初创公司尽调
尽调报告 AI Infrastructure / Enterprise Storage late-stage private 2026-05-27

VAST Data

高增长 AI 数据平台厂商,技术栈差异化明显,$30B 估值已经跑在当前公开披露之前

VAST Data 确实像 AI 基础设施里的赢家:规模真实、生态验证强,数据平台技术也有差异化;但当前 $30 billion 估值仍需先做收入质量、客户集中度和股权结构条款的私下尽调,新资金才有把握承销。

封面要素

Series F 轮估值 01
30000 USD M [CV001]
Series F 轮交易规模 02
1000 USD M [CV001]
承诺 ARR 下限 03
500 USD M+ [CV007]
累计订单额 04
4000 USD M+ [CV008]
披露员工数 05
700 employees+ [CO025]

公司概况

VAST Data 是一家 2016 年创立的私营 AI 基础设施公司,已经从高性能全闪存储平台扩展为它现在所称的 AI Operating System。该平台在 DASE 架构下组合 DataStore、DataBase、DataEngine 和 DataSpace,为企业、新云厂商和主权 AI 工作负载统一存储、数据库与计算。公开证据支撑了有意义的规模:VAST 披露累计订单额超过 $4 billion,上一财年退出时承诺 ARR 超过 $500 million,并且在 2026 年 Series F 前已经实现经营利润率和自由现金流为正。公司目前私募估值为 $30 billion,客户、OEM 和 NVIDIA 相关部署都给出强生态验证,但它对客户集中度、精确收入定义和完全稀释经济条款的公开披露仍然有限。

官网
vastdata.com
成立时间
2016-01-01
创始人
Renen Hallak, Jeff Denworth
创立地点
Tel Aviv, Israel
总部
New York, NY, USA
产品
VAST 销售基于 DASE 架构的 AI Operating System,组合通用文件 / 对象 / 块存储、事务型和分析型数据湖、流式与事件处理服务、全局命名空间,以及面向大规模企业和新云厂商环境的伙伴打包 AI 基础设施。
客户
大型企业、AI 云、研究机构、主权 AI 运营方,以及运行数据密集型 AI、分析、媒体、医疗和 HPC 工作负载的其他组织。
商业模式
通过直销和 OEM 主导的企业渠道销售软件订阅与平台许可,并配套经过验证的硬件包和会随时间扩展的长期基础设施部署。
阶段
late-stage private
融资情况
2026 年 4 月完成约 $1 billion 的 Series F 交易,估值 $30 billion;此前 2023 年完成 $118 million Series E,估值 $9.1 billion。公开证据显示,最新轮次混合了新股资本和老股流动性。
[CO001, CO002, CO003, CO006, CO025, CO027, CO028, CE001]

执行摘要

主要优势

  • 统一的 AI Operating System / DASE 架构让 VAST 拿出差异化的全栈数据层,而不是讲一个商品化阵列故事。
  • 对一家私有基础设施厂商来说,公开披露的商业进展很强:bookings 超过 $4B,committed ARR 超过 $500M,经营利润率和自由现金流均为正,并签下 $1.17B 的 CoreWeave 协议。
  • 客户、OEM 和 NVIDIA 周边生态验证覆盖 AI 云、电信、媒体、医疗、科研和主权 AI 部署。

主要风险

  • 估值达到 $30B 后,除非真实经常性收入明显更接近第三方高端估计、而不是公司自己披露的 ARR 下限,否则相对公开存储可比公司仍偏贵。
  • 公开证据仍未打通 ARR 定义,也没有披露头部客户集中度、续约质量或 Series F 完全摊薄条款。
  • VAST 向平台扩张后,对 NVIDIA、Cisco、OEM 硬件伙伴以及 CoreWeave 等旗舰 AI 云客户的依赖更高。
  • Red Stapler 相关法律 / IP 阴影,以及以色列相关运营连续性、隐私和监管暴露,仍是真实尽调项。

未决问题

  • bookings、committed ARR、projected ARR 以及任何更宽口径 ARR-like 收入结构之间的经审计桥接。
  • CoreWeave 和 neocloud 建设之后的头部客户 ARR、续约队列和集中度暴露。
  • Series F 后股权结构表、清算优先权、期权池影响,以及新股 / 老股交易的精确拆分。
  • 当前员工数、安全 / 合规认证,以及可独立验证的可靠性 / SLA 指标。

目录

Chapter 01

01公司概况

1.1 身份、阶段与产品范围

VAST Data 是一家 2016 年创立的私营 AI 基础设施公司,如今它把自己定位为面向数据密集型应用的 “AI Operating System”,而不是单点存储厂商。留存来源中的公司主线方向一致:VAST 称自己把存储、数据库和计算统一到一个数据平台里;独立报道则把同一转型描述为三阶段,先从面向闪存的存储起步,再进入数据库服务,最后走向更宽的 AI 操作系统编排。产品材料把 VAST DataStore 描述为通用数据存储,既去掉传统存储分层,又能从 TB 扩到 EB;Supermicro 的合作伙伴材料则把该平台放在企业 AI 和 GPU 云部署的数据管理层。总部披露不如产品范围清楚。2023–2026 年官方新闻材料交替使用 “No-Headquarters” 或 “Remote-First Company”,同时把公告锚定在纽约市;Reuters 描述的报道又称公司总部位于纽约。因此,后续章节最稳妥的事实底座是:VAST 是一家远程优先、以纽约为锚点的私营公司,商业界面面向美国,后续地点报道也显示其在以色列有运营根基。[CO001, CO002, CO003, CO004, CO005, CO006]

VAST Data 快照 KPI 表
指标数值 / 状态日期置信度缺口 / 注意事项
成立时间20162016事实稳定;完整创始人名单的佐证不如成立年份清晰
总部 / 运营模式远程优先,官方稿件以纽约市为锚点;Reuters 报道称总部位于纽约2025-08 to 2026-04官方材料没有披露传统意义上的单一总部声明
当前阶段私营、后期、Series F 后2026-04-22IPO 准备已公开讨论,但正式 IPO 流程尚未启动
一句话产品 / 模式统一存储、数据库和计算的 AI 操作系统;通过软件、OEM、伙伴和渠道路径销售2026-05-27留存来源中没有公开价格表
Series F 估值$30B2026-04-22私营估值,未经外部审计
大致累计融资Series F 后隐含 ~$1.38B2026-04-22由 F 轮前累计融资加报道的 F 轮交易规模推算
当前 ARR / CARR 披露> $500M 上一年度承诺 ARR / CARR2026-04-22公司口径;无经审计收入报表
盈利信号据报道经营利润率、自由现金流和 GAAP 盈利均为正2026-04-22私营公司自披露,未经审计
员工规模2023 年末披露 >700 名员工2023-12-06留存来源不支持精确 2026 员工数
客户规模公司称有数千家组织;具名客户标识包括 CoreWeave、Booking、Zoom、Pixar、U.S. Air Force、JPMorganChase 等2023-12 to 2026-04付费客户精确数量未公开披露
办公地点据报道覆盖美国、以色列和伦敦;2023 年报道称扩张至亚太、中东和欧洲2023-12-08留存来源未完整公布 2026 年最新办公室清单

该表混合公司披露、独立报道和一项推算融资总额;不受支持的 2026 员工数、董事会、债务和精确客户数字段以缺口呈现,而不是编造数值。

[CO001, CO003, CO006, CO007, CO008, CO015]
FO002: VAST Data 公司快照逻辑

VAST 的身份、平台架构、客户证明、资本结构和伙伴渠道如何串起当前投资叙事。

[CO002, CO003, CO005, CO014, CO015, CO035]
FO003: VAST Data KPI 披露快照

从披露质量看,哪些头部 KPI 是最新、历史、估计或仍然缺失;即使 VAST 私募估值很高,缺口仍在。

[CO020, CO023, CO024, CO031, CO032, CO034]

1.2 创始人、领导层与治理信号

Renen Hallak 是留存记录中最清晰的领导连续性锚点。无论官方融资稿、产品定位还是媒体采访,他都以创始人兼 CEO 身份出现,因此也是后续尽调最重要、可复用的身份锚点。Jeff Denworth 也被直接证实为联合创始人,并且仍是公司和第三方材料中可见的产品发言人,尤其围绕从统一数据平台走向完整 AI 操作系统的演进。财务负责人现在也变得关键:Amy Shapero 是 VAST 首任 CFO,此前担任 Shopify CFO;CRN 和 Reuters 报道都把她与 IPO 准备度联系起来。这个组合说明公司治理栈在向公开市场纪律成熟,尽管公司尚未启动正式 IPO 流程。云业务领导层的信号更复杂。Jonsi Stefansson 通过 Red Stapler 人才收购加入,并在 Polaris 发布时被 VAST 公开引用,但同一高管仍牵连 NetApp 诉讼风险。公开治理披露比融资披露更薄:留存来源确认 Fidelity 有董事会观察员角色,但没有给出清晰、最新的完整董事会名单;这是需要延续的尽调缺口,不是编造治理细节的理由。[CO009, CO010, CO011, CO012, CO013, CO014]

领导层与创始人表
人物职务 / 状态有证据支持的背景重要性关键人依赖
Renen Hallak创始人兼 CEO在官方融资稿和长篇媒体访谈中反复以创始人 / CEO 及首席战略声音出现牵引公司身份认知、资本叙事和产品定位关键 —— 融资、平台边界和 IPO 准备度的公开代表
Jeff Denworth联合创始人、公开产品发言人在第三方报道中解读统一平台和 Thinking Machine 愿景把架构故事接到商业化叙事和产品演进上高 —— 反复代表联合创始人解读平台路线图
Amy Shapero首任 CFO来自 Shopify,公开与 IPO 准备绑定财务控制与上市公司准备度已成为战略议题高 —— 如果 VAST 走向公开市场,财务负责人是核心
Jonsi Stefansson云解决方案总经理曾任 NetApp CTO,随 Red Stapler 人才收购加入代表云服务扩张和多云运营能力中 —— 战略上有用,但云业务角色被诉讼背景笼罩

仅为部分公开领导层图谱;留存来源未提供完整现任董事会名单或完整高管组织架构。

[CO009, CO010, CO011, CO012, CO013, CO014]

1.3 融资历史、估值与公开规模指标

VAST 的融资曲线即使放在 AI 基础设施里也异常陡峭。公司从 2021 年 5 月 $83 million Series D、估值 $3.7 billion,走到 2023 年 12 月 $118 million Series E、估值 $9.1 billion,再到 2026 年 4 月约 $1 billion Series F 交易、估值 $30 billion。Series F 的意义不只在价格,还有三点。第一,轮次同时包含新股和老股,因此部分资金是在兑现既有持有人,而不只是给资产负债表补弹药。第二,财团现在既有 Drive Capital、Access Industries、Fidelity、NEA 等财务投资人,也有 NVIDIA 作为战略参与方。第三,公司把估值上跳和罕见强的自称经营指标绑在一起:累计订单额超过 $4 billion、上一年承诺 ARR 超过 $500 million、经营利润率和自由现金流为正。这些信号有力,但仍不能替代经审计财务报表。公开规模披露并不均衡。2023 年末来源支持员工超过 700 人,并在美国、以色列和伦敦设有办公室;2026 年精确员工数、客户数以及任何债务或信贷工具,在留存来源包中仍没有支撑。[CO016, CO017, CO018, CO019, CO020, CO021]

利益相关方或投资人图谱
利益相关方角色控制权或经济重要性尽调问题
Drive CapitalSeries F 领投方领投 2026 年融资,确立当前 $30B 估值标记F 轮附带哪些权利、治理影响力或业绩里程碑?
Access IndustriesSeries F 联合领投方共同领投 2026 年融资,并帮助验证扩张期投资财团Access 在治理中有多活跃,还是纯财务参与?
Fidelity Management & ResearchSeries E 领投方;Series F 参与方;Series E 阶段董事会观察员反复出资方,也是少数公开治理信号之一随着公司接近 IPO 准备状态,Fidelity 的观察员身份是否变化?
NEASeries E 参与方;公开引述的战略支持者品牌足够大的成长投资人,并公开背书 AI GPU 投资逻辑NEA 是否持有未公开披露的正式董事席位或其他治理权?
NVIDIA战略投资人兼技术伙伴出现在融资财团及 AI 数据管线的伙伴架构叙事中VAST 的市场牵引有多少依赖 NVIDIA 生态协同?
Tiger Global Management 领投方Series D 领投方支持 2021 年估值重定价至 $3.7B后续融资和老股交易后,Tiger 还剩多少持仓?
员工和早期投资人老股流动性受益方Series F 及部分 Series E 报道显示,新股融资之外还有流动性安排头条融资额中,多少是真正进入资产负债表的资金,多少是给既有持有人提供流动性?

