revenue in 2023, BSFI held 23.8%, fraud detection led applications at 18.8%, North America held 38.0% share, and Asia Pacific was the fastest-growing region at 32.0% CAGR. | 中 | SM016 |
| CM028 | Grand View Research estimates the observability tools and platforms market at USD 2.71 billion in 2023 and USD 5.40 billion by 2030 at a 10.7% CAGR. | 中 | SM017 |
| CM029 | Grand View Research says cloud deployment and large enterprises were the largest current observability segments, and that microservices, containers, and cloud-native complexity are major demand drivers. | 中 | SM017 |
| CM030 | MarketsandMarkets estimates the observability tools and platforms market at USD 2.4 billion in 2023 and USD 4.1 billion by 2028 at an 11.7% CAGR. | 中 | SM018 |
| CM031 | Mordor Intelligence estimates the observability market at USD 3.35 billion in 2026 and USD 6.93 billion by 2031 at a 15.62% CAGR. | 中 | SM019 |
| CM032 | BigQuery positions itself as a fully managed and completely serverless enterprise data warehouse with real-time analytics, built-in AI, vector and hybrid search, and decoupled storage and compute. | 中 | SM020 |
| CM033 | BigQuery pricing includes on-demand query pricing starting at USD 6.25 per TiB scanned and slot-based editions, showing that incumbent warehouse alternatives are sold on explicit usage economics rather than only seat-based pricing. | 中 | SM020 |
| CM034 | Datadog markets a unified observability platform that aggregates logs, metrics, traces, and real-time dashboards across modern infrastructure. | 中 | SM021 |
| CM035 | Datadog pricing separates ingest, indexing, flex storage, archiving, and AI or LLM observability products, demonstrating that observability buyers actively optimize retention tiers and telemetry cost. | 中 | SM022 |
| CM036 | Elastic positions observability as an AI-powered, OpenTelemetry-first platform that unifies logs, metrics, and traces in one system. | 中 | SM023 |
| CM037 | Elastic claims up to 65% storage reduction for logs, up to 50% TCO reduction for long-term log retention, and 40% better latency since January 2026, reinforcing how much observability deals turn on efficiency. | 中 | SM023 |
| CM038 | AWS OpenSearch Service combines managed and serverless deployment for search, observability, and vector database workloads, including log analytics, generative AI, and RAG use cases. | 中 | SM024 |
| CM039 | Grafana predicts that in 2026 unified observability becomes the default operating model, data value overtakes data volume, AI shifts from copilot to collaborator, and OpenTelemetry becomes the default standard. | 中 | SM025 |
| CM040 | IBM argues that 2026 observability strategy will center on AI-driven intelligence, cost management, and compatibility with open standards such as OpenTelemetry, Prometheus, and Grafana. | 中 | SM026 |
| CM041 | The most conservative current TAM floor relevant to ClickHouse already exceeds $10 billion because independent 2026 cloud data warehouse estimates cluster around $14.5-$14.9 billion. | 中 | SM012, SM013 |
| CM042 | No clean public source isolates the narrower market for real-time columnar OLAP databases specifically; warehouse, streaming analytics, and observability estimates overlap and should not be summed. | 中 | SM012, SM013, SM016, SM017, SM019 |
| CM043 | ClickHouse's best market description is analytical data infrastructure spanning warehouse, event analytics, and observability workloads rather than a single software line item. | 中 | SM001, SM003, SM004, SM011 |
| CM044 | BigQuery, Datadog, Elastic, and AWS OpenSearch all show that incumbent substitutes increasingly bundle AI, observability, and managed operations into integrated platforms, which raises switching costs for ClickHouse deals. | 中 | SM020, SM021, SM023, SM024 |
| CM045 | The natural ClickHouse adoption path splits between self-managed control and managed cloud convenience, letting the company sell to both sovereignty-sensitive and operations-sensitive buyers. | 中 | SM002, SM003, SM004, SM009 |
| CP001 | ClickHouse markets itself as the fastest open-source analytical database. | 中 | SP001 |
| CP002 | ClickHouse says it supports data warehousing, real-time analytics, observability, and ML or GenAI workloads in one engine. | 中 | SP001 |
| CP003 | ClickHouse Server can be run locally, in major public clouds, or on on-premises hardware, while the same engine also underpins ClickHouse Cloud. | 中 | SP003 |
| CP004 | ClickHouse Cloud is a fully managed service available on the three major cloud marketplaces. | 中 | SP004 |
| CP005 | ClickHouse public pricing emphasizes separate compute and storage, autoscaling, and scale-to-zero economics instead of fixed always-on capacity. | 高 | SP002, SP004 |
| CP006 | ClickHouse raised a $350 million Series C in May 2025, bringing total funding to more than $650 million. | 中 | SP006 |
| CP007 | ClickHouse said in May 2025 that it served more than 2,000 customers and had grown more than 300% over the prior year. | 中 | SP006 |
| CP008 | ClickHouse positions its benchmark program as public and reproducible, with head-to-head cost and performance comparisons against other cloud data platforms. | 中 | SP007 |
| CP009 | DB-Engines describes ClickHouse as a high-performance column-oriented SQL OLAP system that is available both as open-source software and as a cloud offering. | 高 | SP001, SP008 |
| CP010 | ClickHouse, Inc. incorporated in 2021 around a project that had already been open-sourced in 2016, so the commercial company is younger than the software community around it. | 中 | SP005 |
| CP011 | Snowflake presents itself as a fully managed multi-cloud service with cross-region operation and built-in governance and security features. | 中 | SP009 |
| CP012 | Snowflake cost architecture is broken into compute, storage, and data-transfer charges, with compute consumed as Snowflake credits. | 高 | SP011, SP012 |
| CP013 | Snowflake warehouses can start and stop automatically, resize up or down, and use per-second billing with a 60-second minimum each time a warehouse starts. | 高 | SP011, SP012 |
| CP014 | Snowflake disclosed $9.77 billion of remaining performance obligations, 790 Forbes Global 2000 customers, and 733 $1 million-plus customers as of January 31, 2026. | 中 | SP010 |
| CP015 | Databricks describes its lakehouse as one architecture for integration, storage, processing, governance, sharing, analytics, and AI across major clouds. | 中 | SP014 |
| CP016 | Databricks pricing materials publish undiscounted list prices and SKU groups rather than a single simple platform sticker price. | 中 | SP015 |
| CP017 | Databricks frames the lakehouse as an open architecture that combines the best elements of data lakes and data warehouses on open formats. | 中 | SP017 |
| CP018 | Databricks says more than 20,000 organizations, including 70% of the Fortune 500, rely on its platform and that it works with more than 1,200 global partners. | 中 | SP016 |
| CP019 | BigQuery positions itself as an autonomous data-to-AI platform with built-in predictive analytics, generative-AI functions, graph analysis, and vector or hybrid search. | 中 | SP018 |
| CP020 | BigQuery offers a free tier and a serverless pricing model whose primary components are compute and storage, with slot reservations for committed capacity. | 高 | SP018, SP019 |
| CP021 | Amazon Redshift is positioned as a cloud data warehouse for analytics and agentic-AI workloads, with zero-ETL integrations and a lakehouse path tied to AWS services. | 中 | SP020 |
| CP022 | Redshift public pricing starts at $0.543 per hour for provisioned deployment and $1.50 per hour for serverless deployment, with per-second billing and reservation savings. | 中 | SP021 |
| CP023 | Amazon Athena is positioned as a serverless SQL and Spark analytics layer over S3 and other cloud or on-premises data stores with support for open formats. | 中 | SP022 |
| CP024 | Athena pricing is framed as pay for data processed or compute used, which makes it attractive for intermittent AWS-native analytics use cases. | 中 | SP023 |
| CP025 | DuckDB is an embedded in-process OLAP database with no separate server process, strong portability, vectorized execution, and an MIT license. | 中 | SP024 |
| CP026 | DuckDB is optimized for local or embedded analytical workflows rather than a vendor-managed enterprise control plane. | 中 | SP024 |
| CP027 | StarRocks markets itself as one engine for real-time, lakehouse, and AI analytics with consistent performance at scale. | 中 | SP025 |
| CP028 | Apache Druid positions itself as a high-performance real-time analytics database that supports sub-second queries on streaming and batch data at scale. | 中 | SP026 |
| CP029 | Imply Enterprise is the commercial distribution of Druid and can be deployed on any cloud with management and support tooling. | 中 | SP027 |
| CP030 | Imply Polaris exposes public starter and standard pricing, beginning at $100 per month and $600 per month respectively, which is unusual price transparency for a real-time analytics DBaaS. | 中 | SP028 |
| CP031 | SingleStore Helios is a cloud database service for real-time transactional, analytical, and RAG-style workloads with separate storage and compute. | 中 | SP029 |
| CP032 | SingleStore public pricing is usage-based and illustrated with credits-per-hour and storage-charge examples rather than one fixed enterprise subscription. | 中 | SP030 |
| CP033 | SingleStore also supports self-managed deployment on bare metal, virtual machines, cloud hosts, Docker, and Kubernetes. | 中 | SP032 |
| CP034 | ClickHouse offers more deployment sovereignty than Snowflake, BigQuery, or Athena because it can be self-managed or consumed as a managed cloud service on top of the same analytical engine. | 中 | SP001, SP003, SP009, SP018, SP022 |
| CP035 | Hyperscaler-owned products such as BigQuery, Redshift, and Athena have materially stronger procurement and bundle leverage than ClickHouse because they ride Google Cloud and AWS account control, adjacent services, and annual-report-scale parents. | 中 | SP018, SP020, SP022, SP033, SP034 |
| CP036 | Snowflake and Databricks are the closest broad-platform competitors to ClickHouse because both pair analytics with wider governance or AI suites, while ClickHouse stays more concentrated on high-speed analytical serving workloads. | 中 | SP001, SP009, SP014 |
| CP037 | DuckDB and Druid are narrower substitutes than Snowflake or Databricks: DuckDB focuses on embedded local analytics and Druid focuses on streaming-first real-time analytics. | 中 | SP024, SP026 |
| CP038 | StarRocks and SingleStore overlap with ClickHouse on low-latency analytical serving, but StarRocks leans into lakehouse and AI SQL while SingleStore leans into HTAP-style application databases. | 中 | SP025, SP029, SP031 |
| CP039 | ClickHouse pricing is easier to reason about than Databricks SKU sheets, but AWS, Athena, Imply, and SingleStore publish more concrete public starting prices than ClickHouse does. | 中 | SP002, SP015, SP021, SP023, SP028, SP030 |
| CP040 | Snowflake pricing mechanics are more explicit around credits and warehouse sizes than ClickHouse pricing, which emphasizes autoscaling, compute-storage separation, and scale-to-zero behavior. | 中 | SP002, SP011, SP012 |
| CP041 | ClickHouse has a strong developer-facing open-source posture, while Snowflake, BigQuery, Redshift, Athena, and SingleStore present primarily as proprietary services or products. | 中 | SP001, SP008, SP009, SP018, SP020, SP022, SP029 |
| CP042 | BigQuery, Redshift, and Athena create the most immediate GTM and bundle risk to ClickHouse because they can be sold as one more service inside existing hyperscaler estates. | 中 | SP018, SP020, SP022, SP033, SP034 |
| CP043 | ClickHouse is best aligned to speed-sensitive analytical serving, observability, and real-time warehousing workloads rather than office-suite BI bundling or embedded local analytics. | 中 | SP001, SP004, SP006, SP024 |
| CP044 | ClickHouse is most vulnerable where buyers prioritize a broader governed data-and-AI suite from Snowflake or Databricks or default to native hyperscaler procurement. | 中 | SP009, SP014, SP018, SP020 |
| CP045 | Multi-homing is rational in this category because Snowflake warehouses, BigQuery serverless scans and slots, Redshift or Athena AWS-native workflows, DuckDB local analysis, and ClickHouse real-time serving optimize different jobs. | 中 | SP002, SP011, SP019, SP021, SP023, SP024 |
| CP046 | ClickHouse has enough funding and customer proof to matter globally, but its commercial reach is still smaller than Snowflake’s public scale and Databricks’ customer-plus-partner footprint. | 中 | SP006, SP010, SP016 |
| CP047 | ClickHouse’s open-source and self-hostable posture is a real differentiator for sovereignty-minded buyers, but the same openness means it must continue winning on performance and developer love rather than pure lock-in. | 中 | SP001, SP003, SP024, SP026 |
| CP048 | Among the named alternatives, ClickHouse, DuckDB, and Apache Druid most clearly surface as open-source projects in the reviewed materials, while Databricks emphasizes open formats more than an open-source core. | 中 | SP001, SP017, SP024, SP026 |
| CP049 | SingleStore and ClickHouse both sell separated storage and compute in cloud form, but SingleStore packages that around mixed transactional and analytical application needs while ClickHouse remains analytics-centric. | 中 | SP004, SP029, SP032 |
| CP050 | The main competitive risk to ClickHouse is not one knockout rival but a segmented market where Snowflake and Databricks win breadth, hyperscalers win procurement, and DuckDB, Druid, StarRocks, or SingleStore win narrower use cases. | 中 | SP009, SP014, SP018, SP020, SP022, SP024, SP025, SP026, SP029 |
| CP051 | ClickHouse Cloud’s architecture emphasizes object-backed parallel replicas and separate compute layers for read and write workloads. | 中 | SP004 |
| CP052 | Snowflake highlights encryption, RBAC, network policies, MFA, masking, and Horizon Catalog as key parts of its trust and governance posture. | 中 | SP009 |
| CP053 | Redshift claims up to 2.2x better price-performance and 7x better throughput than other cloud data warehouses, which shows how aggressively incumbents market analytical efficiency. | 中 | SP020 |
| CP054 | Druid says its architecture can support from hundreds to one hundred thousand queries per second, underscoring its relevance for high-concurrency real-time analytics. | 中 | SP026 |
| CI001 | ClickHouse raised $250 million in Series B at a $2 billion valuation on October 28, 2021. | 高 | SI010, SI012 |
| CI002 | The 2021 Series B followed an earlier roughly $50 million Series A raised in August 2021. | 高 | SI010, SI012 |
| CI003 | ClickHouse raised $350 million in Series C financing in May 2025 and said Khosla Ventures led the round. | 高 | SI005, SI006, SI013 |
| CI004 | The May 2025 Series C brought ClickHouse total disclosed funding to more than $650 million. | 中 | SI005, SI006, SI007 |
| CI005 | ClickHouse disclosed a $100 million credit facility alongside the Series C financing. | 高 | SI005, SI006, SI013 |
| CI006 | January 2026 reporting tied ClickHouse to a $15 billion valuation, roughly 2.5 times the $6.35 billion valuation reported for May 2025. | 高 | SI008, SI009 |
| CI007 | Company-linked Series C materials said ClickHouse grew over 300% during the year before the May 2025 financing announcement. | 中 | SI005, SI006 |
| CI008 | Company-linked May 2025 materials said ClickHouse served more than 2,000 customers. | 中 | SI005, SI006 |
| CI009 | By October 2025, ClickHouse said it had more than quadrupled ARR over the prior year while still exceeding 2,000 customers. | 中 | SI004, SI020 |
| CI010 | TechCrunch and Sacra both indicated ClickHouse Cloud ARR was growing more than 250% year over year going into 2026. | 中 | SI008, SI009 |
| CI011 | Sacra estimated ClickHouse reached about $160 million in annualized revenue in 2025, up 256% from a $45 million exit rate at the end of 2024. | 中 | SI009 |
| CI012 | A cautious public range for 2025 ARR or annualized revenue is roughly $150 million to $200 million. | 中 | SI008, SI009 |
| CI013 | ClickHouse monetizes primarily through managed cloud services while the core database remains open source and free to use. | 高 | SI009, SI010, SI012 |
| CI014 | ClickHouse Cloud entered public beta on AWS in October 2022. | 高 | SI011, SI014, SI015 |
| CI015 | ClickHouse Cloud became generally available on December 6, 2022. | 高 | SI003, SI012 |
| CI016 | The December 2022 GA release extended the free trial to 30 days, added low-monthly-spend Development Services, reduced production pricing, and improved compute metering. | 高 | SI003, SI012 |
| CI017 | Current ClickHouse public pages describe usage-based pricing with separate compute and storage scaling and pay-for-use economics. | 高 | SI001, SI002, SI009 |
| CI018 | TechCrunch described ClickHouse Cloud as a PLG motion where users can start with free credits and move to transparent consumption-based monthly billing. | 高 | SI001, SI012 |
| CI019 | m3ter said ClickHouse simplified beta pricing from read and write units to consumed storage plus compute before GA. | 高 | SI014, SI003 |
| CI020 | ClickHouse publicly markets monetization across real-time analytics, data warehousing, observability, and AI-powered data applications. | 高 | SI001, SI023, SI024, SI025 |
| CI021 | ClickPipes and other managed integrations create monetizable adjacencies inside ClickHouse Cloud beyond core query serving. | 中 | SI001, SI009, SI021 |
| CI022 | Dedicated and Bring-Your-Own-Cloud deployment options imply higher-ACV enterprise motions than the open-source core alone. | 中 | SI001, SI009 |
| CI023 | ClickHouse emphasizes compute-storage separation and autoscaling as cost-efficiency drivers relative to self-managed deployments. | 高 | SI001, SI009, SI012 |
| CI024 | Public sources do not disclose realized enterprise pricing, discount levels, or the actual mix between cloud, support, and adjacent products. | 高 | SI001, SI002, SI009 |
| CI025 | Public sources do not disclose ClickHouse gross margin, net revenue retention, or churn. | 中 | SI001, SI009 |
| CI026 | Public sources do not disclose cash on hand, monthly burn, or remaining runway. | 高 | SI005, SI009, SI013 |
| CI027 | Public sources disclose the existence of a $100 million facility but not the draw, covenant, rate, or maturity details needed for full debt underwriting. | 高 | SI005, SI013 |
| CI028 | ClickHouse said in 2025 that it hired Jimmy Sexton as CFO, suggesting a more mature finance function. | 中 | SI004 |
| CI029 | AI-native customers are central to the current ClickHouse Cloud growth narrative. | 中 | SI004, SI005, SI008 |
| CI030 | ClickStack extends ClickHouse monetization into managed observability workloads on the cloud platform. | 高 | SI001, SI024 |
| CI031 | Real-time analytics and data warehousing remain the clearest publicly marketed workload categories and likely core revenue drivers. | 高 | SI001, SI023, SI025 |
| CI032 | The free open-source core functions as top-of-funnel freemium acquisition for the managed cloud offering rather than a separately disclosed revenue line. | 中 | SI009, SI010, SI012 |
| CI033 | By October 2021 ClickHouse pointed to more than 20,000 GitHub stars and 800-plus contributors, indicating a large open-source community funnel. | 中 | SI010 |
| CI034 | TechCrunch reported that early cloud traction already exceeded 100 customers around the December 2022 GA launch. | 中 | SI012 |
| CI035 | Snowflake, MongoDB, Confluent, and Elastic all show current annual report cycles through SEC EDGAR, offering public financial transparency that ClickHouse does not match as a private company. | 高 | SI016, SI017, SI018, SI019 |
| CI036 | Those public-company filing trails provide a more transparent benchmark for discussing category economics and risk than ClickHouse currently offers. | 中 | SI016, SI017, SI018, SI019 |
| CI037 | ClickHouse faces pricing pressure because buyers can compare managed cloud against both self-hosted open source and incumbent analytics platforms. | 中 | SI009, SI016 |
| CI038 | Sacra highlighted a downside case in which incumbent platforms such as Snowflake and BigQuery narrow ClickHouse performance differentiation. | 中 | SI009 |
| CI039 | The disclosed funding stack shows strong capital access, but the absence of current liquidity data prevents a firm runway view. | 中 | SI005, SI008, SI009, SI013 |
| CI040 | Series C proceeds were earmarked for product development, global expansion, and AI-native customer partnerships. | 高 | SI005, SI006, SI013 |
| CI041 | Current ClickHouse pages still frame free-trial credits and pay-for-use economics as customer-acquisition levers in 2026. | 高 | SI001, SI023, SI025 |
| CI042 | As of the 2026 run date, the best public picture is strong growth and fundraising paired with opaque unit economics and liquidity. | 中 | SI008, SI009, SI013 |
| CI043 | The October 2025 Series C extension indicates investors were still willing to add capital after the initial May 2025 round. | 中 | SI004, SI013, SI020 |
| CI044 | Expansion across AWS, GCP, and Azure marketplaces broadens enterprise distribution, but public sources still do not disclose CAC or payback. | 中 | SI001, SI009 |