仅为公开披露的利益相关方图谱;因留存来源未提供,省略未披露投资人、持股比例、清算条款和债务。

[CO016, CO020, CO021, CO022, CO027, CO028]

1.4 里程碑、合作伙伴与不利背景

里程碑记录解释了为什么后续章节不应只把 VAST 当作存储厂商。2016 年创立后,公司第一阶段先打下高性能存储基础,随后在其上叠加数据库和计算功能;到 2026 年,发布重点已经转向控制平面和智能体计算能力,而不只是贴近硬件的软件。合作伙伴图谱同样是主线。VAST 的客户和渠道足迹穿过 AI 云、OEM 和战略技术伙伴,具名关系包括 CoreWeave、Google Cloud、Lambda、Core42、Supermicro、HPE、Cisco 以及 NVIDIA 相关部署。该生态扩大触达,但也让公司依赖伙伴路径和超大规模 AI 资本开支周期。需要保留的主要负面事项是 Red Stapler/NetApp 纠纷。VAST 本身不是被告,NetApp 在 Florida 的案件也因管辖地理由被驳回,但这个事件恰好发生在 VAST 再次募集巨额融资时,仍带来声誉和整合风险。从事实底座看,现有证据下这项法律问题不足以推翻投资逻辑;但它是真实的尽调记忆项,直到冰岛相关程序或上诉程序清楚结束前,都应保持跟踪。[CO035, CO036, CO037, CO038, CO039, CO040]

里程碑表
日期事件类型金额 / 估值 / 状态参与方含义
2016VAST Data 成立成立公司成立Renen Hallak 及其他早期创始人公司从存储逻辑走向更宽数据平台野心的起点
2021-05-04Series D 完成融资$83M,投后估值 $3.7BTiger Global、NVIDIA、既有投资人让 VAST 确立为高估值独立基础设施公司
2023-04后续 Series E 材料强调与 HPE 的战略合作合作HPE GreenLake 文件存储集成VAST Data 与 HPE标志着 VAST 借伙伴分销切入企业文件与 AI 数据基础设施
2023-05后续 Series E 材料强调 NVIDIA DGX SuperPOD 认证合作认证里程碑VAST Data 与 NVIDIA增强其在以 GPU 为中心的 AI 部署中的技术可信度
2023-12-06Series E 完成融资$118M,估值 $9.1BFidelity、NEA、BOND、Drive估值较 Series D 翻三倍,并为更宽 AI 平台扩张提供资金
2025-06-10独立报道在下一轮前披露新估值目标规模化目标估值 ~ $25BTechCrunch 消息源、VAST 管理层背景显示投资人需求和最终 Series F 前的重新定价
2025-11-19Red Stapler / NetApp 纠纷公开反向诉讼与整合悬而未决NetApp、Jonsi Stefansson 与 VAST DataVAST 融资期间产生声誉尽调风险
2026-01-26VAST Amplify 发布产品公司称有效 SSD 容量最高提升 6xVAST Data显示 VAST 从存储硬件经济性扩展到 AI 基础设施效率方案
2026-02-25Polaris 在 VAST Forward 发布产品全球控制平面已可用,并计划扩展能力VAST Data将公司叙事从存储 / 数据平台延伸到多云 AI 编排
2026-04-22Series F 完成融资~$1B 交易,估值 $30BDrive Capital、Access、Fidelity、NEA、NVIDIA 等将公司推入私营 AI 基础设施估值第一梯队

这是公开记录中的时间线,不是内部历史全表;部分 2023 年合作里程碑的确切日期来自后续回顾,而非单独留存的新闻稿。

[CO001, CO016, CO020, CO022, CO027, CO028]
FO001: VAST Data 公司里程碑时间线

从创立到 2026 年 4 月 Series F 完成的公开时间线,展示公司如何从存储基础设施转向 AI 操作系统平台,同时仍有一项明显法律悬而未决问题。

[CO001, CO016, CO020, CO022, CO027, CO028]

1.5 图表要点

Chapter 02

02市场分析

2.1 市场边界、相邻领域与替代方案

VAST 不宜被笼统归为普通 AI 基础设施公司。它自己的定位和白皮书把公司放在原始存储介质与上层 AI 应用软件之间的共享数据层:一个统一平台,覆盖文件、对象、块、表格、向量和流式数据,并能从同一控制平面服务训练、推理、分析和 HPC 工作负载。因此,核心市场边界包括高性能共享存储、元数据与向量检索、全局命名空间与治理,以及让这些数据服务在本地、云和边缘环境保持可用的软件。它不包括 GPU 计算、网络互联、通用应用软件或大块通用 IT 服务支出。实际竞争框架因此窄于整体 AI 基础设施,但宽于传统 NAS 阵列。VAST 自己的方案页面延伸到分析、合规、备份、容器和数据湖仓等相邻用例;白皮书则描述用一个以数据为中心的平台取代分散的对象存储、分析集群和编排层。现状替代方案仍包括既有全闪阵列、横向扩展 NAS、对象存储,以及由分散存储、数据库和编排工具拼出的多产品栈。市场定义很关键:如果把更宽的 AI 基础设施口径当 TAM,会高估 VAST 能现实捕获的支出池;如果只用经典外部存储类别,又会低估 VAST 想卖的软件和编排层。[CM001, CM003, CM004, CM005, CM007, CM008]

市场定义表
细分市场 / 边界纳入支出排除支出买方 / 付款方适配度
核心服务市场共享 AI 数据基础设施:高性能文件、对象、块、元数据、向量检索、治理和全局命名空间服务GPU 计算、网络、通用 AI 软件、咨询服务CIO / 基础设施或平台预算负责人;用户是存储、ML 和数据平台团队直接
AI 工厂邻近市场集成存储 + OEM 计算套件、部署软件、参考架构和 NVIDIA 生态数据服务原始加速器芯片支出和广义数据中心建设中央 AI 平台或基础设施发起方邻近 / 伙伴主导
数据平台邻近市场可落地扩张到同一闪存数据底座中的湖仓、分析、事件、合规和备份工作负载通用 SaaS 分析席位和非数据类应用软件CDO / 数据平台负责人或安全 / 韧性负责人邻近
排除的广义 AI 基础设施共享数据层之外无超大规模云厂商资本开支、服务器、网络、电力、冷却和通用 AI 服务更宽企业或服务商资本开支池作为 TAM 过宽
现状替代方案既有全闪阵列、横向扩展 NAS、对象存储和拼接式多产品栈不是独立支出池;这些是替换路径存储、HPC 和平台运营方竞争同一工作负载

边界逻辑结合 VAST 官方定位与公开市场报告定义;适配度区分直接支出、邻近市场和明确排除项。

[CM001, CM003, CM005, CM007, CM008, CM021]

2.2 规模测算口径与方法冲突

公共市场数据支撑多种可辩护口径,而不是一个干净 TAM。窄口径下,IDC 的外部 OEM 企业存储预测到 2026 年约 $37.7 billion,并明确绑定品牌存储系统,其中全闪阵列受益于 AI 和分析需求。Fortune Business Insights 将 AI 驱动存储置于 2026 年 $44.94 billion,这是更宽口径,包含为 AI 工作负载优化的存储架构,并区分文件型与对象型方法。Research and Markets 与 Coherent Market Insights 都把广义 AI 基础设施放在 2026 年约 $90 billion,但这些报告打包了远多于存储的内容,横跨产品、功能、技术、部署模式和终端用户。Gartner 的 $1.366 trillion AI 基础设施数字更宽,只能作为 VAST 品类周围总生态资本开支有多大的背景。估值时,正确结论不是某一个数字对、其他都错。更准确的是,VAST 位于几个嵌套支出池里,每家发布机构选择了不同边界。最可防守的公开框架,是一个受证据约束的核心口径:下沿由企业外部存储限定,上沿由 AI 驱动存储类别限定;更宽的 AI 基础设施数字只作为相邻市场,而不是 TAM。公开来源还没有披露足够的 VAST 定价、客户或赢单率数据,无法拆出干净的 SAM 或 SOM;因此,应保留互相矛盾的估算和方法限制,而不是把它们平均掉。[CM009, CM011, CM013, CM017, CM018, CM019]

TAM / SAM / SOM 或规模测算视角表
发布方 / 视角年份 / 地域数值CAGR / 增长方法置信度局限
IDC 外部 OEM 企业存储2026 / 全球$37.7B6.3% YoY品牌外部企业存储系统;AI 与分析需求、全闪组合不含超大规模云厂商自建和非 OEM 基础设施
Gartner AI 基础设施2026 / 全球$1,366.4BAI 总支出增长 44%表 1 语境下的全部 AI 基础设施支出;并非仅存储用作 VAST TAM 过宽
Fortune AI 驱动存储2026 / 全球$44.94B25.2% CAGR 至 2034覆盖 SAN、NAS、文件、对象和终端用户细分的 AI 优化存储系统发布方口径宽于共享企业 AI 存储
Research and Markets AI 基础设施2026 / 全球$90.91B25.7% CAGR 至 2030按产品、功能、技术、部署和终端用户划分的 AI 基础设施将存储与更宽基础设施类别打包
Coherent AI 基础设施2026 / 全球$90.0B24.0% CAGR 至 2033广义 AI 基础设施,其中硬件占 54%、本地部署占 46%、企业占 48%、北美占 40%类别覆盖整个基础设施,不是 VAST 专属
受证据约束的 VAST 核心视角2026 / 全球≈$38B-$45Bn/a用 IDC 存储下限和 Fortune AI 存储上限约束共享数据基础设施仍不是公司专属 SAM 或 SOM,因为价格、组合和胜率未公开

除最后的边界估计外,所有数值均为发布方声明的 2026 支出视角;最后一项是窄口径存储与更宽 AI 存储之间的分析桥,并非真正 SAM。

[CM009, CM011, CM013, CM017, CM018, CM019]
FM001: 市场规模测算视角

从总 AI 基础设施到 VAST 可能服务的更窄共享 AI 数据层,按嵌套支出池拆分。

底层故意保持定性,因为没有公开来源单独拆出 VAST 特定的 SAM 或 SOM。

[CM009, CM011, CM013, CM017, CM018, CM020]
FM002: 市场估计区间

已发布的 2025-2026 年支出区间,均以 USD 十亿美元计,说明 VAST 市场规模测算为什么必须保留口径差异。

这些行单位相同,但市场边界不同;这张图比较口径,不是可相加的 TAM 堆栈。

[CM009, CM011, CM013, CM018, CM019]

2.3 买方、用户、付款方与采纳路径

买方图谱取决于细分场景,但模式一致:技术倡导者来自基础设施、数据平台、HPC 或 AI 工程团队;预算批准和风险审查则上移到 CIO、采购、财务和更大的购买委员会。市场报告和厂商参考架构解释了原因。Fortune 的终端用户切分把企业、CSP、电信和政府列为不同需求池。Coherent 对敏感行业本地部署的偏向,加上 VAST 自己的企业 AI 工厂框架,说明对需要控制、治理和数据本地性的买方来说,受监管企业、公共部门或 HPC 用户仍然重要。在这些账户里,存储管理员、ML 平台团队或科研计算负责人可能指定架构,但他们很少单独购买。IDC 称采购、财务和营收运营职能已经成为核心决策者;Forrester 称典型复杂采购现在涉及 13 个内部利益相关方和 9 个外部影响者。这种治理很重的采购流程也解释了采纳路径。试用和可引用案例重要,因为买方必须先证明业务价值,才会扩容。Forrester 称超过 60% 买方会试用,超过 $10 million 的买方中 78% 会试用。Supermicro 和 VAST 卖的是从设计到部署的一站式路径;VAST 的参考架构则强调从小规模起步,随需求扩展。因此,实际漏斗是问题发现、架构验证、价值证明、首个生产集群,再到更多工作负载或站点的横向铺开。公开仍缺的是 VAST 账户穿过这条漏斗的真实转化率。[CM016, CM018, CM022, CM026, CM029, CM031]

细分市场 / 买方图谱
细分市场买方用户付款方 / 预算负责人工作流采用触发因素
企业 AI 工厂CIO、基础设施副总裁或中央 AI 平台牵头人ML 平台、存储和数据工程团队中央基础设施或转型预算统一训练 / 推理 / RAG 堆栈需要让 GPU 保持高利用率,并缩短首 token 时间
受监管企业 / 主权数据CIO 加安全、合规和采购平台工程和数据治理团队核心 IT 预算加经风险审查的资本开支本地或混合 AI 部署需要控制权、可审计性和数据本地化治理
云服务提供商 / GPU 云平台或基础设施总经理SRE、存储和 ML 服务团队基础设施资本开支负责人面向 AI 服务的共享多集群数据层需要可扩展共享存储和快速上线的经济性
科研 / HPC / 公共部门科研计算主管或机构 IT 负责人HPC 管理员、科学家和数据经理项目、经费或机构基础设施预算检查点密集型训练和分析需要高吞吐、一致性和可管理的规模
数据 / 分析扩张CDO 或数据平台负责人数据工程和分析团队中央数据平台预算湖仓、事件、合规和分析工作负载需要一张同时承载结构化和非结构化数据的全闪存数据底座

买方和付费方角色把公开的 2026 年采购群体研究,与 VAST 及其合作伙伴的部署叙事合并;公开资料没有披露 VAST 实际收入的细分结构。

[CM016, CM018, CM026, CM029, CM031, CM032]
FM003: 买方 / 细分市场图

围绕 VAST 邻近的主要买方细分,按角色、部署偏好和治理准备度做矩阵。

这些单元格是定性角色分配,综合了细分市场定义、公开证据中对治理敏感部署偏好和混合控制平面需求的描述, 而非来自 VAST 披露的客户结构。

[CM018, CM026, CM029, CM031, CM032, CM043]
FM004: 采用漏斗 / 价值链图

企业采购从识别问题到规模化生产部署的定性路径。

由于 VAST 未公开阶段转化率,这张图是定性价值链图。

[CM026, CM027, CM029, CM031, CM032, CM033]

2.4 增长驱动、采纳约束与尽调缺口

需求驱动真实且多层。IDC 的存储预测、VAST 自己的 AI 工厂叙事和参考架构都指向同一主题:AI 生产工作负载越多,就需要更多共享数据基础设施、更多全闪需求,并且更在意 time-to-token 和 GPU 利用率。Deloitte 显示企业 AI 活动正在从试点走向生产;Cloudera 将 2026 年描述为从实验转向运营智能编排。NetApp 和 Pure 也强调,既有厂商已经在把同一需求货币化:企业数据需要为 AI 做好准备、可访问。换句话说,VAST 卖的是一个真实品类,不是纯理论市场。但约束同样重要。Gartner 称企业仍对 ROI 敏感,并且在当前 AI 周期里往往偏好既有供应商。Informatica 和 Deloitte 都显示,治理、数据质量和技能缺口正在拖慢扩张。Electronic Design 和 SNIA 把存储视为一阶设计约束,因为它影响性能、功耗、散热和部署风险;Avnet 则显示 SSD 和 DRAM 紧张可能把成本推高,并把供应风险延长到 2027 年。INFUSE 又补上一层买方信任问题:信息更多,不一定带来更强购买信心。这组因素意味着,VAST 估值不能假设采纳毫无摩擦。最大的未解尽调问题是公司特定问题,而不是全市场问题:定价、客户数、工作负载组合、试用到生产的转化率,以及广义 AI 资本开支浪潮到底有多少落到 VAST 服务的共享数据层。[CM010, CM012, CM024, CM028, CM034, CM036]

增长驱动因素与约束表
驱动因素 / 约束方向时点影响尽调问题
AI 数据增长和全闪存需求驱动当前支撑对高吞吐共享存储的持续需求VAST 订单中有多少来自 AI 生产工作负载,而不是传统存储更新?
GPU 利用率和首 token 时间压力驱动当前利好能减少数据瓶颈、让加速器保持忙碌的平台提供首 token 时间和 GPU 利用率提升的基准测试或客户证据
运营型 AI 走出试点驱动2026需求从 PoC 扩大到标准化数据平台采购VAST 管线中,生产扩张与首次试点分别占多少?
采购和财务对 ROI 的审查约束当前拉长销售周期,也削弱纯技术拥护者的影响力VAST 在真实交易中使用什么 ROI 框架和回本周期?
治理和数据质量缺口约束当前在受监管和智能体 AI 用例中抬高采用摩擦赢下受监管客户时,最常用哪些治理、血缘和策略控制?
买方信任缺口和试用依赖约束当前更需要客户背书、外部验证和 PoC 转生产支持VAST 各细分市场的可背书程度和试用转生产转化率如何?
内存 / SSD 紧缺和资本开支通胀约束2026-2027即便需求存在,也可能压缩 ROI 或推迟部署VAST 的 BOM 和定价对 DRAM / SSD 涨价有多敏感?
既有厂商和渠道力量约束当前除非差异化得到证明,大买家可能默认选择 NetApp、Pure、Dell 或捆绑 OEM 栈VAST 在哪些大规模场景替换了既有厂商,赢单靠的是哪类工作负载?