| CE001 | ClickHouse is positioned publicly as both an open-source column-oriented OLAP DBMS and a managed cloud offering. | 中 | SE002, SE016, SE019 |
| CE002 | ClickHouse Cloud is a fully managed service where infrastructure, maintenance, scaling, and operations are handled by ClickHouse. | 中 | SE002 |
| CE003 | ClickHouse Cloud is presented as available on all three major cloud marketplaces/providers. | 中 | SE001, SE021 |
| CE004 | ClickHouse Cloud supports centralized marketplace billing across AWS, Azure, and GCP subscriptions. | 中 | SE010 |
| CE005 | The 2026 cloud changelog names AWS Mexico, Azure Australia East, and GCP London as supported ClickHouse Cloud regions or region additions. | 中 | SE010 |
| CE006 | The cloud overview describes ClickHouse Cloud as providing serverless operations, autoscaling, backups, replication, and high availability. | 中 | SE002 |
| CE007 | ClickHouse says its cloud architecture separates storage and compute with pay-for-use compute scaling. | 中 | SE001 |
| CE008 | ClickHouse Cloud uses object-backed parallel replicas in a shared-nothing architecture to reduce storage duplication and network overhead. | 中 | SE001 |
| CE009 | The Shared database engine works with Shared Catalog to manage databases whose tables use stateless engines such as SharedMergeTree. | 中 | SE006 |
| CE010 | The Shared database engine removes local-disk dependency by storing metadata in a central versioned state that compute nodes fetch on startup. | 中 | SE006 |
| CE011 | The VLDB architecture overview says ClickHouse uses a vectorized query execution engine with optional code compilation. | 中 | SE003 |
| CE012 | The development architecture page says ClickHouse processes data as arrays or chunks and dispatches operations on arrays whenever possible. | 中 | SE004 |
| CE013 | ClickHouse documents its engine as layered into query processing, storage, integration, and access/control components. | 中 | SE003 |
| CE014 | MergeTree-family engines are designed for high ingest rates and large data volumes by creating parts that are merged in the background. | 中 | SE005 |
| CE015 | MergeTree primary keys index blocks of rows called granules rather than individual rows, and the default index granularity is 8192. | 中 | SE005 |
| CE016 | ClickHouse documents sparse primary indexes as memory-efficient for very large tables but notes that a range read can still pull up to index_granularity*2 extra rows per data block. | 中 | SE005 |
| CE017 | ClickHouse explicitly links better compression to sorting data by a consistent primary key. | 中 | SE005 |
| CE018 | The architecture overview says ClickHouse integrates external databases, Kafka and RabbitMQ, lakehouse table formats, and object storage through its integration layer. | 中 | SE003 |
| CE019 | ClickPipes is described as a Cloud-only managed ingestion engine for Kafka, S3, PostgreSQL, MongoDB, GCS, MySQL, and other sources. | 中 | SE001 |
| CE020 | The Kafka engine documentation recommends ClickPipes on ClickHouse Cloud for private networking, independent scaling, and monitoring of Kafka ingestion. | 中 | SE007 |
| CE021 | The Kafka table engine supports configurable consumers, security protocols, schema-aware parsing, and materialized-view based streaming pipelines. | 中 | SE007 |
| CE022 | The dbt-clickhouse adapter supports table, view, incremental, materialized-view, tests, snapshots, and seeds workflows. | 中 | SE008 |
| CE023 | The dbt adapter also exposes ClickHouse-specific codecs, TTLs, indexes, and projections and documents CI/CD patterns for staging and production. | 中 | SE008 |
| CE024 | ClickHouse organizes integrations into core, partner, and community tiers rather than presenting every connector as first-party. | 中 | SE009 |
| CE025 | The Microsoft Power Query connector is GA and supports Power BI semantic models, Dataflows, and Fabric Dataflow Gen2. | 中 | SE020 |
| CE026 | ClickHouse’s Azure GA announcement highlights turnkey integrations with Power BI, Azure Event Hubs, and Azure Blob Storage. | 中 | SE021 |
| CE027 | The Cloudflare Logpush integration guide says ClickPipes can ingest from S3 with exactly-once semantics and replay capability. | 中 | SE011 |
| CE028 | The cloud overview lists GDPR, HIPAA, ISO 27001, ISO 27001 SoA, PCI DSS, and SOC 2 among ClickHouse Cloud compliance programs. | 中 | SE002 |
| CE029 | The cloud overview lists SSO, multi-factor authentication, RBAC, Private Link, Private Service Connect, IP filtering, and CMEK as cloud security controls. | 中 | SE002 |
| CE030 | The Azure GA announcement says ClickHouse Cloud is built with network isolation, traffic encryption, and multi-availability-zone replication. | 中 | SE021 |
| CE031 | The 2026 changelog shows organization spend alerts, dual-window autoscaling, and primary-service idling reaching GA or active rollout during 2026. | 中 | SE010 |
| CE032 | Index sharding entered private preview in 2026 to distribute index analysis across replicas, cut per-replica memory, and improve query performance. | 中 | SE010 |
| CE033 | The cloud changelog says BYOC on GCP became GA and ClickPipes reached AWS region parity in 2026. | 中 | SE010 |
| CE034 | The public GitHub repo showed roughly 47.6k stars, 8.4k forks, and 796 releases on the access date. | 中 | SE014 |
| CE035 | The documentation repo page exposes active maintenance metadata, watchers, and multi-language docs contributions. | 中 | SE015 |
| CE036 | PyPI listed clickhouse-connect 1.1.0, released on 2026-05-26, with Python 3.10-3.14 support plus async, SQLAlchemy, Pandas, and Superset capabilities. | 中 | SE017 |
| CE037 | The npm package page showed @clickhouse/client 1.11.1 with 503,589 weekly downloads and documented streaming support for Node.js and browser clients. | 中 | SE018 |
| CE038 | Docker Hub showed more than 100 million pulls for clickhouse/clickhouse-server and documented the HTTP 8123 and native 9000 interfaces. | 中 | SE019 |
| CE039 | DB-Engines describes ClickHouse as a high-performance column-oriented SQL OLAP DBMS available as open source and as a cloud offering. | 中 | SE016 |
| CE040 | DB-Engines lists access methods including HTTP REST, JDBC, ODBC, and PostgreSQL/MySQL-compatible wire protocols. | 中 | SE016 |
| CE041 | A TrustRadius reviewer credits MergeTree, primary-key data skipping, and compression for strong query performance and cites one workflow improving from more than two minutes to under one second. | 中 | SE022 |
| CE042 | The same TrustRadius review flags gaps in SQL console query-plan UX, cloud role granularity, and some SSO IdP compatibility. | 中 | SE022 |
| CE043 | ClickHouse’s Cloudflare meetup recap says Cloudflare ran ClickHouse in production by the end of 2016 and later exceeded 1,000 active replicas. | 中 | SE013 |
| CE044 | The Cloudflare recap says the deployment processed hundreds of millions of inserted rows per second and maintained 184 dictionaries. | 中 | SE013 |
| CE045 | Cloudflare said moving dictionaries from hashed to hashed-array layouts reduced memory footprints by more than 4x. | 中 | SE013 |
| CE046 | HypeQuery argues that scaled ClickHouse deployments commonly add query translation and semantic layers because optimized schemas are too cognitively expensive for broad self-service use. | 中 | SE023 |
| CE047 | HypeQuery cites Uber, Cloudflare, Instacart, Microsoft, GitLab, Lyft, and Contentsquare as converging on similar abstraction stacks above ClickHouse. | 中 | SE023 |
| CE048 | ClickHouse’s adopters page shows production use across observability, SEO, blockchain, cloud data platforms, and security-related workloads. | 中 | SE012 |
| CE049 | ClickHouse’s user stories and analytics video position the product in observability, customer-facing analytics, and other large-scale analytics workflows. | 中 | SE024, SE025 |
| CE050 | Microsoft’s connector documentation requires an ODBC driver and, for cloud service scenarios, an on-premises data gateway to bridge ClickHouse into Power BI services. | 中 | SE020 |
| CE051 | The cloud overview positions the SQL console and clickhousectl CLI as part of the managed cloud operating surface. | 中 | SE002 |
| CE052 | ClickHouse Cloud documents compute-compute separation, with independent compute layers for read and write workloads. | 中 | SE001 |
| CU001 | ClickHouse's current public proof base spans observability, product analytics, AI and LLM operations, fintech, education, mobility, and industrial analytics rather than a single vertical. | 高 | SU001, SU032 |
| CU002 | The public customer-story set visibly over-indexes toward observability and real-time analytics workloads rather than general-purpose enterprise warehousing. | 中 | SU001 |
| CU003 | In most public ClickHouse deployments, engineering, platform, SRE, or data teams are the buyers and operators, while broader end users consume governed dashboards, APIs, or product surfaces rather than raw ClickHouse directly. | 中 | SU001, SU016, SU018, SU020 |
| CU004 | The official adopter list and external customer lists include Cloudflare, Contentsquare, eBay, Spotify, Uber, and ByteDance, but the quality of evidence differs materially by logo. | 中 | SU002, SU028 |
| CU005 | Cloudflare has used ClickHouse in production since late 2016 and had exceeded 1,000 active replicas by 2023. | 中 | SU004 |
| CU006 | Cloudflare uses ClickHouse for HTTP analytics, DNS analytics, logging analytics, Workers runtime analysis, internal analytics, customer dashboards, Firewall Analytics, and Cloudflare Radar. | 高 | SU004, SU005, SU006 |
| CU007 | Cloudflare's 2025 public demo showed ClickHouse scanning 96 trillion events over one hour and 1.61 quadrillion events over one day in under two seconds. | 中 | SU003 |
| CU008 | Cloudflare's 2018 HTTP analytics pipeline used ClickHouse to support analytics on traffic running at roughly 6 million requests per second. | 中 | SU005 |
| CU009 | Cloudflare's 2022 log analytics migration to ClickHouse stored 100% of events and reduced inserter CPU and memory consumption by eight times compared with the prior Elasticsearch-centered setup. | 中 | SU006 |
| CU010 | Cloudflare's 2026 Ready-Analytics platform uses ClickHouse for millions of calls per day in usage billing, powers hundreds of millions of dollars in usage revenue and fraud workflows, and had grown to more than 2 PiB with millions of rows per second of ingest. | 中 | SU007 |
| CU011 | Contentsquare migrated its main SaaS analytics product from Elasticsearch to ClickHouse and reported 11x lower infrastructure cost, 10x p99 query improvement, and expansion of retention from one month-equivalent historical access to 13 months. | 高 | SU010, SU009 |