这些行把公开的 2026 年需求信号,落到具体采用摩擦和尽调问题;若要精确量化多项约束,还需要公司内部数据。

[CM010, CM012, CM024, CM028, CM033, CM034]

2.5 图表要点

Chapter 03

03竞争格局

3.1 竞争版图、同业类别与替代路径

VAST 的竞争场远不止“其他闪存阵列公司”。Coldago 的 2025 文件存储图谱和 StorageNewsletter 的方法总结,把 VAST 放在企业存储和高性能存储重叠框架中,与 Pure、NetApp、IBM、DDN、Qumulo、Hammerspace、WEKA 并列。因此,直接同业是 AI 优先的文件 / 对象或高性能数据平台厂商,而不只是外形相似的设备厂商。下一圈是既有上市存储厂商。Pure 和 NetApp 现在在宽产品组合上销售用量导向的商业模式;IBM 仍把 Storage Scale 定位为软件定义 AI/HPC 和混合数据平台。相邻圈层也重要:Hammerspace 和 Qumulo 主打编排、全局命名空间和混合云流动性,这会让买方降低 VAST 权重,即便不复制 DASE 架构。替代方案不是理论。AWS FSx for Lustre、S3、Azure Managed Lustre 和 Google Filestore 提供按用量计费、原生账单和既有云合约。内部自建和维持现状也持续存在,因为 Ceph 和 MinIO 可以统一关键存储语义,而不把更多筹码交给单一专有厂商。结果是,VAST 同时面对直接产品竞争、既有厂商捆绑、编排替代、超大规模云原生替代,以及开源反锁定叙事。[CP001, CP002, CP003, CP006, CP011, CP015]

竞争对手画像表
竞争对手 / 类别类型规模 / 融资信号目标客户产品范围 / 战略方向局限 / 风险
Pure Storage上市既有厂商FY2025 收入超过 $3B正在标准化文件 / 对象和 AI 数据管线的企业FlashBlade 加 Evergreen//One,把统一文件 / 对象存储和服务化打包结合起来既有业务宽度可能稀释 AI 专项差异化,定价仍靠谈判
NetApp上市既有厂商上市既有厂商;保留页面强调广泛混合云组合和 Keystone STaaS希望延续 ONTAP 并保留合同灵活性的混合云企业覆盖本地和公有云的统一存储组合,Keystone 支持按用量付费公开标价仍有限,产品铺得太宽可能削弱聚焦 AI 性能的叙事
WEKA直接 AI 性能同业E 轮融资 $140M,估值 $1.6B;2026 年合作伙伴与 NVIDIA 推动AI 工厂、HPC 和多租户 GPU 环境软件定义 NeuralMesh 架构,聚焦 AI 性能、多租户和合作伙伴主导的 AI 工厂私营公司经济性仍部分不透明,更广泛的企业分销还在建设中
DDN直接 AI 性能同业私营公司披露的规模主张包括 10 家领先汽车公司中的 8 家、10 家头部银行 / 证券公司中的 7 家AI 工厂、主权 AI、超大规模云、受监管和 HPC 比重高的客户数据智能平台,以 Infinia 和其他 AI 导向产品覆盖推理、主权 AI 和横向扩展工作负载最强势的一些主张来自公司自述,保留证据中没有更新公开定价或资本结构
IBM Storage上市既有厂商 / 软件定义替代方案全球上市既有厂商;Storage Scale System 6000 获 NVIDIA 认证,强调高吞吐希望用既有 IBM 关系采购软件定义 AI/HPC 数据服务的受监管企业Storage Scale 加 Scale System,销售文件 / 对象统一、内容感知智能和 GPU 就绪性能IBM 更广的产品组合可能拖慢清晰的纯 AI 叙事,定价仍以报价为主
Hammerspace相邻编排挑战者$100M 融资,估值 $500M+;TechCrunch 称 Meta 和 DoD 是客户希望获得数据本地性、编排和混合云迁移能力的 GPU 密集环境Tier 0 本地 NVMe 架构,加全局命名空间和策略驱动放置分销规模小于既有厂商,其最激进的 VAST 批评来自竞争对手
Qumulo相邻混合云挑战者私营公司;保留页面强调 NPS、EB 级云规模和托管 Azure 路径,而非更新融资信息文件密集型企业和云工作负载,需要全局命名空间和托管云选项现代文件 / 对象平台,提供 Cloud Data Fabric、Azure Native 托管服务和混合云部署路径规模、融资和已实现定价的可见度低于上市既有厂商
云厂商原生替代方案替代方案 / 现状延伸既有云合同和公开用量计费结构已锚定 AWS、Azure 或 Google Cloud 的团队托管文件系统和对象服务,具备透明用量计费和原生采购通道数据迁移、闲置容量成本和性能调优,仍可能让云端“廉价”存储在实际中变贵
自建 / 开源替代方案 / 反锁定路径没有供应商股权价值;经济性转向硬件、人力和运维愿意自己承担存储工程和集成的平台团队Ceph 和 MinIO 展示了统一或 EB 级语义,不需要专有供应商锁定运维负担、调优复杂度和更慢的见效时间抬高执行风险

规模 / 融资单元格只使用保留公开语料披露的内容;对私营公司行,在融资、收入或定价未披露时保持这一点。

[CP003, CP005, CP006, CP007, CP010, CP012]
FP001: 竞争定位图

VAST 在 AI 性能专门化上位置较高,但老牌厂商和超大规模云厂商仍主导采购杠杆, 编排挑战者也在挤压控制平面叙事。

分数是基于保留证据的序位综合,不是基准测试输出。x 轴加权 AI 性能和热数据专门化; y 轴加权装机基础、合同杠杆和市场路径权力。

[CP005, CP010, CP015, CP020, CP027, CP035]

3.2 竞争对手画像与能力取舍

留存资料显示,同一预算有三种攻法。Pure 和 NetApp 从既有厂商端切入,把多协议存储广度和更让采购放心的包装绑在一起;这些采购团队本来就在标准化上市公司供应商。WEKA、DDN 和 IBM 则强调 AI 性能、GPU 效率或横向扩展数据服务,比设备熟悉度更重要。Hammerspace 和 Qumulo 转向编排和数据本地性:Hammerspace 的 Tier 0 故事是本地 NVMe 加共享访问能击败外部存储瓶颈;Qumulo 强调全局命名空间、云原生部署模式和 Azure 托管路径。VAST 的公开叙事宽于原始存储硬件,因为它现在声称自己是覆盖存储、数据库和计算的 AI 操作系统。这个更宽叙事有帮助,但竞争端负面很明显:一旦 VAST 说自己是智能体 AI 的数据和执行底座,它就不再只和文件存储设备比。它会被拿来和所有声称统一数据服务的厂商、所有承诺数据本地执行的编排器、所有提供透明计量和既有合约的云替代方案比较。因此能力必须同时通过两条轴线检验:AI 工作负载热数据平面的深度,以及围绕它的工作流、命名空间、部署和采购选择广度。[CP001, CP003, CP006, CP008, CP011, CP013]

功能 / 能力矩阵
采购标准VASTPureNetAppWEKADDNIBMHammerspaceQumulo超大规模云 / 自建
统一文件 + 对象定位中-强不一
AI / HPC 性能专精很强很强很强本地性强不一
全局命名空间 / 编排中-强不一
GPU 本地性 / Tier 0 角度弱-中弱-中很强弱-不一
托管服务 / 云路径未知 / 销售主导中,借助 Evergreen强,借助 KeystoneUnknownUnknown在 Azure 中强很强
公开定价透明度Unknown部分部分UnknownUnknownUnknownUnknownUnknown
采购 / 合规覆盖面中-强中-强
反锁定叙事

单元格是基于保留公开来源综合出的定性评分;“未知”表示公开语料没有披露足够可支撑细节。

[CP001, CP003, CP006, CP008, CP013, CP015]
FP002: 功能广度 / 能力图

主要分野不只是原始功能数量,而是供应商更强调热数据性能、编排广度、云采购便利性, 还是反锁定定位。

[CP017, CP025, CP028, CP034, CP035, CP037]

3.3 定价透明度、切换摩擦与分销力量

定价可见性是语料中最清楚的结构性差异之一。AWS、Azure 和 Google 都为文件或对象服务发布明确或可由计算器驱动的用量结构;专业存储厂商大多仍把买方导向销售、演示或谈判式订阅结构。这不意味着超大规模云原生替代方案在每个真实工作负载里都更便宜,但它确实降低采购摩擦,也更容易在既有云合约里比较。Pure 和 NetApp 又用 STaaS 模型包装平台,进一步缩小差距,而不是只卖资本开支型设备。VAST 的切换成本仍然存在,因为 PB 级数据集、调优过的 AI 流水线和运营熟悉度都很粘。但证据说明,这些成本不像垄断式锁定那样硬:买方可以在云对象、托管文件系统、编排层和开源层之间多栖,几个替代方案还明确销售反锁定或可移植叙事。分销进一步放大这一点。上市既有厂商和超大规模云厂商已经掌握大量账户控制权,WEKA 和 Hammerspace 还在投入扩张销售或合作伙伴。因此,VAST 要保持差异化,不能只靠速度宣称;当买方把它和云账单便利性、既有厂商账户杠杆放在一起比较时,产品、迁移路径和商业模式都必须继续站得住。[CP004, CP007, CP018, CP021, CP022, CP023]

定价 / 打包对比
方案商业模式公开定价可见度计量 / 单位影响
VAST Data定制企业报价 / 产品导航提到消费型选项协商定价不透明提高尽调负担,也削弱并列表比较的便利性
Pure Evergreen//One带 SLA 的存储即服务部分已用容量、服务层级、性能消费模式缩小相对专精厂商的资本开支劣势
NetApp Keystone混合云 STaaS部分用量层级、期限、性能 / 容量让 NetApp 能用运营开支口径和云端弹性扩展回应 AI 存储招标
WEKA软件订阅 / 合作伙伴主导的 AI 工厂打包协商高性能叙事很强,但买方仍需要定制报价
DDN / Infinia企业平台报价协商DDN 靠 AI ROI 和密度竞争,而不是透明标价
IBM Storage Scale软件或一体机报价协商IBM 可以打包软件定义和一体机选项,但公开定价稀疏
Hammerspace软件层企业报价协商可能减少外部闪存需求,但实际节省需要验证
Qumulo托管 Azure 加自托管 AWS/GCP 模式低到部分托管服务或自托管云资源足迹云路径提高采购灵活性,尽管标价仍然稀疏
AWS FSx for Lustre 托管文件服务按用量付费存储、吞吐、元数据 IOPS、备份、传输透明定价加既有合同,让它成为容易采用的基准替代方案
AWS S3按用量付费存储、请求、传输及相关功能冷数据层便宜,但无法完整替代高性能共享文件系统
Azure Managed Lustre按用量付费 / 报价计算器每 GiB 每月和每小时、性能层级对已在 Azure 的 HPC 和 AI 团队,托管服务降低运维负担
Google Filestore预置型云文件系统预置容量和 IOPS 特性预置计费让闲置容量成本显性化
Ceph / MinIO 自建软件加自运营基础设施不一硬件、云和人力,而不是供应商订阅避免供应商锁定,但工程和支持负担落到内部团队

这里没有把费率归一到统一 TCO 模型;本表比较打包方式和公开可见度,不比较折扣后谈判净价。

[CP004, CP007, CP022, CP023, CP024, CP026]

3.4 护城河耐久性、多栖与反向读数

VAST 的护城河真实存在,但留存证据显示它是有条件的,不是绝对的。Coldago 仍把 VAST 视为领导者,VAST 自己的规模叙事也延伸到 $30 billion 估值和 AI 操作系统叙事。这支撑乐观情景:一家具有可信规模、强 AI 定位和共享数据平面的厂商,可以整合存储、元数据和部分 AI 工作流。反向证据重要,因为它攻击同一个投资逻辑。theCUBE 认为 VAST 的 AI-OS 野心走在当前平台成熟度之前;StorageMath 称 VAST 的数据缩减叙事夸大现实,并把客户推向更深的生态锁定。Hammerspace 的竞争简报从 Tier 0 和云规模编排上攻击 VAST。与此同时,Pure 和 NetApp 并未停步;它们已经把技术广度和更安全的商业包装配在一起。云替代方案持续挤压成本和便利性端,内部自建选项也让反锁定替代继续存在。现实读法是,VAST 的护城河取决于它能否在热 AI 数据性能和一体化工作流价值上保持显著优势,同时阻止买方得出“既有厂商包装、编排覆盖层或云原生替代已经够好”的结论。多栖是关键泄压阀:如果客户只把 VAST 留给最热层,其他部分交给别处解决,护城河耐久性会变弱。[CP018, CP027, CP028, CP029, CP030, CP031]

护城河耐久性 / 竞争风险登记表
护城河主张威胁严重性证据 / 影响缓释措施 / 尽调问题
AI 优先的共享数据平面WEKA 和 DDN 也在讲 AI 原生性能叙事VAST、WEKA、DDN、IBM 和 Pure 的工作负载重叠面很宽按工作负载类别测试赢率,不要预设谁拥有通用 AI 存储领先地位
AI Operating System 整合叙事独立分析师认为产品成熟度仍落后于湖仓 / 超大规模云厂商同类方案theCUBE 认为 AI OS 目标走在当前平台成熟度前面验证数据库、向量和智能体运行时层的生产引用,不只看存储
数据缩减和效率营销反向来源认为 Flash Reclaim / Amplify 的说法被夸大,并加深锁定效应StorageMath 直接质疑 VAST 的数据缩减算法和锁定激励要工作负载级证明,不要只看组合平均值
热数据性能与密度如果「足够好」更便宜或更容易采购,超大规模云厂商和传统厂商的打包方案仍能胜出云替代方案公布价格,传统厂商则叠加 STaaS 包装对标总工作流成本,包括闲置 GPU 和迁移成本,不只看原始容量
命名空间与多站点控制Hammerspace 和 Qumulo 更明确强调编排与数据放置中高相邻挑战者可以与 VAST 并存,而不是直接替代它检查 VAST 能否掌握控制平面,而不只是最热的存储层
受监管企业可信度上市传统厂商和超大规模云厂商有更宽的采购通道Azure Government 的采购表述和 DDN 的主权 AI 叙事,把信任变成选择维度并排梳理认证和公共部门采购路径
存量客户杠杆Pure 和 NetApp 可向既有账户交叉销售上市传统厂商把广覆盖产品组合和 STaaS 包装结合起来在传统厂商占比较高的账户中量化替换空间,再承销扩张
落地后切换概率低多栖部署和开源替代方案让退出路径保持开放中高Ceph、MinIO、云文件系统和编排覆盖层降低全有或全无依赖访谈客户,确认随时间推移有多少数据留在 VAST,多少流向二级层