| CU012 | Contentsquare built AST-based query optimization on top of ClickHouse, and an external architecture roundup describes this abstraction layer as driving 10x speedups on the slowest 5% of queries. | 高 | SU010, SU009 |
| CU013 | OpenAI ingests petabytes of log data per day, says volume is growing by more than 20% per month, and runs its ClickHouse observability system with 90 shards and two replicas before adding a third query replica. | 中 | SU011 |
| CU014 | After GPT-4o image generation launched in March 2025, OpenAI experienced a 50% overnight log-volume spike and then achieved a 40% CPU reduction from a one-line Bloom-filter optimization in ClickHouse. | 中 | SU011 |
| CU015 | The accessible public proof for Uber is mainly secondary architecture coverage and older adopter-list references rather than a fresh customer-authored 2026 engineering case study. | 中 | SU009, SU002, SU029 |
| CU016 | The accessible public proof for Spotify is mainly an adopter-list entry tied to older slides rather than a fresh detailed deployment write-up. | 中 | SU002, SU030 |
| CU017 | LeadCognition lists ByteDance alongside Cloudflare, Uber, eBay, and Spotify as a production ClickHouse user, but the fetched proof quality is lower than current flagship case studies. | 低 | SU028 |
| CU018 | Anthropic says ClickHouse played an instrumental role in shipping Claude 4 and now operates a custom air-gapped version of the ClickHouse Cloud architecture inside its secure compute environment. | 中 | SU012 |
| CU019 | Tesla's Comet platform on ClickHouse currently ingests tens of millions of rows per second and survived a one-billion-rows-per-second load test running for 11 days. | 中 | SU013 |
| CU020 | The same Tesla load test crossed one quadrillion rows without reported instability, making it one of the strongest public proofs of ClickHouse's telemetry-scale ceiling. | 中 | SU013 |
| CU021 | Microsoft Clarity launched publicly with ClickHouse at its core and says the system now supports millions of projects, hundreds of trillions of events, and hundreds of petabytes of data. | 高 | SU014, SU001 |
| CU022 | Microsoft Clarity says heat map generation moved from a roughly 30-minute offline workflow to an instantaneous task after choosing ClickHouse. | 中 | SU014 |
| CU023 | Replo's official customer story describes an analytics system trusted by more than 4,000 Shopify merchants that processes more than 100 billion events and ingests 3,000 to 5,000 events per second. | 中 | SU015 |
| CU024 | Mintlify says ClickHouse reduced dashboard latency from tens of seconds to sub-one-second, removed weekly analytics bug reports, improved estimated NPS by roughly 30%, and cut cost by around 60% versus PostHog. | 中 | SU016 |
| CU025 | Padlet reports around 40 million monthly unique users, usage in 242 of 246 countries, roughly 8 billion events in a month, 45 millisecond median query latency, and 690 millisecond p99 latency on its ClickHouse analytics layer. | 中 | SU017 |
| CU026 | Buildkite Test Engine grew from roughly 3 billion to 12 billion test executions per month in six months, now stores about 70 billion records in ClickHouse, peaks above 25,000 events per second, and saves eight dollars elsewhere for each dollar spent on ClickHouse. | 中 | SU018 |
| CU027 | Ramp says it serves more than 50,000 customers and found that 16,000 randomized queries across 60 million transactions completed in about 12 seconds on ClickHouse, while enterprise reports that timed out after 40 seconds on Postgres returned in milliseconds after migration. | 中 | SU019 |
| CU028 | Qonto serves more than 600,000 small businesses and freelancers across eight countries and says ClickHouse stores 231 TB of uncompressed span attributes in 376 GB, implying a 99.84% compression ratio and about $70,000 of annual storage savings. | 中 | SU020 |
| CU029 | Langfuse says its newer immutable Events table on ClickHouse uses roughly three times less memory and produces up to 20 times faster queries than its older mutable model. | 中 | SU021 |
| CU030 | Lyft's ClickHouse Cloud deployment reads more than 450 TB per day, writes about 4 TB per day, and handles hundreds of queries per second on average with peaks in the thousands. | 中 | SU022 |
| CU031 | Polymarket says it implemented a ClickHouse data warehouse in weeks and now serves its leaderboard API at 100s of requests per second with average latency around 25 milliseconds. | 中 | SU023 |
| CU032 | ClickHouse Cloud markets separate storage and compute, pay-for-use pricing, major cloud-marketplace availability, and reduced shard and replica management overhead. | 高 | SU024, SU025 |
| CU033 | Open-source availability and SQL compatibility are repeatedly cited in public materials and reviews as reducing vendor lock-in and lowering the friction of evaluation or migration. | 中 | SU011, SU027 |
| CU034 | TrustRadius reviewers praise ClickHouse's MergeTree performance and real-time warehouse utility but flag limitations in SQL-console features, cloud role granularity, and SSO support. | 中 | SU026 |
| CU035 | PeerSpot gives ClickHouse an average rating of 8.6 out of 10 and highlights speed, scalability, compression, and lack of vendor lock-in as strengths. | 中 | SU027 |
| CU036 | PeerSpot review summaries also surface recurring complaints around documentation, UI and security/admin maturity, setup complexity, and cloud pricing visibility. | 中 | SU027 |
| CU037 | Public proof shows broad logo diversity across AI, fintech, developer tools, education, mobility, and industrial analytics, which reduces obvious single-vertical concentration risk. | 中 | SU001, SU032 |
| CU038 | ClickHouse does not publicly disclose customer NRR, GRR, logo churn, renewal rates, or top-customer ARR share in the fetched materials, leaving durability and concentration as unresolved diligence items. | 中 | SU001, SU026, SU027, SU024 |
| CU039 | Public customer stories repeatedly show a land-and-expand motion in which ClickHouse lands as the fix for one urgent analytics or observability bottleneck and then expands into adjacent product or operational use cases. | 中 | SU010, SU016, SU018, SU019, SU020, SU023 |
| CU040 | Evidence quality splits into high-confidence customer-quoted production case studies, medium-confidence secondary architecture summaries, and low-confidence adopter-list or logo-directory proof, so marquee-logo claims should be weighted accordingly. | 高 | SU003, SU009, SU002, SU028 |
| CR001 | In May 2025 ClickHouse announced a $350 million Series C round at a $6.35 billion valuation. | 中 | SR001, SR002 |
| CR002 | The same financing package brought ClickHouse total funding to over $650 million and included a $100 million credit facility led by Stifel and Goldman Sachs. | 高 | SR001, SR002, SR003 |
| CR003 | ClickHouse said it grew over 300% during the prior year and now serves more than 2,000 customers. | 中 | SR001, SR002 |
| CR004 | The May 2025 financing announcement named Anthropic, Tesla, Mercado Libre, Sony, Meta, Memorial Sloan Kettering, Lyft, and Instacart as customers or recent wins. | 中 | SR001, SR002 |
| CR005 | ClickHouse Cloud is marketed as a fully managed serverless service with pay-for-use compute and autoscaling. | 高 | SR004, SR005 |
| CR006 | ClickHouse Cloud publicly offers a 30 day trial with $300 credits, reinforcing a self-serve developer acquisition motion. | 高 | SR004, SR013, SR014 |
| CR007 | The pricing page says ClickHouse separately meters storage and compute, scales unused resources toward zero, and lets customers set autoscaling limits to contain bill shock. | 中 | SR005 |
| CR008 | ClickHouse documents SLAs only for select committed-spend contracts rather than for every cloud user. | 中 | SR006 |
| CR009 | The public status page reported 98.62% aggregate uptime for February through May 2026 while AWS components showed 100% uptime. | 中 | SR007 |
| CR010 | ClickHouse Cloud says it has maintained SOC 2 Type II since 2022 and ISO 27001 since 2023. | 中 | SR008 |
| CR011 | ClickHouse says it self-certified to the U.S. Data Privacy Framework in 2024 and maintains internal GDPR and CCPA compliance programs. | 高 | SR008, SR009 |
| CR012 | HIPAA support is available only on the Enterprise plan and PCI service-provider compliance was added in 2025. | 中 | SR008 |
| CR013 | Privacy and compliance disclosures imply that ClickHouse must continuously manage cross-border data-transfer and privacy obligations for cloud customers. | 中 | SR008, SR009 |
| CR014 | ClickHouse documented CVE-2025-1385 as a route to arbitrary code execution on misconfigured self-managed servers that enable library bridge plus file upload paths. | 高 | SR019, SR020, SR022 |
| CR015 | Both ClickHouse’s security changelog and its GitHub advisory say ClickHouse Cloud was not vulnerable to CVE-2025-1385. | 中 | SR019, SR020 |
| CR016 | A GitHub advisory describes a query-cache bug in which switching roles under a single user can bypass role-based access controls and expose unauthorized rows. | 中 | SR021, SR019 |
| CR017 | ClickHouse offers cloud, server, local CLI, clickhouse-local, and embedded chDB deployment modes on the same core engine. | 中 | SR012 |
| CR018 | ClickHouse’s product and community pages advertise roughly 2.9k contributors, 29k pull requests, and 47.6k GitHub stars. | 高 | SR011, SR015 |
| CR019 | The community page says ClickHouse has more than 12k product makers on Slack, underscoring a developer-led funnel rather than a pure top-down enterprise motion. | 中 | SR015 |
| CR020 | Altinity argues that important capabilities such as SharedMergeTree, lightweight updates, and S3 role-based access are now cloud-only, creating open-core and fork risk for the ecosystem. | 中 | SR023 |
| CR021 | The same Altinity analysis says community trust is strained by an unclear OSS roadmap and by the ClickHouse team acting as a pull-request bottleneck. | 中 | SR023 |
| CR022 | ClickHouse Cloud markets ClickPipes and several managed integrations as cloud-only conveniences that do not exist in the same form for self-managed users. | 中 | SR004 |
| CR023 | In DB-Engines May 2026 rankings, Snowflake was #6 and Databricks #7 while ClickHouse ranked #26. | 中 | SR026 |
| CR024 | The same ranking placed DuckDB at #44 and StarRocks at #142, showing that both alternatives continue to accumulate mindshare even from much smaller bases. | 中 | SR026 |
| CR025 | Exasol’s February 2026 benchmark found ClickHouse’s successful query runtime degraded 1.39x between 1 and 16 concurrent streams. | 中 | SR024 |
| CR026 | The same benchmark found ClickHouse query success fell from 82% at one node to 36% at eight nodes on TPC-H style distributed tests because large joins hit per-shard memory limits. | 中 | SR024 |