这张清单是判断层,不是基准测试;严重性取决于每项威胁能多直接压缩 VAST 的差异化或定价权。

[CP018, CP029, CP031, CP033, CP035, CP036]
FP003: 护城河 / 准备度 KPI

公开数字标记显示,即使还没把谈判折扣纳入模型,VAST 面对的竞争对手和替代方案已经有不小资本、 规模或成本信号。

这些值混用不同单位,是决策辅助,不是估值模型。每一项纳入,是因为它公开且与竞争耐久性直接相关。

[CP005, CP010, CP014, CP016, CP019, CP044]

3.5 图表要点

Chapter 04

04财务情况

4.1 收入模式、定价与确认问题

VAST 的货币化最好描述为通过企业和伙伴渠道销售的基础设施软件,而不是纯自助式存储产品。公司营销的是统一 AI Operating System,组合存储、数据库和计算;较早公司材料明确称 Gemini 是一种软件消费模型,把硬件和软件经济性拆开。公开商业页面强调联系销售和演示,而不是自助定价。因此,证据足以识别收入机制,但不足以直接观察实际价格表或折扣行为。Cisco 的 Global Price List 提供了一条已验证的伙伴采购路径;Polaris 免费包含在内,则说明至少部分控制平面被捆进平台。CoreWeave 协议证明 VAST 能拿下非常大的伙伴相关承诺,但公开记录仍没有打通订单额、承诺经常性收入,以及任何用量或系统交付组成部分如何进入同一套收入定义。[CI001, CI002, CI003, CI004, CI005, CI006]

收入流表
收入流机制单位当前数值 / 状态质量尽调问题
AI Operating System 软件通过企业合同销售的统一存储、数据库和计算软件合同 / 年度承诺核心经常性层已确认;纯软件收入占比未披露存在性高,构成占比中索取按模块拆分的产品级收入构成和已确认经常性收入。
OEM / 渠道绑定部署VAST AI OS 随 Cisco 等 OEM 硬件栈销售合作伙伴采购部署Cisco GPL 确认渠道采购路径;实际合作伙伴加价未披露索取合作伙伴价目表、分成率和渠道来源订单额占比。
Neocloud / 超大规模云厂商承诺以 AI 云部署为锚的大型战略合同多年商业协议CoreWeave 协议价值 $1.17B;集中度占比未披露索取头部客户订单额、ARR 和续约时间表。
捆绑控制平面服务Polaris 等平台服务包含在 AI OS 包内捆绑功能无额外费用包含;没有披露独立服务收入存在性中,利润率影响低索取捆绑服务的附着率和支持成本分摊。
存量客户扩张 / 优化Amplify 等项目靠提升客户自有闪存利用率变现,而不只卖新硬件容量优化项目商业路径存在,但价格和收入贡献未披露索取存量客户扩张项目的收入贡献和毛利率。
支持 / 服务支持、部署和交付经济性大概率存在,但未单独披露unknown未发现公开分部拆分索取按客户层级拆分的支持收入、服务交付成本和附着率。

收入机制有证据支持,但公开资料没有按收入流拆分。收入确认似乎混合了合作伙伴绑定的系统经济性、已承诺经常性软件收入,以及更宽泛的非承诺收入结构;因此每一行讲的是机制,而不是经审计的收入构成。

[CI001, CI003, CI004, CI005, CI006, CI018]
定价 / 变现表
价格 / 单位 / 合同标价与实际价格折扣 / 未知项来源 / 可支持信号
Gemini 软件消耗模型无公开数字标价硬件 / 软件拆分和折扣表未披露Series D 新闻稿和 About 页面描述了软件消耗模型。
通过联系销售签订企业直销合同无公开标价所有实际 ASP 均未披露公开商业页面将买家导向联系销售和演示。
Cisco GPL 渠道采购可通过合作伙伴价目表获得,但公开单价不可见Cisco 加价和返利结构未披露StorageNewsletter 确认 GPL 可用且由 Cisco 支持。
历史初始落地规模2021 年平均初始投资约 $1M当前 ACV 可能显著更高;未披露当前下限Series D 新闻稿。
前 100 名新客户承诺平均超过 $1.2M均值周围的分布未披露TNW 和 Sacra 报道。
大型战略协议$1.17B CoreWeave 商业协议确认节奏、期限和未履约订单处理未披露CoreWeave 官方公告和 CRN 报道。
合同期限TNW 报道称通常为五到七年续约结构和终止权未披露TNW 报道。

VAST 有明确变现机制,但几乎所有实际价格点仍未公开。因此表格区分有证据支持的商业结构和缺失的实际定价数据,而不是假装存在公开价目卡。

[CI002, CI003, CI004, CI007, CI017, CI024]
FI001: 收入模型桥

VAST 如何通过软件、合作伙伴硬件和战略性云合同,把 AI 基础设施需求转化为订单和经常性收入。

这张桥接图有公开来源支撑,但只能指示方向,无法完整对账。公开来源没有拆分与系统销售绑定的收入、承诺经常性收入和更广泛非承诺收入的确切占比。

[CI001, CI003, CI004, CI005, CI006, CI012]

4.2 GTM 路径与销售效率代理指标

VAST 的公开 GTM 证据指向高接触企业销售,平均落地规模异常大,扩张动态强。2021 年公司称初始投入平均约 $1 million,扩张平均 328%;FY22 又称 NRR 超过 300%。更近期第三方报道显示,前 100 个新客户平均支出超过 $1.2 million,合约期限常为 5 到 7 年。这些都是企业收入耐久性的强代理指标,尤其是叠加 Cisco GPL 入口,以及云伙伴把公司带进 AI 基础设施预算。公司似乎也受益于巨型灯塔账户:仅 CoreWeave 就签下 $1.17 billion 协议,多家媒体还提到 xAI 和其他 AI 云客户。公开记录没有披露 CAC、CAC 回本周期、平均销售周期或伙伴抽成。因此,本章可以认为 VAST 很可能具备高效的先落地再扩张动态,但不能把这些需求信号换算成闭环销售效率模型。[CI006, CI007, CI008, CI017, CI018, CI024]

FI002: 单位经济性桥

从大型初始落地和扩张,到利润率和现金生成说法,梳理公开可见的单位经济性代理指标。

该图使用公开代理指标,而非闭合的单位经济模型。VAST 未公开披露 CAC、回本周期和周期时长,因此这些节点仍是定性判断。

[CI007, CI008, CI017, CI021, CI024, CI033]

4.3 成本结构、毛利率驱动与资本强度

最强的公开利润率信号指向软件主导的成本结构,尽管 VAST 卖入的是硬件很重的市场。2022 和 2023 年公司披露称业务建立在软件毛利率之上,后来又提到接近 90% 毛利率和正现金流。架构和商业模式在这里很重要:Gemini 切开硬件和软件经济性;白皮书称 Cisco、SuperMicro 等 OEM 可以销售标准服务器 VAST 集群;Cisco 现在也把平台放上 GPL。这些因素意味着,部分库存、制造和资产负债表营运资本负担可能落在 OEM 伙伴,而不是 VAST 自己身上。同时,Polaris 等捆绑功能和 Amplify 等已装机基础项目表明,支持、控制平面交付和优化服务嵌入了平台经济性,但没有单独披露。结果是一条看起来有利的毛利率路径,表观资本强度低于专有阵列厂商,但仍缺少当前分部利润率桥接和营运资本披露。[CI003, CI019, CI020, CI021, CI022, CI023]

单位经济表
指标数值 / 状态置信度重要性尽调问题
已承诺经常性收入下限> $500M CARR,上一财年结束时公开记录中最扎实的当前经常性收入下限提供 ARR/CARR 桥接和季度队列视图。
历史 ARR 锚点2023 年底 ARR 为 $200M为轨迹分析提供基准点确认 FY24 和 FY25 经审计 ARR。
扩张 / 留存代理指标2021 年平均扩张率 328%;FY22 NRR >300%显示历史上极强的落地后扩张表现提供按队列拆分的当前 NRR 和 GRR。
毛利率信号2023 年披露毛利率接近 90%;软件毛利率口径反复出现如果收入构成没有恶化,支撑类软件经济性提供当前综合和分部毛利率。
平均承诺规模历史初始落地约 $1M;前 100 名新客户平均 >$1.2M低到中大 ACV 可抵消较长的企业销售周期提供按细分市场拆分的 ACV 分布和成交额中位数。
合同期限代理指标2026 年报道常见为五到七年长期限提升收入耐久性和回本容忍度提供期限分布、续约权和提前终止数据。
CAC / CAC 回本周期未公开披露判断 AI 需求增长是高效还是单纯扩张的关键按渠道提供 CAC、回本周期和配额产出率。
销售周期 / 合作伙伴经济性未公开披露用来理解采购摩擦和渠道杠杆提供销售周期长度、合作伙伴分成率和附着率数据。

本表刻意混合已确认指标和缺失指标。VAST 的公开证据强力支持扩张质量和高毛利潜力,但还无法闭合 CAC / 回本模型。

[CI007, CI008, CI009, CI010, CI012, CI017]
FI004: 资本强度 / 现金流图

VAST 模型哪些部分看起来像软件,哪些公开披露仍让资本暴露不清晰,该图给出定性映射。

这是定性暴露图,不是现金流量表。公开来源能看出经济性大概率落在哪些环节,但看不出每项暴露的确切规模。

[CI019, CI020, CI022, CI023, CI030, CI031]

4.4 公开牵引力与资本充足性

对一家私营基础设施公司来说,公开牵引力异常强,但资本充足性仍只能通过不完整披露来读。官方来源显示,从 2023 年底 $200 million ARR,到上一财年退出时超过 $500 million 承诺 ARR,再加上 $4 billion 累计订单额、经营利润率和自由现金流为正,形成了一条阶梯。独立媒体把画面拉得更远:Reuters 相关报道谈到 $600 million ARR;Calcalist/TNW 谈到包含非承诺组成部分在内、最高约 $2 billion 的类 ARR 收入。这个差距不只是噪音;它告诉投资人,VAST 公开引用的收入指标用了不同定义。资本充足性方面,2026 年 Series F 对公司估值 $30 billion,总交易规模约 $1 billion,但公开报道显示,名义数字中很大一部分是老股交易。Fidelity 的 SEC 估值标记独立佐证了估值上跳,但公司仍未公开披露在手现金、烧钱速度、现金跑道、债务或融资触发条件。[CI010, CI011, CI012, CI014, CI015, CI016]

资本充足性表
项目当前数值 / 状态影响尽调问题
Series F 交易价值总额约 $1B,估值 $30B显示投资人需求很深,名义融资能力充足提供已签署融资文件和公司实际收到的新股现金总额。
新股与老股交易拆分官方称新股与老股交易混合;外部报道称老股交易 >$500M,且可能占多数名义融资额可能高估新增经营现金提供准确的新股募资额、老股交易额和卖方名单。
资金用途新股资金指定用于全球增长和战略交易指向扩张,而非救助性融资提供预算资金用途表和 M&A 储备假设。
盈利能力 / 资本依赖据报道经营利润率和自由现金流为正;CEO 称运营不需要外部资本支持资本充足性,但仍不能替代资金数据提供月度现金流报表和当前现金余额。
第三方估值标记2024 至 2026 年间,Fidelity N-PORT 对同一优先股篮子的标记从约 $1.08M 升至约 $3.51M独立组合标记印证估值上调提供最新投资人标记和股权结构表调节。
现金 / 烧钱速度 / 现金跑道未公开披露无法做真正的下行情景现金跑道分析提供资金资料包和 12/24 个月现金跑道情景。
债务 / 项目融资义务所审阅来源未公开披露无法排除隐藏债务契约或设备融资压力提供债务时间表、债务契约包和任何设备融资协议。

本表关注前瞻资本充足性,而不是轮次年表。核心问题不是 VAST 能否融到钱,而是 2026 年名义轮次中有多少真正补强了资产负债表,以及还有哪些未披露义务压在表面之下。

[CI026, CI027, CI028, CI029, CI030, CI031]
FI003: 财务估计区间

VAST 经常性收入和估值信号的公开数字区间,显示投资评估仍需要一座指标定义桥。

该图有意混用承诺 ARR、第三方 ARR 类估计和估值标记,因为核心尽调问题是定义不一致,而不是增长方向缺少证据。

[CI012, CI014, CI015, CI016, CI026, CI027]

4.5 财务判断与尽调阻塞项

财务读数有利,但还不能完全支撑最终投资判断。正面看,VAST 同时具备软件式利润率信号、多年合约证据、非常大的平均承诺额、正自由现金流说法,以及 AI 云客户带来的异常需求。和多数后期基础设施创业公司公开披露相比,这是更强的收入质量画像。谨慎点在于,几个最重要输入仍是私有信息:客户集中度、实际现金余额和现金跑道、CAC 与回本周期、实际定价、债务义务,以及订单额、CARR 和任何更宽 ARR 或收入口径之间的桥接。大型 AI 云账户也会引入集中度和续约时点风险,因为它们能如此剧烈地撬动业绩。因此,正确投资判断不是否定商业模式,而是把公开记录能支撑的内容,与仍需要公司披露的内容切开。VAST 的财务强度足以值得严肃尽调,但证据集仍不足以完成可签署交易级别的财务模型。[CI030, CI036, CI037, CI038, CI039, CI040]

公开财务缺口表
缺失的私有指标对承销的影响精确尽调路径
订单额到收入再到 ARR 的桥接没有指标定义桥接,公开收入质量无法与同业标准化对比索取从订单额到确认收入再到 CARR / ARR 的季度桥接,包括非承诺收入。
当前现金余额和现金跑道无法形成真正的下行流动性视图索取资金报告和董事会现金跑道情景。
债务 / 债务契约包无法评估隐藏融资压力或设备融资义务索取债务时间表、债务契约摘要和有担保债务的法律确认。
客户集中度与续约大型 AI 云交易可能主导风险和收入确认节奏索取前 10 大客户 ARR/订单额和续约日历。
CAC、回本周期和销售周期数据销售效率仍是定性判断,无法直接建模按渠道索取漏斗转化、CAC、回本周期和销售周期指标。
实际成交价格和折扣表公开定价机制无法转换成实际 ASP 或利润率假设索取价目表、订单样本和合作伙伴定价表。
分部毛利率和支持 / 服务成本当前利润率路径无法按收入流审计索取产品线和渠道毛利率桥接,以及支持成本分摊。

这些是把强公开叙事转成真正投资级财务模型所需的最低缺失输入。公司看起来商业表现亮眼,但数据室仍必须补上承销核心。

[CI031, CI032, CI033, CI034, CI037, CI040]