| CR027 | Exasol’s benchmark says DuckDB still degrades 41% at 16 concurrent streams but remains attractive where single-process simplicity matters more than cluster concurrency. | 中 | SR024 |
| CR028 | A StarRocks-sponsored benchmark article claimed ClickHouse could not complete its TPC-H test set and that StarRocks was 2.2x faster on wide-table SSB queries. | 低 | SR025 |
| CR029 | ClickHouse’s own product page explicitly pitches migrations from Snowflake, Redshift, Elastic, and Druid, confirming that incumbent warehouse and analytics budgets are the primary target market. | 中 | SR011 |
| CR030 | ClickHouse Cloud says it is available on all three major cloud marketplaces and handles updates, backups, scaling, and security patches automatically. | 中 | SR004 |
| CR031 | ClickHouse claims its lower cloud cost comes from compute-storage separation, autoscaling, object-backed parallel replicas, and lower replica overhead. | 中 | SR004, SR005 |
| CR032 | ClickHouse argues that many teams move from open source to ClickHouse Cloud because they do not want to manage shards, replicas, upgrades, or security patches themselves. | 中 | SR004, SR012 |
| CR033 | Use-case pages position high concurrency, millions of rows per second, and interactive dashboards as core reasons to adopt ClickHouse Cloud. | 中 | SR013, SR014 |
| CR034 | The public financing disclosures emphasized customers and growth but did not publish ARR, revenue, gross margin, or profitability metrics. | 高 | SR001, SR002, SR003 |
| CR035 | Because the May 2025 $6.35 billion mark came without public unit-economics disclosure, valuation underwriting depends on future cloud monetization and margin capture that outsiders cannot yet verify. | 中 | SR001, SR004, SR005 |
| CR036 | The combination of free trial credits, local deployment options, and an open-source core creates a product-led funnel that can expand adoption faster than it converts to durable paid cloud spend. | 中 | SR004, SR012, SR013, SR015 |
| CR037 | Public materials disclose more than 2,000 customers and committed-spend SLAs but no top-customer share, cohort retention, or NRR, leaving concentration risk materially under-documented. | 中 | SR001, SR006 |
| CR038 | Enterprise-only HIPAA and PCI features imply that regulated high-spend customers likely matter disproportionately to cloud monetization. | 中 | SR004, SR008 |
| CR039 | Offering the same engine across OSS, local, embedded, and fully managed modes reduces raw lock-in for buyers but increases cannibalization and upgrade-path risk for the paid cloud business. | 中 | SR012, SR017, SR023 |
| CR040 | DB-Engines describes ClickHouse as both an Apache 2.0 open-source product and a cloud offering with row policies, quotas, resource limits, and multiple wire protocols. | 中 | SR027 |
| CR041 | Public uptime documentation and aggregate status metrics show that ClickHouse Cloud reliability remains a live execution variable rather than a solved background condition. | 中 | SR006, SR007 |
| CR042 | ClickHouse’s security changelog records repeated crash, ACL, and RCE-class issues across recent OSS releases, making patch discipline part of the enterprise trust story. | 高 | SR019, SR022 |
| CR043 | The CVE-2025-1385 advisory explicitly tells maintainers of forked ClickHouse versions to port the fix themselves, raising maintenance burden for any serious fork. | 中 | SR020 |
| CR044 | TechCrunch framed ClickHouse as a direct Snowflake and Databricks challenger, reinforcing that investor expectations are tied to taking share from much larger data-platform incumbents. | 中 | SR029 |
| CR045 | The Tinybird alternatives page and ClickHouse’s own deployment docs show developers have multiple ways to use ClickHouse-compatible or adjacent analytics stacks without defaulting to ClickHouse Cloud. | 低 | SR012, SR030 |
| CR046 | ClickHouse’s real-time analytics, warehousing, and adopters surfaces show broad workload coverage that expands addressable market but also broadens the product and support execution surface. | 中 | SR013, SR014, SR016 |
| CR047 | The combination of strong community breadth and visible roadmap tension means governance missteps would be amplified across a large developer base rather than staying a niche issue. | 中 | SR015, SR023 |
| CR048 | The cleanest public diligence asks are revenue quality by cohort, top-customer exposure, cloud gross margin, incident history by service tier, and an explicit OSS-versus-cloud roadmap split. | 中 | SR001, SR006, SR007, SR023 |
| CV001 | ClickHouse raised $350 million in a Series C round on May 29, 2025 led by Khosla Ventures. | 高 | SV001, SV013 |
| CV002 | The Series C included BOND, IVP, Battery Ventures, Bessemer Venture Partners, and existing investors Index Ventures, Lightspeed, GIC, Benchmark, Coatue, FirstMark, and Nebius, taking total funding to more than $650 million. | 高 | SV001, SV014 |
| CV003 | ClickHouse also secured a $100 million credit facility led by Stifel and Goldman Sachs alongside the Series C financing. | 中 | SV001, SV003 |
| CV004 | At the time of the round, ClickHouse said it had grown more than 300% over the prior year and served more than 2,000 customers. | 高 | SV001, SV004 |
| CV005 | ClickHouse describes itself as an open-source columnar database management system built for real-time analytics and large-scale analytical workloads. | 高 | SV001, SV008 |
| CV006 | ClickHouse Cloud monetizes through usage-based pricing with separate compute and storage charges rather than fixed-seat software pricing. | 高 | SV006, SV007 |
| CV007 | ClickHouse Cloud emphasizes pay-for-use compute, separate storage, and managed autoscaling as core commercial mechanics. | 高 | SV006, SV007 |
| CV008 | ClickHouse says its cloud architecture separates storage and compute and can offer faster warm-up and better economics than Snowflake for real-time analytics use cases. | 中 | SV007, SV009 |
| CV009 | Sacra estimated that ClickHouse reached about $160 million in annualized revenue in 2025. | 中 | SV011, SV012 |
| CV010 | Sacra reported that ClickHouse Cloud ARR was growing more than 250% year over year as of January 2026. | 中 | SV011, SV012 |
| CV011 | Sacra said ClickHouse had roughly 46,000 GitHub stars and broad open-source adoption across user-facing analytics products by February 2026. | 中 | SV012 |
| CV012 | Independent coverage pegged ClickHouse’s May 2025 Series C at approximately $6.35 billion post-money. | 中 | SV003 |
| CV013 | A $6.35 billion valuation on $160 million ARR implies about 39.7x trailing ARR. | 中 | SV003, SV012 |
| CV014 | Using a $150 million to $185 million ARR underwriting range, the Series C valuation implies roughly 34.3x to 42.3x ARR. | 中 | SV003, SV012 |
| CV015 | Snowflake generated $4.68 billion of revenue in fiscal year 2026. | 高 | SV021, SV022 |
| CV016 | Snowflake’s market capitalization was about $61.55 billion in late May 2026. | 中 | SV019, SV020 |
| CV017 | Snowflake traded at roughly 13.1x revenue in May 2026 based on a $61.55 billion market cap and $4.68 billion FY2026 revenue. | 中 | SV019, SV020 |
| CV018 | Databricks announced a $5.4 billion revenue run-rate and an approximately $134 billion valuation in February 2026. | 高 | SV015, SV016 |
| CV019 | Databricks’ February 2026 financing implied an enterprise value to revenue multiple of about 24.8x. | 中 | SV015, SV016 |
| CV020 | Databricks disclosed more than 65% year-over-year growth, net retention above 140%, and more than 20,000 organizations on the platform. | 中 | SV015, SV018 |
| CV021 | SingleStore reported ARR above $123 million in Q2 fiscal 2026, up 23% year over year. | 高 | SV026, SV028 |
| CV022 | SingleStore ended Q2 fiscal 2026 with more than $150 million in cash, zero debt, and free cash flow nearly breakeven over the prior twelve months. | 高 | SV026, SV028 |
| CV023 | Tracxn lists SingleStore’s last known valuation at $1 billion as of October 3, 2022. | 中 | SV027 |
| CV024 | A $1 billion valuation on $123 million ARR implies an approximate 8.1x ARR multiple for SingleStore. | 中 | SV026, SV027 |
| CV025 | MongoDB generated $2.46 billion of revenue in fiscal year 2026. | 中 | SV023 |
| CV026 | MongoDB’s market capitalization was about $24.74 billion in late May 2026. | 中 | SV023, SV024 |
| CV027 | MongoDB traded at roughly 10.0x revenue in May 2026. | 中 | SV023, SV024 |
| CV028 | DB-Engines ranked MongoDB fifth, Snowflake sixth, and ClickHouse twenty-sixth in its May 2026 popularity table, highlighting ClickHouse’s smaller installed-base footprint. | 中 | SV025 |
| CV029 | ClickHouse’s implied Series C multiple sits above Snowflake’s ~13x, MongoDB’s ~10x, and SingleStore’s ~8x, and even above Databricks’ ~25x despite materially smaller scale. | 中 | SV003, SV015, SV019, SV023, SV026, SV027 |
| CV030 | The gap between ClickHouse’s ~40x trailing ARR and Databricks’ ~25x suggests investors were underwriting extraordinary forward growth rather than current scale parity. | 中 | SV012, SV015, SV016 |
| CV031 | At Snowflake’s ~13.1x revenue multiple, ClickHouse would need roughly $485 million of ARR or revenue to justify a $6.35 billion valuation. | 中 | SV003, SV019, SV020 |
| CV032 | At Databricks’ ~24.8x multiple, ClickHouse would need roughly $256 million of ARR or revenue to justify $6.35 billion. | 中 | SV003, SV015, SV016 |
| CV033 | At MongoDB’s ~10.0x multiple, ClickHouse would need roughly $632 million of ARR or revenue to justify $6.35 billion. | 中 | SV003, SV023, SV024 |
| CV034 | Open-source distribution gives ClickHouse a premium narrative because free adoption can feed ClickHouse Cloud, but public open-source comp MongoDB shows the premium normalizes near ~10x once scale matures. | 中 | SV011, SV012, SV023, SV024 |
| CV035 | ClickHouse’s AI and observability customer mix shows real demand from sophisticated users, but those logos do not yet prove public-company durability on retention, margins, or workload concentration. | 中 | SV001, SV011 |
| CV036 | Reuters noted that software stocks were under pressure in 2026 over fears that fast-moving AI could disrupt software economics, reinforcing public-market multiple compression risk. | 中 | SV016 |
| CV037 | Because ClickHouse’s public revenue evidence is third-party-estimated rather than audited, the Series C price has less objective support than Databricks’ official run-rate disclosure or Snowflake’s filing-backed revenue base. | 中 | SV012, SV015, SV022 |
| CV038 | A bullish underwriting case requires ClickHouse to keep converting open-source adoption into enterprise cloud ARR and to cross roughly $300 million ARR quickly so the current mark moves from ~40x trailing to ~20x forward. | 中 | SV012, SV015, SV019 |