4.6 图表要点

Chapter 05

05产品与技术

5.1 按客户工作流理解产品范围

VAST 把平台卖成 AI Operating System,而不是单一存储 SKU。按客户工作流看,留存记录显示买方想把数据路径多个阶段压到同一个共享底座上:通过文件、对象、流式或 Kubernetes 接口摄取数据;通过 DataBase 编目和查询;在 DataEngine 中触发事件驱动处理;借助 DataSpace 让同一份数据跨站点可见,而不是复制到不同数据孤岛。AI 参考架构明确说明,公司想用一个闪存原生平台覆盖数据捕获、整理、模型训练、模型服务、审计日志和 RAG 式检索。Kubernetes 材料把这一点落到运维操作层:同一集群既可以暴露为 NFS,也可以暴露为 CSI 配置的持久卷。因此,工作流价值主张不只是速度,而是减少数据移动、避免分散的存储 / 数据库 / 流式栈,并给基础设施团队一个控制模型。[CE001, CE002, CE003, CE004, CE005, CE007]

产品模块 / 资产矩阵
模块 / 资产主要用户 / 买方公开状态 / 成熟度功能差异化尽调缺口
Universal Storage / DataStore存储架构师、AI 基础设施团队、平台运维成熟核心;在 2026 年 AI OS 重新叙事之前就已反复记录以 NFS、SMB 和 S3 暴露的共享文件和对象数据平面,底层元素相同一条写入路径可通过多种协议呈现,而不是形成割裂孤岛需要按工作负载给出当前公开基准测试和参考部署细节,而不只是架构主张。
VAST DataBase数据工程师、分析团队、RAG 管道负责人已记录模块;在 2024-2026 年叙事中显著扩展绑定目录、下推、SQL 分析和数据溯源工作负载的 ACID 表格引擎把结构化查询接到存储非结构化数据的同一平台上需要针对最新 GPU 加速数据库路径的公开 GA 文档和客户引用。
VAST DataEngine管道构建者、平台工程师、AI 应用团队可见,但公开文档深度仍不如存储核心面向函数、触发器、管道、主题和计算资源的事件驱动处理环境把平台从被动存储变成执行平面需要更多 DataEngine 工作负载的公开生产案例和运营 SLO。
DataSpace / Polaris 控制层多站点运营方、云团队、治理负责人DataSpace 有文档;Polaris 更新,更多来自公告面向分布式 VAST 环境的全局命名空间和混合控制编排让混合和多站点运营成为产品叙事的一部分,而不是外部覆盖层Polaris 需要公开管理文档、客户引用和 GA 状态清晰度。
AI OS 加速器(CNode-X、PolicyEngine、TuningEngine)企业 AI 平台团队、智能体 AI 构建者2026 年发布阶段的能力面;文档深度尚不如平台核心GPU 加速计算路径、内联策略执行和闭环模型调优把 VAST 从存储推向全栈 AI 工作流控制与加速需要证明客户生产使用、实际可用性和支持边界。
开发者 / 运维工具SRE、自动化团队、Kubernetes 运维方活跃且公开CSI 驱动、Helm chart、Python SDK、Go 客户端、Terraform provider 和 DataEngine CLI降低集成摩擦,并给产品一个可见的实践者界面除仓库存在和活跃度外,还需要使用遥测或下载 / 安装证据。

状态 / 成熟度反映留存公开证据的深度,而不是内部发货数据。核心存储平台的文档比最新 AI OS 控制和加速层更充分。

[CE001, CE003, CE005, CE007, CE008, CE009]
工作流 / 用例表
用户任务当前工作流 / 痛点VAST 方案可衡量 / 公开收益限制
整合 AI 数据摄取和模型服务数据路径团队在分离的归档、训练和推理栈之间复制数据面向捕获、精炼、模型服务和 RAG 检索的单一闪存原生平台公开架构把收益描述为减少复制,并让流水线各阶段共用一个平台公开语料偏架构;这里留存的客户特定 ROI 数据有限。
为 Kubernetes 工作负载提供持久存储原生容器存储依赖主机本地或手工配置将 VAST 用作 NFS 后端存储,或通过 CSI 动态配置官方文档展示两种模式和官方 CSI 分发路径未留存关于 CSI 采用的大规模部署公开指标。
在同一数据平面上跑实时分析和向量检索查询引擎和向量系统往往位于存储旁边,而非嵌在存储内部DataBase 加 GPU 加速 SQL,以及 CNode-X 上基于 cuVS 的向量检索2026 年材料声称查询时间 / 成本更低,检索加速最强数字来自公司发布的基准测试主张,不是独立测试。
在不新增独立 SAN 或 Kafka 资产的情况下加入块和流式能力企业把块工作负载或事件流留在分离孤岛上在共同平台上提供 NVMe/TCP 块能力和 Event Broker / Kafka 兼容摄取公开报道称块能力继承快照、克隆、复制和 QoS最新能力的生产规模公开证据仍比文件 / 对象核心薄。
构建具备网络韧性的备份或受治理的数据留存工作流备份平台常需要单独不可变目标和额外存储层Commvault 验证 VAST 可作为 NFS 或 S3 目标,并提供 WORM 不可变选项公开合作伙伴文档支持磁盘库和云库使用场景,并支持不可变副本存在验证,但本章没有留存更广泛的多厂商备份引用。

收益仅限留存来源直接支持的范围。独立部署结果数据对集成存在性最强,而非覆盖广泛公开 ROI 基准测试集。

[CE001, CE002, CE009, CE021, CE023, CE024]
FE002: 客户工作流 / 运营流程

公开定位中,VAST 如何把数据从摄取一路承载到查询、事件处理和受治理的 AI 使用。

[CE001, CE002, CE005, CE007, CE008, CE009]

5.2 架构与运行模型

架构重心是 DASE,也就是 VAST 的解耦式 shared-everything 模型。公开技术文档描述,无状态计算节点处理协议和软件服务,持久数据和系统状态则放在 NVMe 支撑的存储机箱里,从而让性能和容量可以独立扩展。这一点重要,因为平台叙事要求所有协议看到同一个底层数据平面:DataStore 暴露文件和对象接口,DataBase 增加表格和查询服务,Event Broker 把平台延伸到 Kafka 式流式场景,而块访问现在正通过 NVMe/TCP 加入。架构也被表述为部署灵活。较早参考架构仍展示经典 CBox 和 DBox 布局;Supermicro 和 Cisco 的当前伙伴材料则展示打包方案,在保留 VAST 软件抽象的同时,适配 OEM 硬件和 AI 工厂参考设计。净读数是,VAST 产品不是一堆松散独立产品;它是一个分层软件平台,其价值取决于共享数据平面能否在多种访问方式和节点拓扑下真正保持一致。[CE003, CE014, CE015, CE016, CE017, CE018]

技术 / 运营架构表
层 / 组件角色依赖关键风险
接入协议与客户端通过 NFS、SMB、S3、SQL/表、CSI 以及较新的 NVMe/TCP 块接口呈现共享数据客户端兼容、协议转换,以及稳定的共用元数据语义较新的块和流式接口成熟度可能落后于传统文件 / 对象接口。
CNodes / CBoxes / CNode-X运行 VAST 软件服务、协议处理、AI 加速和管理功能CPU 或 GPU 服务器供应,以及 Cisco 和 Supermicro 等 OEM 伙伴的封装加速路线图执行会绑在伙伴硬件和 NVIDIA 配套支持上。
DNodes / DBoxes / NVMe fabric 架构承载持久化元数据和数据状态,同时在集群内开放共享访问NVMe 闪存经济性、fabric 设计和数据保护算法硬件拓扑和闪存假设仍是性能与重建主张的核心。
DataBase 和目录层提供 ACID 表、目录元数据、下推、分析和溯源支持查询引擎集成、GPU 库,以及对象 / 文件元数据的一致捕获最新加速 SQL 主张的公开部署证据,比架构图更薄。
DataEngine 和 Event Broker运行事件驱动函数、触发器、流水线、主题和流式工作流开发者工具、运行时支持和流水线编写体验DataEngine 已公开可见,但运行时成熟度比存储成熟度更难尽调确认。
DataSpace 和 Polaris把可见性、控制和编排延伸到混合与多站点部署基于代理的集群管理、云集成,以及一致的策略传播Polaris 的公开文档仍追不上发布叙事的野心。

这是一张定性的运营模型图,不是包级规格说明。风险单元格聚焦公开证据比产品主张更薄的地方,尤其是最新层。

[CE014, CE015, CE016, CE017, CE018, CE020]
FE001: 产品架构图

从多协议访问,到共享数据服务、编排和底层 DASE 基础设施,VAST 平台中有公开证据支撑的各层。

[CE003, CE014, CE015, CE017, CE021, CE024]

5.3 部署、集成与路线图成熟度

VAST 的公开集成面足够让它看起来能在现代基础设施环境中运行,而不只是能演示。Juniper 验证设计暴露了具体集群拓扑、VIP 再均衡行为、GUI 和 CLI 管理,以及底层 REST API。Kubernetes 知识库文章加上 CSI 仓库和 Helm 文档,显示容器平台不是事后补上的。GitHub 组织、Python SDK、Terraform provider、DataEngine CLI 和 Go client 共同构成一个公开运维工具链,可用于配置、查询、自动化和流水线管理。不过,路线图推进速度快于最深层公开文档。块存储和 Event Broker 在 2025 年浮出水面,2026 年叙事又加入 Polaris、CNode-X、PolicyEngine、TuningEngine 和开放 DataEngine 蓝图。这些动作把买方故事从存储扩展到编排和智能体 AI 运维,但也意味着最新层更多由发布材料和评论支撑,而不是像核心平台和 Kubernetes 集成那样,有成熟管理员文档可见。[CE009, CE010, CE011, CE012, CE013, CE020]

路线图 / 发布 / 开发阶段表
日期 / 阶段功能 / 里程碑公开状态含义来源信号
2025 年发布通过 NVMe/TCP 提供原生块存储已公开宣布,并由新闻报道说明把 VAST 从文件 / 对象 / 表扩展到接近 SAN 的工作负载和远程启动场景DBTA 与 The Next Platform 报道
2025 年发布Event Broker / Kafka 兼容流处理已公开宣布,并由新闻报道说明让流式数据在同一平台上成为文件 / 对象 / 表数据的一等同级对象DBTA 与 The Next Platform 报道
2026 年公告CNode-X 加速服务器路径随 OEM 市场路径一同宣布加深对 NVIDIA 的依赖,同时明显拓宽 AI 计算叙事VAST 新闻稿、StorageNewsletter、Cisco/Supermicro 引用
2026 年公告Polaris 全球控制平面已宣布,并由独立会议报道说明把 VAST 从存储 / 数据层扩展到跨本地部署、neocloud 和公有云的集群编排SiliconANGLE 与 TechArena
2026 年路线图PolicyEngine计划在 2026 年内 / 到年底推出内联治理可能成为受监管智能体 AI 工作流的关键差异点SiliconANGLE 与 TechArena
2026 年路线图TuningEngine计划在 2026 年内 / 到年底推出推动 VAST 走向企业边界内的闭环模型改进SiliconANGLE 与 TechArena
2026 年生态推进Open DataEngine 蓝图已公开宣布表明 VAST 想要可复用的参考工作负载,而不只是基础设施原语VAST 新闻稿和 StorageNewsletter

2025 年项目作为已发布能力扩展,证据更扎实;几个 2026 年控制平面和智能体治理项目仍更像路线图阶段或早期推出功能,而不是有充分文档的普遍可用界面。

[CE022, CE024, CE025, CE027, CE029, CE030]

5.4 差异化、依赖与技术风险

可支撑的差异化逻辑更多来自共享数据平面的广度,而不是每一层新 AI-OS 能力今天都完全成熟。留存语料中,VAST 确实有真实产品区别:一个平台声称同时支持多协议文件、对象、表格、流式以及现在的块访问;DASE 被反复定位为平台能够扩展、同时避开传统 shared-nothing 或分层存储取舍的原因;与 NVIDIA 的合作则把 GPU 加速带入 SQL、向量搜索、检索和长上下文推理路径。话虽如此,同一证据也暴露了重要依赖。加速部署依赖 NVIDIA 库、DPU 和网络假设;Cisco 和 Supermicro 是 CNode-X 的明确 GTM 和硬件路径;最新治理和调优层仍更像已宣布能力,而不是被广泛证明的生产默认项。因此,产品看起来有技术差异化,但风险登记册落在接缝处:伙伴硬件可用性、前瞻性智能体 AI 功能执行,以及如何让更宽的 AI-OS 叙事显得和底层存储核心一样成熟。[CE016, CE018, CE021, CE025, CE026, CE027]

FE003: 关键依赖图

VAST 较新的产品层明显依赖合作伙伴硬件、NVIDIA 软件以及周边运营者生态。

[CE018, CE020, CE025, CE027, CE030, CE037]
FE004: 产品成熟度 / 能力图

核心平台和运营者工具的公开证据成熟度最高,最新 AI-OS 控制与加速层较低。

评级是基于文档深度、代码库可见度和发布阶段证据得出的等级判断,不是内部出货、采用或支持指标。

[CE003, CE009, CE012, CE021, CE029, CE039]

5.5 信任、合规与控制态势

公开信任证据真实存在,但并不均匀。最强材料不是一套打磨好的公开信任中心,而是部署和加固材料。VAST 的 Security Configuration Guide 称平台是 STIG 加固设备,记录了联邦采购导向,并说明产品列入 DoDIN Approved Products List。该指南还列出具体控制面:基于 HTTPS 的 VMS 管理、CLI 访问、包含 CSI 专用权限的预定义角色、与 Active Directory、LDAP 和 NIS 的身份集成,以及配置 SSO 和 MFA 前所需的可信 TLS 前提。AI 参考架构又补充了更宽的设计主张:零信任、RBAC、ABAC、加密、审计、网络隔离和租户特定密钥管理。Commvault 的集成指南通过记录 VAST 支撑的云库工作流上的 WORM 模式不可变备份副本,增加了网络韧性证据。谨慎点在于,本章留存语料仍没有提供宽泛的公开认证目录、清晰 SLA 或事故历史证据,也没有强公开引用证明最新 Polaris 和 Policy/Tuning 层已经在生产中广泛部署。[CE032, CE033, CE034, CE035, CE036, CE037]

信任 / 质量 / 合规表
控制 / 工件状态范围支撑证据缺口 / 风险
STIG 加固设备姿态公开文档说明联邦部署和加固基线Security Configuration Guide 称平台已按 STIG 加固,并列入 DoDIN APL加固信号很强,但不等同于覆盖广泛的公开企业信任包。
零信任与租户控制已公开声称并部分文档化RBAC、ABAC、审计、租户隔离、QoS、网络隔离、租户专属加密密钥AI 架构和安全指南描述这些控制需要更广泛的独立验证和当前认证映射。
身份与管理控制平面公开文档说明HTTPS VMS、CLI、角色、AD/LDAP/NIS/本地用户、SSO/MFA 之前的 TLS安全指南给出具体访问方法和角色细节需要关于密码策略默认值、事件响应 SLA 和外部审计的最新公开证据。
不可变保护与网络韧性公开支持通过伙伴工作流提供快照和 WORM 模式不可变备份副本AI 架构及 Commvault 集成指南留存材料没有包含更广泛的公开恢复测试或勒索软件恢复案例研究。
AI 生命周期安全扩展2026 年公布CrowdStrike 遥测,以及覆盖摄取、训练、运行时、推理的协同检测 / 响应CrowdStrike 新闻稿和 VAST Forward 报道发布材料明确提醒,部分引用功能可能尚未普遍可用。
公开信任可见性部分存在合规营销和加固文档合规页面和安全文档提及治理、安全和受监管场景这里未保留当前审计报告、SLA 承诺或事件历史证据的广泛公开目录。

本表区分已文档化的安全控制和企业买家可能仍会期待的更广义信任姿态。主要缺口不是没有证据,而是这些证据的公开包装不均衡。

[CE032, CE033, CE034, CE035, CE036, CE037]

5.6 图表要点

Chapter 06

06客户情况

6.1 客户分层与买方 / 用户 / 付款方图谱

公开记录支撑 VAST 拥有广泛但尚未完全枚举的客户基础。官方客户证明页面和名单披露显示,VAST 触达 AI 云运营商、电信主权 AI 建设者、金融服务机构、医疗与生命科学团队、政府研究机构和媒体组织。可见的买方 / 用户 / 付款方模式因细分市场而异,而不是统一模板。在媒体领域,NHL 把采购围绕由媒体运营团队运行的生产和归档工作流展开,服务下游编辑、制作人、播出方,并最终服务 AI 辅助内容发现。在主权 AI 和 AI 云场景中,技术运营方是平台或基础设施团队,但平台采购是为了让企业、研究和公共部门租户消费 GPU 和数据服务。在医疗、基因组学和量化交易里,最终用户是研究人员、科学家和量化人员,中央基础设施预算则为平台买单。因此,VAST 不是卖给单一通用画像;它卖的是深层基础设施层,商业发起方通常是中央 IT 或平台工程,经济合理性来自更上层的数据团队和业务工作流。[CU001, CU002, CU003, CU004, CU024, CU037]

客户细分表
细分市场代表性买方 / 付款方主要用户具名证据用例 / 规模缺口
AI 云 / neocloud云基础设施负责人和平台预算模型构建者和租户 AI 团队CoreWeave;2023 年名单还点名 Lambda 和 Core42AI 云、训练、推理、共享客户交付的主要数据底座直接终端客户收入分成和头部客户组合未披露
电信 / 主权 AISK Telecom 基础设施和国家 AI 计划预算政府、研究和企业 AI 租户SK Telecom主权 GPUaaS / AI 云,1,000 块 Blackwell GPU,5–10 分钟开通没有公开合同金额或续约结构
金融服务中央数据 / 创新 / 交易技术预算银行 AI 团队、资产服务团队、量化研究员HSBC、CACEIS、Jump TradingAI 驱动银行、客户沟通、算法交易、低延迟 HPC大多数留存页面只有标题级信息或量化很轻
医疗健康 / 生命科学研究 IT 和生物信息学基础设施预算科学家、遗传分析师、基因组研发团队Invitae、PacBio、HHS、Boston Children’s 名单提及基因组学、测序、医学研究、公共卫生科学工作流按细分市场划分的客户数未披露
政府 / 研究公共部门和实验室基础设施预算研究人员和 HPC / ML 团队HHS、LLNL、U.S. Air Force、DOE 名单提及、Chan Zuckerberg Initiative药物疗效、公共卫生研究、政府 AI 和科学工作负载具名 logo 的生产深度差异很大
媒体 / 体育联盟媒体运营和内容平台预算编辑、制片人、档案管理员、转播方NHL20+ PB 档案、全部 32 个场馆、实时复制、户外赛事试点没有公开收入或合同期限披露

细分图结合了直接案例研究证据和更广泛名单披露。买方、用户和付款方根据具名部署描述推断,而不是来自单一公司细分披露。

[CU001, CU002, CU003, CU004, CU013, CU024]
FU001: 客户旅程图

VAST 客户通常从识别问题出发,经历技术共创、生产上线,再扩张到相邻工作流。

[CU004, CU006, CU011, CU015, CU043]

6.2 生产采纳轨迹与具名客户证据

最强证据来自 VAST 或客户对已上线部署的描述,并给出范围、时间线或可衡量结果。NHL 关系是最清晰的长期案例:VAST 称该联盟约六年前首次采用平台,把超过 20 PB 存档内容迁入平台,把覆盖范围扩到全部 32 座场馆和总部,并在 2026 年成功试点面向户外赛事的便携节点,安全复制了超过 2 TB 素材。SK Telecom 是最清晰的主权 AI 和电信样本;VAST 接入 Petasus AI Cloud 后,GPU 环境可在约 5–10 分钟内开通,并扩展到 1,000 块 Blackwell GPU 栈。CoreWeave 是押注最大的 AI 云样本:VAST 在 2025 年末公布一项 $1.17 billion 商业协议,官方和第三方来源都把 VAST 描述为 CoreWeave 云里的关键数据底座。除这些旗舰案例外,留存来源还给出 Invitae 和 PacBio 的量化医疗成效,以及 HSBC、CACEIS、Jump Trading、HHS、LLNL 和 Chan Zuckerberg Initiative 等按标题可识别的案例页;合起来看,VAST 的垂直行业确有铺开,只是每页证据深度不一。[CU005, CU006, CU007, CU008, CU009, CU010]

客户增长 / 采用轨迹表
指标 / 里程碑数值 / 状态日期来源依据置信度含义 / 缺失分母
平均初始落地规模初始投资约 $1M2021-05-04Series D 官方发布显示历史企业 ACV 下限,但不是当前组合或中位数
存量客户扩张平均净收入扩张 328%;数个客户 >$10M2021-05-04Series D 官方发布落地后扩张信号很强,但没有当前 NRR/GRR 披露
具名名单广度2023 年官方名单覆盖 Booking、U.S. Air Force、DOE、Verizon、Boston Children’s、Pixar、Zoom 及 AI 云伙伴2023-12-06Series E 官方发布确认垂直行业广度,但不确认付费客户数
CoreWeave 扩张$1.17B 商业协议和扩大的合作关系2025-11-06官方新闻稿加独立新闻最大公开扩张证据也提高集中度风险
SK Telecom 部署成熟度主权 AI 云上 GPU 环境可在 5–10 分钟内开通2025-08-14 / 2025客户页面、博客和新闻稿显示生产就绪度和运营收益
NHL 生产推出全部 32 个场馆加总部使用 VAST;2026 年户外试点成功迁移 >2 TB2025-2026客户页面和 2026 年博客强重复使用信号,但未披露商业支出
医疗健康结果证据Invitae 报告数据访问提升 6x、IOPS 提升 30x;PacBio 无故障增加 2 PB2026客户证据页面部分客户有量化结果,不是全量客户基础

轨迹表混合历史扩张指标、具名部署里程碑和结果披露。它突出公开可支撑的内容,同时保留总客户数、按队列留存、按账户收入占比等缺失分母。

[CU008, CU009, CU012, CU015, CU020, CU021]
具名客户证据表
客户细分市场部署 / 用例生产 vs 试点结果 / 证据质量限制
NHL媒体 / 体育档案现代化、32 个场馆复制、实时媒体工作流、户外赛事节点生产部署,2026 年试点扩展6 年关系、>20 PB 档案、全部 32 个场馆、2026 年试点复制 >2 TB商业价值和续约条款未公开
SK Telecom电信 / 主权 AI面向主权 AI 工作负载的 Petasus AI Cloud / GPUaaS生产级部署5–10 分钟开通、1,000 块 Blackwell GPU、接近裸机吞吐量的声明没有公开合同金额或租户数
CoreWeaveAI 云 / neocloudAI 云和共享客户工作负载的主要数据底座生产 / 扩张账户$1.17B 协议、23 个数据中心、250,000+ 块 GPU、500+ PB、99.9999% 正常运行时间声明VAST 直接收入分成和终端客户归因仍不清楚
Invitae医疗健康 / 生命科学在 VAST Universal Storage 上做遗传学分析生产部署引用数据访问快 6x、IOPS 提升 30x单页结果证据;没有支出或留存数据
PacBio医疗健康 / 生命科学测序和基因组数据平台生产部署又增加 2 PB 且没有问题;称其契合可扩展、经济的 HPC仍是公司托管证据,而不是独立客户证言

列举有意保持部分化,仅限本章留存的最强具名公开证据。还有许多额外 logo,但这些行之所以入选,是因为它们包含生产细节、可衡量结果或新近证据,而不只是 logo 营销。

[CU005, CU006, CU007, CU008, CU009, CU011]
FU002: 采用 / 部署漏斗

从评估到规模化生产:最清晰具名 VAST 客户证据所显示的路径。

[CU006, CU009, CU012, CU015, CU020, CU043]
FU003: 客户证据矩阵

公开客户证据并不均匀;最强证据同时有新近日期、量化结果和明确的生产成熟度。

定性单元格反映本章保留的公开证据质量,而不是内部客户阶段。它们只比较所保留证据的新鲜度和具体度。

[CU023, CU024, CU025, CU026, CU042]

6.3 留存、重复使用与满意度信号

耐久性可以支撑,但证据仍是间接的。VAST 2021 年 Series D 新闻稿仍是最干净的历史扩张数据点:客户初始投入平均约 $1 million,平均净收入扩张 328%,且已有数名客户支出超过 $10 million。较新的公开证据显示的是具名账户中的重复使用和扩张,而不是组合层面的留存指标。NHL 关系从存档现代化扩到实时场馆到总部工作流,又延伸到 2026 年户外赛事试点。CoreWeave 从既有合作演进为披露的 $1.17 billion 协议。PacBio 明确表示,它在 VAST 集群上又增加了 2 PB 且没有问题;Invitae 则提到数据访问快 6x、IOPS 提升 30x。独立满意度证据方向为正但偏薄:FeaturedCustomers 汇总了大量证言和参考评分,PeerSpot 则在小样本中给出 5.0/5 和 100% 推荐意愿。需要警惕的是,留存来源没有披露当前 GRR、NRR、流失、队列续约或头部客户占比,所以公开材料中的耐久性案例偏正面,但不完整。[CU020, CU021, CU022, CU029, CU030, CU031]

留存 / 重复使用 / 满意度表
信号数值 / 状态细分市场 / 账户置信度说明什么尽调需索取
历史扩张平均净收入扩张 328%广泛存量客户基础(2021)早期落地后扩张行为强索取当前按队列和细分市场拆分的 NRR / GRR
大额重复支出到 2021 年数个客户 >$10M广泛存量客户基础部分客户远超首单继续扩张索取当前前 20 大客户扩张曲线和 logo 年份
长期具名关系约 6 年合作NHL支持其不只是试点 logo,而有持续性索取合同期限、续约节奏和年度支出
由伙伴扩张为巨额交易已披露 $1.17B 协议CoreWeave重大扩张证据,但也有集中度风险索取 ARR / bookings 占比和续约时间表
独立评论口碑1,417 份参考评分为 4.8/5;65 份推荐语;37 个案例研究广泛评论 / 参考客户基础第三方满意度信号方向上正面按队列和垂直行业匹配客户参考访谈
当前组合留存指标未公开披露公司整体GRR、NRR、流失和续约存在重大公开缺口索取当前 NRR、GRR、流失、续约和队列桥接表

公开耐久性证据混合了历史公司指标、具名账户重复使用和第三方评论口碑。它不能替代当前队列级留存披露。

[CU006, CU015, CU020, CU021, CU022, CU029]
FU004: 公开持续性记分卡

公开证据显示持续性偏正面,但最强信号是间接的,核心队列指标仍未披露。

该记分卡混合披露指标、具名账户重复使用和评价情绪,是公开证据视角,不是公司披露的留存仪表盘。

[CU021, CU030, CU032, CU040, CU044]

6.4 扩张、集中度、渠道依赖与未解缺口

同一组证明采用的公开证据,也把风险磨得更尖。CoreWeave 同时是 VAST 最好的公开扩张案例,也是最清楚的集中度警示:一份披露的伙伴关联协议就值 $1.17 billion,第三方报道还称 VAST 支撑了 CoreWeave 交付给 Meta、OpenAI 和 Microsoft 的工作负载。VAST 公开材料也把采用放在 NVIDIA、Supermicro、HPE、Lambda 和 Core42 等伙伴框架下,说明伙伴主导的分销对获客和部署在结构上重要。公开来源没给出的东西同样关键:没有留存披露显示当前总客户数、渠道组合、销售周期长度、伙伴抽成率、队列留存或头部客户收入占比。用户指定的 BMW、Nasdaq、Goldman Sachs 和 Dell 也有揭示性。留存的 NVIDIA 关于 BMW 和 Nasdaq 的页面证明了相邻 AI 基础设施可信度,但没有提到 VAST。Sacra 只能证明 Goldman Sachs 和 Dell Technologies Capital 是投资人,而非付费客户。因此,本章可以支撑真实采用、真实扩张和真实伙伴杠杆,但无法为每个头部客户标识补上直接证据,也无法给出承销级客户耐久性所需的集中度数学。[CU018, CU019, CU033, CU034, CU035, CU036]

扩张与集中度风险表
扩张驱动 / 风险公开证据潜在影响置信度尽调路径
落地后扩张动作2021 年官方指标:净收入扩张 328%,数个客户 >$10M账户增长和钱包份额获取的正面指标索取按队列和产品模块拆分的当前扩张
CoreWeave 集中度与一个已披露 AI 云伙伴 / 客户签订 $1.17B 协议可能在单一 neocloud 账户形成收入和续约集中度索取头部客户 ARR / bookings 集中度和合同时间
共享客户依赖SDxCentral 称 CoreWeave 旗下 VAST 支撑 Meta、OpenAI 和 Microsoft 工作负载敞口可能是间接且经伙伴中介,而非直接拥有索取直接终端客户与伙伴导流客户之间的拆分
伙伴主导的市场路径官方资料重点提到 CoreWeave、Lambda、Core42、NVIDIA、HPE、Supermicro扩大触达范围,但也加重对生态优先级和认证的依赖索取伙伴来源订单额占比、销售管线份额和抽成率
采购阻力可见度留存资料没有公开标价、销售周期披露或伙伴经济条款很难判断采购流程在标杆客户之外能否规模化按细分市场索取平均销售周期、折扣惯例和采购卡点
标杆名称证据缺口BMW 和 Nasdaq 只是 NVIDIA 相邻引用;Goldman 和 Dell 出现在投资人名单里,不是留存客户证据如果把生态引用误当成付费客户,客户名录含金量会被高估索取直接客户背书包,包含生产状态、支出区间和背调联系人

本表把真正公开的扩张证据,和影响集中度、渠道依赖投资判断的缺口拆开。最关键的私有材料缺口,是按 ARR / 订单额口径披露客户集中度,以及伙伴来源组合。

[CU018, CU019, CU033, CU034, CU035, CU036]