| CV039 | A bearish underwriting case assumes growth slows toward public and open-source comps, which could compress valuation toward a low-teens revenue multiple and materially below the Series C mark. | 中 | SV016, SV019, SV023 |
| CV040 | SingleStore’s near-breakeven profile at a much lower implied multiple shows investors pay very different prices for real-time database vendors once hypergrowth cools. | 中 | SV026, SV027, SV028 |
| CV041 | Databricks’ richer valuation is tied not only to growth but also to broader platform breadth, more than 20,000 organizations, and a disclosed retention profile, all of which ClickHouse has not matched publicly. | 中 | SV015, SV018 |
| CV042 | ClickHouse’s company materials argue that real-time analytics and lower cost per query versus legacy warehouses are central to the product moat. | 中 | SV009, SV010 |
| CV043 | The available evidence supports a track recommendation: product-market pull and open-source distribution are strong, but the May 2025 price already capitalized much of the next leg of execution. | 中 | SV003, SV012, SV015, SV019 |
| CV044 | A move from track to buy would require audited financials or management disclosure on ARR quality, retention, gross margin, and the economics of converting large open-source users to paid cloud. | 中 | SV012, SV022 |
| CV045 | Thesis-break triggers are a sharp deceleration below the growth needed for ~$300 million ARR, failure to close enterprise feature gaps, or public comps de-rating further from the current 10x to 25x range. | 中 | SV016, SV019, SV023 |
| 编号 | 出版方 | 标题 | 引文 |
|---|---|---|---|
| SO001 | ClickHouse | Our Story - ClickHouse | Work on ClickHouse began in 2009 ... 2016 ClickHouse releases as an open-source project under the Apache 2 license ... 2021 ClickHouse, Inc. incorporates in Delaware, with our headquarters in the San Francisco Bay Area. |
| SO002 | ClickHouse | We Stand With Ukraine | ClickHouse, Inc. is a Delaware company with headquarters in the San Francisco Bay Area. We have no operations in Russia, no Russian investors, and no Russian members of our Board of Directors. |
| SO003 | GitHub | GitHub - ClickHouse/ClickHouse: ClickHouse® is a real-time analytics database management system | |
| SO004 | Business Wire | ClickHouse Raises $250M Series B To Scale Groundbreaking OLAP Database Management System Globally | ClickHouse ... has raised $250 million in Series B funding at a $2 billion valuation. The investment was led by Coatue and Altimeter, with participation from Index Ventures, Benchmark, Lightspeed, Redpoint, Almaz, Yandex N.V., FirstMark and Lead Edge. |
| SO005 | Business Wire | ClickHouse Raises $350 Million Series C to Power Analytics for the AI Era | ClickHouse ... has raised $350 million in Series C financing. The round was led by Khosla Ventures ... Today's round follows earlier investments of over $300 million, bringing total funding to over $650 million. |
| SO006 | FirstMark | ClickHouse Raises $350 Million Series C to Power Analytics for the AI Era | |
| SO007 | Coatue | Our Partnership with ClickHouse: Powering Analytics for the AI Era | |
| SO008 | Goodwin | Goodwin Advises ClickHouse on $350 Million Series C Financing and Extension to Accelerate Growth in Real-Time Analytics for the AI Era | |
| SO009 | Silicon Valley Daily | ClickHouse Clicks With $350 Million Series C Round | |
| SO010 | Business Wire via Ritzau | ClickHouse Extends Series C Financing and Expands Leadership Team to Fuel Growth | Business Wire | The company bolstered its leadership team with three key hires. In July, Kevin Egan joined as Chief Revenue Officer ... In August, Mariah Nagy came on board as Vice President of People ... Jimmy Sexton joined as Chief Financial Officer. |
| SO011 | Sacra | ClickHouse at $160M ARR | Sacra estimates ClickHouse hit $160M ARR by the end of 2025, growing 256% year-over-year. |
| SO012 | Forbes | ClickHouse | Company Overview & News | ClickHouse creator Alexey Milovidov cofounded the Portola Valley, California-based company in 2021 with ex-Salesforce exec Aaron Katz and ex-Netflix exec Yury Izrailevsky. |
| SO013 | Craft | ClickHouse CEO and Key Executive Team | Craft.co | |
| SO014 | Unify | Employee Data and Trends for ClickHouse | Unify | |
| SO015 | JFrog | 7 RCE and DoS vulnerabilities Found in ClickHouse DBMS | The JFrog Security research team ... discovered seven new security vulnerabilities in ClickHouse DBMS. |
| SO016 | Ubuntu | USN-6933-1: ClickHouse vulnerabilities | Ubuntu security notices | Ubuntu | It was discovered that ClickHouse incorrectly handled memory, leading to a heap-based buffer overflow ... or execute arbitrary code. (CVE-2021-43305) |
| SO017 | DEV Community | Lessons Learned #2: Your new feature could introduce a security vulnerability to your old feature (Clickhouse CVE-2024-22412) | |
| SO018 | Tracxn | ClickHouse - Funding Rounds & List of Investors | |
| SO019 | PitchBook News | Big Data wars: ClickHouse has a playbook to beat Snowflake at its own game | This past May, it raised a mammoth $350 million round led by Khosla Ventures ... and the company just hit $100 million in annualized revenue. |
| SO020 | Lightspeed Venture Partners | ClickHouse | |
| SO021 | Index Ventures | The Fast and the Furious: How ClickHouse, the world's fastest open-source database, is creating the first real-time data warehouse | In August of 2021, they announced the incorporation of ClickHouse, Inc., along with $50 million in Series A funding led by Index Ventures and Benchmark. |
| SO022 | Colorado Department of State via OpenGovCO | ClickHouse, Inc. · 4113 Alpine Rd, Portola Valley, CA 94028 | The entity was formed on August 25, 2021 in the jurisdiction of Delaware. The registered office location is at 4113 Alpine Rd, Portola Valley, CA 94028. |
| SO023 | Craft | ClickHouse Corporate Headquarters, Office Locations and Addresses | Craft.co | |
| SO024 | PitchBook | ClickHouse 2026 Company Profile: Valuation, Funding & Investors | PitchBook | ClickHouse is headquartered in San Francisco, CA. ClickHouse has 531 total employees. |
| SO025 | Tracxn | ClickHouse Company Profile | ClickHouse has 569 employees as of Apr 26. |
| SO026 | AIM Media House | ClickHouse Raises $350M at $6.35B: Future of Analytics | ClickHouse has now confirmed a $350 million Series C at a valuation of $6.35 billion. |
| SO027 | Index Ventures | Index Ventures: Mike Volpi | Index Ventures | He's currently serving on the boards of Aurora, ClickHouse, Cockroach Labs, Cohere, Confluent, Covariant.ai, Kong, Scale, Sonos, and Wealthfront. |
| SO028 | Forbes | Peter Fenton | He currently serves on the boards of Airtable, ClickHouse, Cockroach Labs, Docker, CarbonDrop, Mercor, Sorare, Timescale, and Wildlife Studios. |
| SM001 | ClickHouse | Fast Open-Source OLAP DBMS - ClickHouse | ClickHouse is a fast open-source column-oriented database management system that allows generating analytical data reports in real-time using SQL queries. |
| SM002 | ClickHouse Docs | Deployment modes | ClickHouse Docs | ClickHouse Server can be installed locally, deployed to AWS GCP or Azure, or run on on-premises hardware; ClickHouse Cloud is the fully managed deployment mode. |
| SM003 | ClickHouse | ClickHouse Cloud | Cloud Based DBMS | ClickHouse | ClickHouse Cloud is the fastest, most cost-efficient way to build real-time analytics, observability, and AI-powered data applications, and is available on all three major cloud marketplaces. |
| SM004 | ClickHouse | ClickStack: High-Performance Open Source Observability | Logs, Metrics, Traces with ClickHouse | ClickStack is open source observability for OpenTelemetry at scale with sub-second queries, 10-100x cost savings, and logs metrics traces and session replays powered by ClickHouse. |
| SM005 | GitHub | GitHub - ClickHouse/ClickHouse: ClickHouse® is a real-time analytics database management system | |
| SM006 | DB-Engines | ClickHouse System Properties | ClickHouse is a high-performance column-oriented SQL DBMS for OLAP and is available as both open-source software and a cloud offering. |
| SM007 | ClickHouse | Welcome to the ClickHouse Community | The ClickHouse community page reports 12k+ Slack members, 2.9k+ contributors, 29k+ PRs, 796 releases, and 47.6k+ GitHub stars. |
| SM008 | ClickHouse Docs | ClickHouse adopters | ClickHouse Docs | A list of companies using ClickHouse and their success stories. |
| SM009 | ClickHouse | ClickHouse Pricing | ClickHouse pricing automatically scales compute up and down, scales unused resources down to zero, and separates storage and compute. |
| SM010 | ClickHouse | Real-time Analytics with ClickHouse | ClickHouse highlights continuous ingest, high query concurrency, and low-latency analytics for interactive apps and dashboards. |
| SM011 | ClickHouse | Data warehousing with ClickHouse | ClickHouse positions itself as a real-time data warehouse for BI with faster queries at a fraction of the cost. |
| SM012 | Mordor Intelligence | Cloud Data Warehouse Market Share & Size 2031 Outlook | The Cloud Data Warehouse Market worth USD 14.94 billion in 2026 is growing at a CAGR of 26.86% to reach USD 49.12 billion by 2031. |
| SM013 | Research and Markets | Cloud Data Warehouse Market Report 2026 - Research and Markets | The Cloud Data Warehouse Market, valued at USD 14.53B in 2026, is projected to reach USD 31.7B by 2030, growing at a 21.5% CAGR. |
| SM014 | MarketsandMarkets | Cloud Data Warehouse Market Share, Forecast | Growth Analysis & Opportunities | The global market for cloud data warehouse is categorized by application, vertical, deployment model, type, and region. |
| SM015 | IndustryARC | Cloud Data Warehouse Market size, Industry outlook, Market forecast, Demand Analysis, Market Share, Market Report 2021-2026 | Cloud Data Warehouse Market is forecast to reach $39.1 billion by 2026 after growing at a CAGR of 31.4% during 2021-2026. |
| SM016 | Grand View Research | Streaming Analytics Market Size | Industry Report, 2030 | The global streaming analytics market was valued at USD 23.4 billion in 2023 and is projected to reach USD 128.4 billion by 2030 at a CAGR of 28.3%. |
| SM017 | Grand View Research | Observability Tools And Platforms Market Size Report, 2030 | The global observability tools and platforms market size was estimated at USD 2.71 billion in 2023 and is projected to reach USD 5.40 billion by 2030 at a CAGR of 10.7%. |
| SM018 | MarketsandMarkets | Observability Tools and Platforms Market Size & Trends, Growth Analysis, Industry Forecast [2030] | The global observability tools and platforms market is projected to grow from USD 2.4 billion in 2023 to USD 4.1 billion by 2028 at a CAGR of 11.7%. |