6.5 证据图表

Chapter 07

07风险

7.1 风险排序与投资含义

VAST 不是丑闻,也不是明显产品失败案例;风险主要是公开披露尚未跟上,而 AI 基础设施估值处在周期高位,承销需要承担不确定性。公司确实有真实证明点:它完成了 $30 billion Series F,称累计预订额超过 $4 billion、承诺 ARR 超过 $500 million,拿下 $1.17 billion CoreWeave 合作,并已通过经过验证的 NVIDIA、Cisco 和 Supermicro 架构出货。但同一组事实也构成最主要的剩余风险栈。第一是估值重置风险,因为价格假设 AI 建设能持久延续,且私人指标会继续被市场接受。第二是 Nvidia、Cisco、OEM 硬件和少数头部 AI 云客户之间的集中度风险。第三是执行宽度,VAST 正从存储拉伸到完整 AI 操作系统和云控制平面。第四是围绕 Red Stapler 和新数据规则的法律及合规暴露。第五是以色列相关的连续性和人才暴露。投资含义:这笔交易仍可尽调,但前提是法律来源、集中度和合规证据从开放式风险变成有边界的风险。[CR001, CR002, CR005, CR006, CR007, CR011]

FR001: 风险热力图

按 2026 年私募市场价格评估 VAST 时,最关键风险的定性严重度图。

这是一张基于现有证据搭建的定性矩阵,不是概率模型。

[CR006, CR018, CR021, CR030, CR036, CR045]
FR002: 风险传导图

VAST 的主要风险如何传导到收入、利润率、融资选项和退出价值。

这张 DAG 表达方向关系,不做数学计算;它用来显示上游小冲击会在哪些位置放大成更大的承销问题。

[CR006, CR021, CR025, CR030, CR036, CR038]

7.2 法律、监管、隐私与主权风险

最清楚的实时法律阴影来自 Red Stapler 链条。NetApp 起诉前 CTO Jonsi Stefansson,指称其滥用商业秘密,事项与 VAST 收购的初创公司有关;佛罗里达起诉虽被驳回且不影响再诉,但驳回依据是审判地问题,NetApp 已上诉。这意味着 VAST 尚未面对已证实的责任事件,但事件也没有彻底熄火。另一个层面是,公开法律面显示,公司通过网站和云栈处理联系人、设备和候选人数据,同时借 Cisco 的医疗和金融服务参考设计更深进入监管行业。因此,2025 年大宗数据制度和更广泛的美国隐私执法,对 VAST 的相关性高于一家简单存储供应商。以色列又加一层。VAST 公开绑定一个以色列主权 AI 云,并运营以特拉维夫为中心的大型研发足迹;以色列法律评论则强调出口管制、两用技术和 IIA 补助限制,这些都可能影响融资、IP 转让和退出。公开证据足以把这些列为真实法律和监管风险,但文档审查前还无法关闭。[CR021, CR022, CR023, CR024, CR025, CR026]

监管 / 法律风险登记表
排名规则 / 许可 / 案件司法辖区状态可能性严重性缓释措施剩余敞口尽调路径
1NetApp / Red Stapler 知识产权纠纷美国 / 冰岛佛州诉状被驳回且可再诉;上诉仍在继续;VAST 未被列为被告对收购代码和人员来源设防隔离;依靠 SPA 赔偿条款和外部律师审查中-高获取收购文件、代码来源备忘录、赔偿条款包,以及冰岛诉讼的任何更新
2DOJ 批量数据规则,以及不断扩张的美国隐私执法美国2025 年起生效;适用于部分供应商、投资人、员工和数据访问流合同控制、数据流映射、最小权限访问和客户专属安全项目可以缓释索取数据流图、子处理方清单、跨境访问控制和隐私合规矩阵
3以色列出口管制 / 军民两用 / IIA 资助限制以色列行业层面的义务明确;VAST 专属资助或许可状态未公开低-中中-高早期法律架构设计和本地合规审查可以约束交易风险确认是否存在 IIA 资助、国防关联客户、出口许可或 know-how 转让限制
4受监管工作负载合规义务多法域公司在营销金融服务和医疗健康用例,但公开证明材料不完整中-高Cisco 零信任叙事和企业架构有帮助,但公开证据仍薄中-高索取当前 SOC 2、ISO 27001、渗透测试摘要和受监管客户审批背书

截至 runDate,本表只覆盖影响 VAST 的主要法律和监管事项中可公开核验的一部分;私有客户审批、资助和内部控制仍不在公开视野内。

[CR021, CR022, CR023, CR024, CR025, CR026]

7.3 运营、质量与依赖风险

运营风险不太来自公开记录里的故障史,而是来自 VAST 有多少关键层并不完全受自己控制。当前产品故事依赖 NVIDIA 加速、Cisco 分销、Supermicro 参考系统,以及来自多家也服务 VAST 竞争对手的 OEM 硬件。CoreWeave 合同和 SDS 主权云建设证明 VAST 能赢下超大部署,但也说明收入和路线图执行可能被少数资本开支密集型对手方绑住;这些对手方自身 GPU 预算、数据中心节奏和采购优先级都可能突然变化。竞争压力让问题更尖锐。在已审阅证据中,Dell 不是伙伴;它是活跃竞争对手,公开攻击 VAST 的电力、空间和基准测试叙事,StorageMath 也批评过 VAST 的一项优化主张。这并不能证明产品弱,但意味着 VAST 在跨关键任务环境扩张时,必须同时守住技术可信度和伙伴协同。[CR007, CR008, CR009, CR011, CR012, CR013]

运营 / 质量 / 安全风险登记表
排名失效模式可能性严重性缓释成熟度剩余敞口未解决缺口
1NVIDIA 路线图或资源分配扰动拖慢 AI OS 部署未公开披露配额保护、替代加速器策略或供给承诺
2Red Stapler 云控制平面整合在法律或工程压力下延期低-中中-高未公开代码来源审查或收购后整合 KPI 包
3安全与合规证明落后于受监管客户拓展目标低-中中-高留存资料未显示当前公开认证或审计材料
4大型主权云或新型云厂商部署会带来可靠性和多租户复杂度中-高留存资料未发现公开 SLA、事故日志或复盘材料
5营销和基准测试争议削弱竞争性交易中的技术可信度公开记录中,对争议主张的独立技术验证有限

本登记表聚焦外部可观察的运营模式;没有公开宕机历史,不等于没有故障,仍作为尽调缺口处理。

[CR013, CR014, CR015, CR017, CR019, CR020]
伙伴 / 依赖风险登记表
排名依赖项交易对手角色集中度失效情景严重性缓释措施剩余敞口
1加速 AI 栈NVIDIAGPU、库、认证和参考架构骨干关键NVIDIA 路线图或配额变化削弱 VAST 现有 AI OS 定位认证设计和深度合作
2标杆新型云厂商收入CoreWeave主要披露的大型交易和云分发锚点资本开支暂停、续约压力或重新定价,会打击增长观感和订单集中度多年合作和技术整合
3企业渠道与采购CiscoGPL 路径、已验证栈和企业触达中-高Cisco 优先推进既有替代方案,或联合栈在一线落地不及预期直接进入 GPL,并获得联合方案支持中-高
4已验证硬件平台Supermicro 加 OEM 栈参考系统和部署硬件中-高硬件约束、规格变化或 OEM 冲突拖慢部署中-高采取多 OEM 姿态,而非依赖单一整机
5以色列主权云足迹SDS 和以色列本地运营区域旗舰部署和本地执行基地地缘政治扰动或本地劳动力压力拖慢交付和支持中-高远程优先组织形态和全球客户基础

交易对手集中度按已披露战略重要性排序,而不是按未披露的收入占比;头部客户和头部伙伴经济性仍属私有信息。

[CR007, CR008, CR011, CR012, CR013, CR014]
FR003: 依赖图

当前 VAST 技术栈依赖外部 GPU、渠道、硬件、云和区域执行节点,而不是一个单一自洽的平台。

这张依赖图是敞口图,不是完整伙伴名单;它突出证据支持下风险传导最强的节点。

[CR007, CR011, CR013, CR014, CR017, CR027]

7.4 财务、商业与竞争风险

公开记录中的财务模型风险不是现金即将吃紧,而是估值脆弱性和集中度脆弱性。VAST 的融资能力显然很强,但本轮结构混合新股和老股,所以头条规模夸大了新增经营现金。与此同时,最强的规模指标仍是公司口径,未公开勾稽。这一点重要,因为宏观背景已不再一边倒支持。Gartner 仍预测 2026 年 AI 基础设施支出巨大,但也称该行业处在幻灭低谷,且既有厂商会捕获大量企业 AI 需求。Goldman 把建设争论定义为对假设敏感,而非板上钉钉。两件事可以同时成立:支出可以增长,而市场对高溢价私营估值的容忍度下降。对 VAST 来说,下行情景不只是收入增长放慢;它还包括增长放慢叠加 Dell 等既有厂商更强竞争,以及 VAST 填上披露缺口前公开市场倍数先重置。[CR001, CR002, CR004, CR005, CR006, CR018]

7.5 人员、执行、缓释因素与否决标准

人员风险有分量,因为 VAST 的市场叙事仍高度依赖创始人 CEO Renen Hallak,而云控制平面扩张路径又依赖 Red Stapler 团队;其加入正是 NetApp 纠纷触及的领域。还有后期私营公司的典型尽调限制:公开来源集仍没有给出头部客户集中度、续约排期、债务、当前安全认证或清晰的资金管理视图。好消息是,部分缓释因素确实存在。VAST 融资通道很深,有 Cisco 和 Supermicro 的伙伴验证,也在大型 AI 环境中有可信采用。坏消息是,这些缓释因素本身无法回答对投资逻辑最敏感的问题。因此,投资人应把几件事列为明确的签约后监控项,而不是模糊观察点:任何把 VAST 直接拖入 Red Stapler 争端的新诉讼进展、客户集中度比预期更差的证据、无法为监管工作负载提供当前合规证明,或任何暗示估值重置的融资事件。这些否决标准比泛泛担心竞争或宏观噪音更清楚。[CR022, CR037, CR038, CR039, CR042, CR048]

人员 / 执行风险登记表
排名角色 / 职能依赖或缺口可能性严重性缓释措施尽调路径
1创始人 / CEO商业、产品和资本叙事仍高度围绕 Renen Hallak补强高管梯队厚度和董事会流程索取继任计划、董事会委员会结构和升级路径图
2云控制平面领导层Red Stapler 团队是超大规模云野心的核心,也贴近正在发酵的法律叙事隔离 IP 审查,并建立整合治理索取收购后路线图里程碑和代码所有权审查
3合规与安全市场进入受监管行业推进速度,看起来快于公开可见的合规证明中-高借助 Cisco 和伙伴验证,同时搭建直接控制体系索取当前认证、客户问卷和安全人员配置计划
4董事会与披露深度以 $30B 公司标准看,公开披露中的董事会构成、债务和资金管理细节仍偏薄中-高后期公司 CFO 纪律和投资人流程索取董事名单、债务明细表、现金政策和治理日历
5以色列 R&D 和人才集中以色列工程团队占比较大,面临宏观劳动力和地缘政治压力远程优先组织形态和全球商业覆盖按地域索取组织架构图、流失趋势和业务连续性计划

本表按单一团队或控制缺口对路线图交付、融资能力或受监管客户转化的直接影响排序。

[CR021, CR022, CR029, CR030, CR037, CR038]
缓释与叫停标准表
风险可监测触发点阈值 / 事件行动含义
估值重置融资或 IPO 信号降价轮、结构化救助融资,或 IPO 定价显著低于近期私募估值标记暂停新增资本部署,并按公开可比公司倍数重新测算
客户集中度尽调披露头部客户或头部伙伴占比显著超过投资假设,尤其是围绕 CoreWeave降低持仓规模,或要求更强下行保护
法律 / IP 阴影案件状态VAST 被直接点名、面临禁令风险,或无法证明代码来源干净在法律风险量化前,视为投资逻辑破裂
安全 / 合规缺口认证证据公司无法在尽调窗口内拿出受监管工作负载的当前证明材料仅按非受监管增长情景测算,或直接停止
伙伴依赖路线图 / 渠道延误NVIDIA、Cisco 或 OEM 路线图延误,使重大部署推迟超过一个规划周期下修增长和利润率预期
以色列连续性风险人员与运营韧性以色列领导层出现显著流失,或主权云中断且没有快速后备方案重新给执行风险定价,并在交割前要求连续性计划

这些触发点旨在尽调期和交割后第一年内可监测,而不是依赖模糊的宏观情绪。

[CR036, CR042, CR045, CR046, CR047, CR048]

7.6 证据图表

Chapter 08

08估值

8.1 建议取决于价格,不是公司质量判决

VAST Data 更像真实品类领导者,而不是纯炒作资产。作为一家私营存储与数据平台,公司把 AI 基础设施相关性、头部客户、快速增长和正面利润率信号组合得异常强。问题不在业务是否重要;问题在于,新资金以 $30 billion 估值入场,是否还能仅凭公开证据赚到有吸引力的回报。在官方披露的超过 $500 million 承诺 ARR 下限上,本轮价格隐含约 60x。即使用较早 Reuters 式的 $600 million ARR 预测,也仍约 50x。只有最激进的外部估计——约 $2 billion ARR——才把入场价格拉低到约 15x,但相对上市存储可比公司仍偏贵。因此,建议对价格和证据都敏感:新资金应观察或继续研究;只有成本基础显著更低时才持有;不要把强公司和合理入场价混为一谈。[CV001, CV007, CV009, CV019, CV020, CV021]

建议摘要表
维度当前判断证据基础决策含义
建议新资金应观察 / 继续研究;只有低成本老股持有人适合持有新资金测算只有在乐观情景下才过关,但公司质量信号仍强没有私有尽调包或更好进入价格,不应按本轮价格新进。
信心核心融资事实真实,但收入定义和优先股堆叠缺口仍未解决将该判断视为证据敏感,而非最终结论。
风险评级当前价格需要溢价增长假设、AI 资本开支持续支撑,以及干净的集中度动态先测算下行情景。
估值立场昂贵本轮估值意味着约 60x 官方 CARR、约 50x Reuters 式 ARR 预测,以及约 15x 激进外部 ARR 估计当前价格已经计入成功,但公开证据尚未充分证明。
目标回报 / 持有 / 退出新资金要实现 2x 总回报需要约 $60B,3x 总回报需要约 $90B;现有持有人可为 IPO 可选性继续持有对照当前公开可比公司倍数,回报门槛很难站住新买方需要估值重置,或强得多的证据。

本表明确对价格敏感。回报门槛按 $30B 本轮估值做算术推导,未调整未来稀释或融资条款。

[CV019, CV020, CV021, CV044, CV050, CV051]
FV001: 推荐逻辑

推荐判断基于真实产品和需求证明;这些证据还要跨过未解的收入桥,最后接受昂贵价格测试。

这条流程是分析框架,不是机械模型。它说明推荐结论如何来自证据质量和价格支撑,而不是解释 VAST 产品在技术上如何运作。

[CV007, CV014, CV024, CV027, CV030, CV044]