| SM019 | Mordor Intelligence | Observability Market Size, Report, Share & Competitive Landscape 2031 | The Observability Market worth USD 3.35 billion in 2026 is growing at a CAGR of 15.62% to reach USD 6.93 billion by 2031. |
| SM020 | Google Cloud | BigQuery | AI data platform | EDW | BigQuery is Google Cloud's fully managed and completely serverless enterprise data warehouse with real-time analytics, built-in AI, and decoupled storage and compute. |
| SM021 | Datadog | Infrastructure & Application Monitoring as a Service | Datadog | Datadog presents a unified observability platform that aggregates metrics events logs and traces and supports real-time interactive dashboards. |
| SM022 | Datadog | Pricing | Datadog | Datadog pricing breaks observability into ingest, indexing, storage, archiving, and AI or LLM observability products, underscoring how buyers manage telemetry cost by tier. |
| SM023 | Elastic | Full-stack observability solution — built on the Elasticsearch Platform | Elastic describes an AI-powered OpenTelemetry-first observability platform with best-in-class efficiency for logs and metrics and one platform for everything. |
| SM024 | Amazon Web Services | Open Source Search Engine - Amazon OpenSearch Service - AWS | Amazon OpenSearch Service simplifies AI-powered search, observability, and vector database operations with both managed clusters and serverless deployment. |
| SM025 | Grafana Labs | 2026 observability trends and predictions from Grafana Labs | Grafana Labs | Grafana argues that in 2026 unified observability becomes the default operating model, data value overtakes data volume, AI becomes a collaborator, and OpenTelemetry becomes the default. |
| SM026 | IBM | Observability Trends 2026 | IBM | IBM argues that 2026 observability strategies must become more intelligent, cost-effective, and compatible with open standards as AI adoption grows. |
| SM027 | Altinity | Ecosystem Projects | Altinity describes itself as the second-largest contributor to ClickHouse and highlights open-source tools including the Kubernetes operator, clickhouse-backup, and a Grafana plugin. |
| SP001 | ClickHouse | Real-Time Data Analytics Platform | ClickHouse | The fastest open-source analytical database. |
| SP002 | ClickHouse | ClickHouse Pricing | We scale storage and compute separately, due to our flexible architecture. |
| SP003 | ClickHouse | Deployment modes | ClickHouse Docs | ClickHouse Server can be installed on your local machine... deployed to major cloud providers... or set up on your own on-premises hardware. |
| SP004 | ClickHouse | ClickHouse Cloud | Cloud Based DBMS | ClickHouse | ClickHouse Cloud: the fastest, most cost-efficient way to build real-time analytics, observability, and AI-powered data applications. |
| SP005 | ClickHouse | Our Story - ClickHouse | 2021: ClickHouse, Inc. incorporates in Delaware... |
| SP006 | Business Wire | ClickHouse Raises $350 Million Series C to Power Analytics for the AI Era | ClickHouse, Inc. ... has raised $350 million in Series C financing. |
| SP007 | ClickHouse | ClickHouse benchmarks: Performance, cost & scalability compared | All proven by benchmarks that can be reproduced by anyone. |
| SP008 | DB-Engines | ClickHouse System Properties | A high-performance, column-oriented SQL DBMS for online analytical processing. |
| SP009 | Snowflake | The Snowflake Platform | Snowflake supports multi-cloud and cross-region operations. |
| SP010 | Snowflake | Snowflake - Investor Relations | 790 Forbes Global 2000 Customers ... 733 $1M+ Customers ... 125% Net Revenue Retention Rate. |
| SP011 | Snowflake | Understanding overall cost | Snowflake Documentation | The total cost of using Snowflake is the aggregate of the cost of using data transfer, storage, and compute resources. |
| SP012 | Snowflake | Overview of warehouses | Snowflake Documentation | Snowflake utilizes per-second billing (with a 60-second minimum each time the warehouse starts). |
| SP013 | U.S. Securities and Exchange Commission | EDGAR Entity Landing Page | |
| SP014 | Databricks | Data Lakehouse Architecture | Databricks | One architecture for integration, storage, processing, governance, sharing, analytics and AI. |
| SP015 | Databricks | Databricks Pricing: Flexible Plans for Data and AI Solutions | The Price List displays Databricks' undiscounted price for each SKU. |
| SP016 | Databricks | About Databricks: The data and AI company | More than 20,000 organizations worldwide ... and 70% of the Fortune 500 rely on the Databricks Data Intelligence Platform. |
| SP017 | Databricks | What Is a Lakehouse? | A lakehouse is a new, open architecture that combines the best elements of data lakes and data warehouses. |
| SP018 | Google Cloud | BigQuery | AI data platform | EDW | BigQuery is the autonomous data to AI platform. |
| SP019 | Google Cloud | BigQuery | BigQuery is a serverless data analytics platform. |
| SP020 | AWS | Cloud Data Warehouse - Amazon Redshift - AWS | Amazon Redshift is built on cloud economics that scale with your usage. |
| SP021 | AWS | Amazon Redshift Pricing | Redshift Provisioned starts at $0.543 per hour, while Redshift Serverless begins at $1.50 per hour. |
| SP022 | AWS | Interactive SQL - Amazon Athena - AWS | Get streamlined, near-instant startup of SQL or Apache Spark analytics workloads with a serverless experience. |
| SP023 | AWS | Amazon Athena Pricing | Pricing is simple: you pay based on data processed or compute used. |
| SP024 | DuckDB Foundation | Why DuckDB | DuckDB does not run as a separate process, but completely embedded within a host process. |
| SP025 | StarRocks | StarRocks | A High-Performance Analytical Database | One Engine for Real-Time, Lakehouse, and AI. |
| SP026 | Apache Druid | Apache Druid | Apache® Druid | A high performance, real-time analytics database that delivers sub-second queries on streaming and batch data at scale and under load. |
| SP027 | Imply | Imply Enterprise - Imply | Commercial distribution of Druid. |
| SP028 | Imply | Imply Database as a Service Cost | Real-Time Analytics Database-a-Service Cost | Starter ... Starts at $100/month. Standard ... Starts at $600/month. |
| SP029 | SingleStore | Product Overview | SingleStore Helios cloud service | SingleStore Helios is a cloud database-as-a-service available on leading public clouds. |
| SP030 | SingleStore | SingleStore Pricing | The cost of SingleStore is determined by actual usage. |
| SP031 | SingleStore | About SingleStore | SingleStore brings you the world’s fastest distributed SQL database for real-time applications and analytics. |
| SP032 | SingleStore | Deploy · SingleStore Self-Managed Documentation | SingleStore can be deployed on bare metal, on virtual machines, or in the cloud. |
| SP033 | Alphabet | Alphabet Investor Relations - Investors | |
| SP034 | Amazon | Annual reports, proxies and shareholder letters | |
| SI001 | ClickHouse | ClickHouse Cloud | Cloud Based DBMS | ClickHouse | |
| SI002 | ClickHouse | ClickHouse Pricing | |
| SI003 | ClickHouse Docs | Cloud changelog - 2022 | ClickHouse Docs | |
| SI004 | ClickHouse | ClickHouse Extends Series C Financing and Expands Leadership Team to Fuel Growth | |
| SI005 | Yahoo Finance | ClickHouse Raises $350 Million Series C to Power Analytics for the AI Era | |
| SI006 | FirstMark | ClickHouse Raises $350 Million Series C to Power Analytics for the AI Era | |
| SI007 | FinSMEs | ClickHouse Raises $350M in Series C Funding | |
| SI008 | TechCrunch | Snowflake, Databricks challenger ClickHouse hits $15B valuation | TechCrunch | |
| SI009 | Sacra | ClickHouse revenue, funding & news | |
| SI010 | Business Wire | ClickHouse Raises $250M Series B To Scale Groundbreaking OLAP Database Management System Globally | |
| SI011 | Business Wire | Announcing ClickHouse Cloud: Democratizing lightning-fast insights and analytics | |
| SI012 | TechCrunch | ClickHouse launches ClickHouse Cloud, extends its Series B | TechCrunch | |
| SI013 | Goodwin | Goodwin Advises ClickHouse on $350 Million Series C Financing and Extension to Accelerate Growth in Real-Time Analytics for the AI Era | News & Events | Goodwin | |
| SI014 | m3ter | The ClickHouse Story with m3ter | |
| SI015 | App Developer Magazine | ClickHouse Cloud beta released on AWS | |
| SI016 | Snowflake via SEC EDGAR | EDGAR Search Results | |
| SI017 | MongoDB via SEC EDGAR | EDGAR Search Results | |
| SI018 | Confluent via SEC EDGAR | EDGAR Search Results | |
| SI019 | Elastic via SEC EDGAR | EDGAR Search Results | |
| SI020 | Business Wire | ClickHouse Extends Series C Financing and Expands Leadership Team to Fuel Growth | |
| SI021 | ClickHouse Docs | ClickHouse Docs | ClickHouse Docs | |
| SI022 | ClickHouse Docs | Changelog 2026 | ClickHouse Docs | |
| SI023 | ClickHouse | Real-time Analytics with ClickHouse | |
| SI024 | ClickHouse | ClickStack: High-Performance Open Source Observability | Logs, Metrics, Traces with ClickHouse | |
| SI025 | ClickHouse | Data warehousing with ClickHouse | |
| SE001 | ClickHouse | ClickHouse Cloud | Cloud Based DBMS | ClickHouse | |
| SE002 | ClickHouse | Introduction | ClickHouse Docs | |
| SE003 | ClickHouse | Architecture overview | ClickHouse Docs | |
| SE004 | ClickHouse | Architecture Overview | ClickHouse Docs | |
| SE005 | ClickHouse | MergeTree table engine | ClickHouse Docs | |
| SE006 | ClickHouse | Shared | ClickHouse Docs | |
| SE007 | ClickHouse | Kafka table engine | ClickHouse Docs | |
| SE008 | ClickHouse | Integrating dbt and ClickHouse | ClickHouse Docs | |
| SE009 | ClickHouse | Integrations | ClickHouse Docs | |
| SE010 | ClickHouse | Cloud changelog - 2026 | ClickHouse Docs | |
| SE011 | ClickHouse | Monitoring Cloudflare logs with ClickStack | ClickHouse Docs | |
| SE012 | ClickHouse | ClickHouse adopters | ClickHouse Docs | |
| SE013 | ClickHouse | Trouble will find you: How Cloudflare uses ClickHouse to scale analytics at quadrillion-row scale | |
| SE014 | GitHub | GitHub - ClickHouse/ClickHouse: ClickHouse® is a real-time analytics database management system | |
| SE015 | GitHub | GitHub - ClickHouse/clickhouse-docs: Official documentation for ClickHouse | |
| SE016 | DB-Engines | ClickHouse System Properties | |
| SE017 | PyPI | clickhouse-connect · PyPI | |
| SE018 | npm | @clickhouse/client - npm | |
| SE019 | Docker | clickhouse/clickhouse-server - Docker Image | |
| SE020 | Microsoft Learn | Power Query ClickHouse connector - Power Query | |
| SE021 | Business Wire | ClickHouse Cloud Is Now Generally Available on Microsoft Azure | |
| SE022 | TrustRadius | ClickHouse Reviews & Ratings 2026 | TrustRadius | |
| SE023 | HypeQuery | Seven Companies, One Pattern: Why Every Scaled ClickHouse Deployment Looks the Same | |
| SE024 | ClickHouse | User stories - ClickHouse | |
| SE025 | ClickHouse | How ClickHouse powers Netflix, Uber and Spotify’s Analytics | Aaron Katz, CEO of ClickHouse | |
| SU001 | ClickHouse | User stories | |
| SU002 | ClickHouse | ClickHouse adopters | |
| SU003 | ClickHouse | Trouble will find you: How Cloudflare uses ClickHouse to scale analytics at quadrillion-row scale | A single query scanned 96 trillion events in an hour and returned in less than two seconds. |
| SU004 | ClickHouse | London Meetup Report: How Cloudflare processes hundreds of millions of rows per second with ClickHouse | This year we actually exceeded a thousand active replicas. That's processing hundreds of millions of inserted rows every second. |
| SU005 | Cloudflare | HTTP Analytics for 6M requests per second using ClickHouse | Cloudflare has grown tremendously... from under 1M requests per second to current levels of 6M requests per second. |
| SU006 | Cloudflare | Log analytics using ClickHouse | CPU and memory consumption on the inserter side were reduced by eight times. |
| SU007 | Cloudflare | Our billing pipeline was suddenly slow. The culprit was a hidden bottleneck in ClickHouse | This pipeline powers hundreds of millions of dollars in usage revenue, fraud systems, and more. |
| SU008 | ClickHouse | How ClickHouse powers Netflix, Uber and Spotify’s Analytics | |
| SU009 | HypeQuery | Seven Companies, One Pattern: Why Every Scaled ClickHouse Deployment Looks the Same | Uber's QueryBridge migration preserved 10,000+ Kibana dashboards with zero user retraining. |
| SU010 | ClickHouse | Contentsquare migration from Elasticsearch to ClickHouse | ClickHouse turned out to be 11 times cheaper and allowed us to have a 10x performance improvement in our p99 for queries. |
| SU011 | ClickHouse | Why OpenAI chose ClickHouse for petabyte-scale observability | Every day, the company ingests petabytes of log data... and that volume is growing by more than 20% each month. |
| SU012 | ClickHouse | How Anthropic is using ClickHouse to scale observability for the AI era | ClickHouse played an instrumental role in helping us develop and ship Claude 4. |
| SU013 | ClickHouse | How Tesla built a quadrillion-row-scale observability platform on ClickHouse | Over one quadrillion rows ingested—with not a single hiccup, not a single issue. |
| SU014 | Microsoft Clarity | Why Microsoft Clarity chose ClickHouse | Heat map generation became an instantaneous task to do, and it was even orders of magnitude cheaper to run. |
| SU015 | ClickHouse | Replo uses ClickHouse to power real-time merchant analytics | Replo, an AI-powered page builder trusted by more than 4,000 Shopify merchants... [is] capable of processing and analyzing more than 100 billion events. |
| SU016 | ClickHouse | Mintlify boosts NPS 30% and saves 60% with real-time analytics on ClickHouse Cloud | Dashboards that took tens of seconds to load in PostHog now return results in under a second. |
| SU017 | ClickHouse | How Padlet uses ClickHouse Cloud to power real-time classroom analytics | Padlet ingested roughly 8 billion events into ClickHouse in a single month. |
| SU018 | ClickHouse | How Buildkite transformed test analytics and cut costs with ClickHouse Cloud | For every dollar spent on ClickHouse, the team is saving eight dollars elsewhere. |
| SU019 | ClickHouse | Just OLAP it: How Ramp rebuilt its analytics platform on ClickHouse Cloud | When those customers tried to run reports... charts that once timed out after 40 seconds were returning in milliseconds. |
| SU020 | ClickHouse | Goodbye limitations, hello data: How Qonto is rethinking observability with ClickHouse Cloud | Qonto’s ResourceAttributes and SpanAttributes columns store 231 TB of uncompressed data in 376 GB. |
| SU021 | ClickHouse | How Langfuse is scaling LLM observability for the agentic era with ClickHouse | Compared to the old approach, the new model delivered around three times less memory usage and up to 20 times faster queries. |
| SU022 | ClickHouse | Lyft analytics on ClickHouse Cloud | Reading more than 450 terabytes of data per day and writing around 4 terabytes... |
| SU023 | ClickHouse | How Polymarket scales data with Postgres and ClickHouse | The API now handles 100s requests per second, with an average latency of approximately 25 milliseconds. |
| SU024 | ClickHouse | ClickHouse Cloud | Available on all three major cloud marketplaces. |
| SU025 | ClickHouse | Pricing philosophy | We automatically scale unused resources down to zero so that you don’t pay for idle services. |
| SU026 | TrustRadius | ClickHouse reviews | Managing roles in the Cloud could have more options... [and] it still misses SSO compatibility with some Identity Providers like Okta. |
| SU027 | PeerSpot | ClickHouse reviews | My experience with pricing, setup cost, and licensing indicates that it is very expensive—ClickHouse is the most expensive option. |
| SU028 | LeadCognition | Companies and developers using ClickHouse | ClickHouse is used in production at Cloudflare, Uber, eBay, Spotify... |
| SU029 | Uber Engineering | Uber Engineering blog homepage | |
| SU030 | Spotify Engineering | Spotify Engineering homepage | |
| SU031 | G2 | ClickHouse reviews | |
| SU032 | CaseStudies.com | ClickHouse B2B case studies and customer successes | |
| SR001 | Business Wire | ClickHouse Raises $350 Million Series C to Power Analytics for the AI Era | The company grew over 300% during the past year and now serves over 2,000 customers. |
| SR002 | Yahoo Finance | ClickHouse Raises $350 Million Series C to Power Analytics for the AI Era | |
| SR003 | Goodwin | Goodwin Advises ClickHouse on $350 Million Series C Financing and Extension to Accelerate Growth in Real-Time Analytics for the AI Era | News & Events | Goodwin | |
| SR004 | ClickHouse | ClickHouse Cloud | Cloud Based DBMS | ClickHouse | ClickHouse Cloud offers a serverless hosted DBMS solution. Automatic scaling and no infrastructure to manage at consumption-based pricing. |
| SR005 | ClickHouse | ClickHouse Pricing | |
| SR006 | ClickHouse Docs | Service uptime | ClickHouse Docs | |
| SR007 | ClickHouse Cloud Status | ClickHouse Cloud Status | |
| SR008 | ClickHouse Docs | Compliance overview | ClickHouse Docs | |
| SR009 | ClickHouse | ClickHouse Privacy Policy | |
| SR010 | ClickHouse | ClickHouse benchmarks: Performance, cost & scalability compared | |
| SR011 | ClickHouse | Real-Time Data Analytics Platform | ClickHouse | |
| SR012 | ClickHouse Docs | Deployment modes | ClickHouse Docs | |
| SR013 | ClickHouse | Real-time Analytics with ClickHouse | |
| SR014 | ClickHouse | Data warehousing with ClickHouse | |
| SR015 | ClickHouse | Welcome to the ClickHouse Community | |
| SR016 | ClickHouse Docs | ClickHouse adopters | ClickHouse Docs | |
| SR017 | GitHub | GitHub - ClickHouse/ClickHouse: ClickHouse® is a real-time analytics database management system | |
| SR018 | GitHub | Contributors to ClickHouse/ClickHouse | |
| SR019 | ClickHouse Docs | Security changelog | ClickHouse Docs | |
| SR020 | GitHub Security Advisory | CVE-2025-1385- Fail input validation in clickhouse-library-bridge API could lead to RCE | |
| SR021 | GitHub Security Advisory | Role-based Access Control is bypassed when query caching is enabled. | |
| SR022 | NIST National Vulnerability Database | NVD - CVE-2025-1385 | |
| SR023 | Altinity | Is ClickHouse® Moving Away from Open Source? | Important new features are available only in ClickHouse Cloud. |
| SR024 | Exasol | How 5 Databases Actually Scale across Concurrency, Data, and Nodes | |
| SR025 | Habr | StarRocks vs. ClickHouse, Apache Druid, and Trino | |
| SR026 | DB-Engines | DB-Engines Ranking | |
| SR027 | DB-Engines | ClickHouse System Properties | |
| SR028 | Chaos and Order | Database Engines 2026 Deep-Dive — Postgres Won the API, ClickHouse and DuckDB Won Analytics | |
| SR029 | TechCrunch | Snowflake, Databricks challenger ClickHouse hits $15B valuation | TechCrunch | |
| SR030 | Tinybird | Honest guide to the best ClickHouse® alternatives in 2026 | |
| SV001 | ClickHouse / BusinessWire | ClickHouse Raises $350 Million Series C to Power Analytics for the AI Era | |
| SV002 | Yahoo Finance | ClickHouse Raises $350 Million Series C to Power Analytics for the AI Era | |
| SV003 | AIM Research | ClickHouse Raises $350M at $6.35B: Future of Analytics | |
| SV004 | SiliconANGLE | ClickHouse reels in $350M for its high-speed columnar database | |
| SV005 | FinSMEs | ClickHouse Raises $350M in Series C Funding | |
| SV006 | ClickHouse | ClickHouse Pricing | |
| SV007 | ClickHouse | ClickHouse Cloud | Cloud Based DBMS | ClickHouse | |
| SV008 | ClickHouse Docs | What is ClickHouse? | ClickHouse Docs | |
| SV009 | ClickHouse | Snowflake vs ClickHouse | |
| SV010 | ClickHouse | Real-time Analytics with ClickHouse | |
| SV011 | Sacra | ClickHouse revenue, funding & news | |
| SV012 | Sacra | ClickHouse at $160M ARR | |
| SV013 | FirstMark | ClickHouse Raises $350 Million Series C to Power Analytics for the AI Era | |
| SV014 | MarketScreener | ClickHouse, Inc. announced that it has received $350 million in funding from a group of investors | |
| SV015 | Databricks | Databricks Grows >65% YoY, Surpasses $5.4 Billion Revenue Run-Rate, Doubles Down on Lakebase and Genie | |
| SV016 | Reuters / U.S. News | Databricks Raises $5 Billion in Latest Funding, Defying Software Selloff | |
| SV017 | CNBC | Databricks raises capital at $134 billion valuation in latest funding round | |
| SV018 | Sacra | Databricks revenue, valuation & funding | |
| SV019 | Stock Analysis | Snowflake (SNOW) Revenue 2019-2026 | |
| SV020 | CompaniesMarketCap | Snowflake (SNOW) - Market capitalization | |
| SV021 | Nasdaq | Snowflake Reports Financial Results for the Fourth Quarter and Full-Year of Fiscal 2026 | |
| SV022 | Securities and Exchange Commission | Snowflake 10-K for fiscal year ended January 31, 2026 | |
| SV023 | Stock Analysis | MongoDB (MDB) Revenue 2016-2026 | |
| SV024 | CompaniesMarketCap | MongoDB (MDB) - Market capitalization | |
| SV025 | DB-Engines | DB-Engines Ranking | |
| SV026 | SingleStore | SingleStore Delivers Record Performance in the Second Quarter of Fiscal Year 2026 | |
| SV027 | Tracxn | SingleStore | |
| SV028 | StorageNewsletter | SingleStore Delivers Record Performance in the Second Quarter of Fiscal Year 2026 | |
| SV029 | Owler | SingleStore’s Competitors, Revenue, Number of Employees, Funding, Acquisitions & News - Owler Company Profile | |
| SV030 | ClickHouse | Learn about the latest ClickHouse tips, tricks and company announcements. |