8.2 融资背景、持仓估值与入场纪律

融资记录方向上有支撑,但在承销最关键的地方仍不完整。2026 年 3 月报道提前浮出这轮融资;公司在 2026 年 4 月 22 日正式披露 Series F 完成。管理层确认约 $1 billion 交易混合新股和老股;外部报道暗示早期持有人流动性可能约占头条金额的一半。这一点重要,因为老股占比重的一轮,对新增资产负债表实力的信号弱于纯新股融资。Fidelity 组合文件显示,VAST 优先股的独立持仓估值从 2024 年 8 月约 $18.16 / 股升至 2026 年 2 月约 $59.22 / 股,所以估值跃升并非凭空而来。但同一组来源仍未披露确切的优先股堆叠、稀释或收入定义桥。入场纪律必须简单:假设公开记录支持低于本轮的估值,除非尽调证明经常性收入更接近高端估计,且后期融资条款干净。[CV004, CV005, CV006, CV017, CV018, CV019]

投资逻辑 / 反向逻辑表
视角投资逻辑反向逻辑什么会改变判断
市场AI 基础设施需求真实,VAST 正处在大型建设周期里。如果泡沫叙事占上风,或利用率不及预期,AI 资本开支倍数可能急剧压缩。用客户续约证明需求持久,并在 AI 支出走弱时仍保持韧性增长。
产品统一的 AI Operating System 可以支撑相对传统存储厂商的溢价。如果产品广度不能转化为可变现的经常性收入,这个溢价就不值得投。披露模块级变现和附加率。
客户标杆客户和伙伴支撑战略相关性。公开资料没有量化集中度,因此一两家大型新型云厂商可能主导经济性。分享头部客户 ARR 和续约时间表。
财务官方信号显示 CARR 超过 $500M、订单额超过 $4B,并声称利润率 / 现金流为正。即便如此,距离支撑当前价格所需的约 $2B ARR 估计,仍有很大缺口。以审计形式打通订单额、CARR、ARR 和确认收入。
竞争如果 VAST 真是 AI 基础设施平台,而不是存储盒子公司,就应高于 NetApp 和 Pure 交易。除非高端 ARR 估计成立,否则即便是高溢价公开可比公司,交易倍数也远低于 VAST 隐含倍数。在大得多的规模上,证明软件式耐久性和利润率质量。
融资 / 治理Fidelity 独立估值标记支撑估值确有上台阶。老股占比较高的结构和未披露的优先条款削弱价格信心。提供新股 / 老股拆分和完整股权结构瀑布。

每一行都把真实利好与相应投资测算反对点配对。只有满足「什么会改变判断」这一列,本章建议才会上调。

[CV018, CV024, CV027, CV030, CV035, CV036]
FV002: 估值敏感性

估值争论最敏感的变量,是收入定义清晰度和公开市场倍数选择;客户集中度和倍数压缩是主要负向抵消。

数值是相对十几 B 美元中段基准情景的方向性十亿美元级增减,用来标出哪些假设最关键,而不是给出精确 DCF。

[CV019, CV021, CV024, CV027, CV030, CV038]

8.3 情景与可比估值锚

公开可比公司把估值问题框得很清楚。Everpure,即更名后的 Pure Storage,约按 7.7x EV / 过去十二个月收入交易;NetApp 约 4.1x;CoreWeave 是最清楚的公开 AI 基础设施溢价参照,约按 10.9x EV / 一季度年化收入交易,同时杠杆重得多。Lambda、Crusoe 和 Together AI 等私营 AI 基础设施融资,确认投资人愿意为 GPU 云和 AI 平台资产付高价,但这些里程碑没有一个能填上 VAST 官方 $500 million-plus CARR 下限与 $30 billion 价格之间的缺口。因此基准情景低于本轮估值,因为中点收入估计叠加公开 AI 基础设施溢价后,估值仍落在百亿美元级的低十几到高十几区间。乐观情景只有在约 $2 billion ARR 估计方向正确、集中度可控、公开市场仍愿给 AI 基础设施高溢价倍数时,才到 $20B 高段至 $30B 中段。悲观情景则在 AI 支出放缓或收入桥接失败时迅速回落到上市存储公司的区间。[CV024, CV027, CV030, CV031, CV032, CV033]

乐观 / 基准 / 悲观情景表
情景假设估值区间(USD B)相对 $30B 进入价的隐含结果概率信号 / 触发点
乐观公开高端 ARR 估计(~$2B)方向正确;集中度可控;在 IPO 市场开放时,VAST 获得 14x-18x 溢价倍数。28-36新买方持平到小幅正回报;对低成本持有人有吸引力需要私有尽调验证收入规模和退出准备度。
基准标准化经常性收入能力约为 $1.3-$1.5B;VAST 可相对 Pure / NetApp 获得溢价,但不是泡沫倍数。13-21低于当前轮估值;新进不具吸引力最符合当前公开记录。
悲观可持续经常性收入更接近 $0.8B;AI 倍数向公开存储公司区间压缩;集中度或退出时点恶化。6.4-8.0相对本轮价格大幅减记收入桥接失败、客户集中或 IPO 窗口关闭会触发。

情景数值是估计股权价值区间,不是谈判后的条款清单结果。刻意使用宽区间,因为公开记录不支持精确到点。

[CV041, CV042, CV043, CV044, CV045, CV051]
可比估值表
可比公司状态当前价值收入锚点隐含倍数 / 里程碑参考价值局限
Pure Storage / Everpure上市公司EV ~$28.0B,2026-05-27TTM 收入 ~$3.66B~7.7x EV / 收入最适合作为上市存储平台高溢价可比公司,利润率和现金状况都不错但比 VAST 更成熟,AI 原生程度也更低。
NetApp上市公司EV ~$27.1B,2026-05-27TTM 收入 ~$6.70B~4.1x EV / 收入可作为存储老牌公司的下限参考传统业务组合和较慢增长可能低估 AI 平台上行空间。
CoreWeave上市公司EV ~$90.7B,2026-05-27Q1'26 收入年化 ~$8.3B~10.9x EV / 年化收入最接近的上市 AI 基础设施高溢价参考杠杆很高、耗电重;不能直接做完全可比公司。
Lambda私募轮次$480M Series D 轮;Reuters 报道估值 ~$2.5B公开收入未披露仅作里程碑参考说明投资人愿意为 AI 云容量支付高价目前没有收入或利润率披露。
Crusoe私募轮次$1.375B Series E 轮,估值 >$10B公开收入未披露仅作里程碑参考垂直整合 AI 基础设施的大额融资估值没有配套公开的经常性收入细节。
Together AI私募轮次$305M Series B 轮,估值 $3.3B公开收入未披露仅作里程碑参考另一个云 / AI 平台估值标尺模型组合和财务画像与存储基础设施不可比。

上市公司行使用 runDate 市场数据;CoreWeave 收入以 2026 年 Q1 年化,仅用于方向性比较; 私募行只是里程碑参考,因为当前收入未公开披露。这里列的是最相关、可观测可比对象的一部分, 并非穷尽晚期 AI 基础设施公司。

[CV024, CV027, CV030, CV031, CV032, CV033]
FV003: 估值 / 回报区间

情景区间显示,公开证据支持的基准情景低于本轮价格;从 $30B 入场,新资金回报门槛很高。

情景区间是估算股权价值。回报门槛行只是基于本轮价格的算术,用来说明为什么估值纪律重要。

[CV041, CV042, CV043, CV051, CV052]

8.4 退出准备度、投资逻辑破裂触发点与最终尽调事项

最能支撑的退出路径是最终 IPO,但管理层自己表示 VAST 尚未启动正式 IPO 流程,也尚未聘请投行。这意味着投资人不应把近期上市退出当作已经启动来建模。公开记录也留下了最常压垮后期交易的问题:收入中有多少来自少数超大规模云厂商或 neocloud 账户,这些合同有多可重复,Series F 后股权结构表是什么样,审计后的报告系统是否经得起公开市场审视。在这个估值上,这些不是装饰性细节。如果 ARR 定义无法勾稽,如果客户集中度很高,或如果公司在要求上市公司式定价时仍远未准备好 IPO,投资逻辑应立即破裂。因此,尽调议程很直接:先要私人指标桥、集中度表、融资瀑布和 IPO 准备路线图,再把 $30 billion 这一轮视为可投资,而不是仅仅令人印象深刻。[CV039, CV040, CV045, CV046, CV048, CV049]

投资逻辑失效与终止触发条件表
触发条件阈值为什么会打破投资逻辑行动含义
收入桥接失效经审计 ARR 或经常性收入明显低于 ~$1.5B当前轮价格成立的前提是规模远高于官方 $500M CARR 下限停止交易,或重新定价。
客户集中度过高前一两大客户主导经常性收入或即将续约的收入如果需求面窄且不稳定,晚期溢价倍数应被压缩提高下行情景权重,或退出。
AI 倍数压缩上市 AI 基础设施可比公司估值向传统存储区间下修VAST 已按溢价倍数定价价格重置前避免新进入。
IPO 准备停滞虽然定价已接近上市公司高溢价水平,但没有投行、内控或报告路线图持有期拉长,估值支撑同步走弱按更长久期私有资产处理,或拒绝。
优先股悬顶超预期Series F 清算瀑布或反稀释条款大幅压低普通股价值名义估值不再对应投资人经济收益从股权结构表重建回报模型,而不是从名义估值出发。

终止触发条件故意设得具体、以交易为导向。它们关注的是会推翻当前定价的情形, 而不是只会拖慢增长的情形。

[CV036, CV038, CV046, CV048, CV049]
最终尽调索取清单表
主题缺失证据为什么重要负责人 / 尽调路径
收入桥接按年度和头部客户拆分的预订额 → CARR → ARR → 确认收入勾稽这是公允价值里最大的摆动因素财务尽调 / CFO 材料包。
新股与老股占比已签署融资摘要,列明准确新增现金和卖方流动性决定稀释幅度,也决定现金跑道实际改善多少法务 + 投资者关系尽调。
股权结构表 / 优先权Series F 条款清单、清算瀑布和完全稀释后持股名义估值可能高估普通股价值法务尽调 / 董事会材料。
客户集中度前 20 大 ARR、续约日期、利润率画像和超大规模云厂商敞口集中度可能迅速压垮乐观情景商业尽调 / 收入运营导出。
IPO 准备度内控、审计状态、董事会构成和投行准备里程碑IPO 是最站得住的退出路径,但尚未正式启动CEO/CFO 尽调 + 治理审查。

这些问题最能快速改变建议。没有这些证据,公开记录足以让人认可公司, 但不足以支撑以 $30 billion 估值新进场。

[CV037, CV039, CV048, CV049]
FV004: 投资 KPI

供 IC 使用的记分卡显示:市场和产品证明很强,证据质量居中,而当前入场价格下估值支撑偏弱。

分数是 0-10 分委员会式启发判断,用来概括证据质量和价格支撑,不是加权算法。

[CV035, CV036, CV044, CV045, CV047, CV049]

8.5 证据图表

免责声明

本报告仅供信息参考,基于截至 2026-05-27 可获得的公开来源。VAST Data 是一家公开披露有限的私营公司,因此若干财务、治理和客户集中度结论依赖管理层表述和第三方报道,而非经审计文件。情景分析和估值观点仅作说明,不构成投资建议。

证据索引

结论
编号陈述可信度来源
CO001 VAST Data was founded in 2016. SO007, SO026
CO002 VAST describes itself as an AI Operating System company and a global data platform for agentic and data-intensive applications. SO001, SO029
CO003 VAST says its platform natively unifies storage, database, and compute in one system. SO011, SO029
CO004 VAST DataStore is marketed as a universal data store that eliminates storage tiers and scales from terabytes to exabytes. SO005, SO018
CO005 VAST attributes its platform economics and scale claims to its DASE architecture. SO007, SO008, SO025
CO006 Recent official materials anchor VAST to New York City while also branding the company as remote-first or no-headquarters. SO007, SO011, SO008
CO007 Reuters-described coverage characterized VAST as New York-headquartered in 2025. SO028
CO008 After closing its Series F financing, VAST remained a private, late-stage company rather than a public issuer. SO007, SO016
CO009 Renen Hallak is VAST's founder and CEO or co-founder and CEO in retained public sources. SO007, SO011, SO016
CO010 Jeff Denworth is a co-founder who remains a recurring public spokesperson for VAST's platform strategy. SO011, SO017, SO023
CO011 Amy Shapero is VAST's first CFO and joined from Shopify as the company prepared for public-company readiness. SO016, SO028
CO012 Jonsi Stefansson became VAST's general manager of cloud solutions after VAST bought Red Stapler and hired its team in 2025. SO020, SO021, SO009
CO013 Fidelity's 2023 investment gave it a board observer role rather than a publicly disclosed board seat. SO025
CO014 Hallak is the main public face across VAST's financing, product-positioning, and IPO-readiness disclosures in the retained source set. SO007, SO016, SO028
CO015 VAST appears to use multiyear licensing, OEM relationships, resellers, and cloud partners as major routes to market rather than a purely direct sales model. SO016, SO018, SO026
CO016 VAST raised $83 million in Series D in May 2021 at a $3.7 billion post-money valuation led by Tiger Global Management. SO010, SO019
CO017 VAST said NVIDIA also participated in the Series D round. SO010, SO019
CO018 VAST said its balance sheet was $230 million strong after adding the Series D proceeds to prior rounds. SO010, SO019
CO019 VAST said it exited its second year with nearly $100 million of annualized software run rate and cash-flow positivity around the Series D announcement. SO010, SO019
CO020 VAST raised $118 million in Series E in December 2023 at a $9.1 billion valuation led by Fidelity Management & Research Company. SO011, SO023, SO024
CO021 NEA, BOND Capital, and Drive Capital also participated in the Series E round. SO011, SO024
CO022 Independent coverage said part of the Series E proceeds would support a stock buyback or secondary sale for existing shareholders. SO023, SO024
CO023 At the end of fiscal Q3 in 2023, VAST said cumulative software bookings had surpassed $1 billion. SO011, SO023, SO024
CO024 VAST said late-2023 growth was 3.3 times year over year with nearly 90 percent gross margin and positive cash flow. SO011, SO023, SO025
CO025 VAST disclosed more than 700 employees worldwide in late 2023. SO011, SO025
CO026 Enterprise Times reported that VAST had offices in the US and Israel plus an office in London in late 2023. SO025
CO027 VAST closed its Series F financing in April 2026 at a $30 billion valuation. SO007, SO014, SO016
CO028 The Series F transaction value was approximately $1 billion and included both primary and secondary capital. SO007, SO014, SO015
CO029 Drive Capital led the Series F round and Access Industries acted as co-lead. SO007, SO014, SO016
CO030 Fidelity, NEA, and NVIDIA were named as participating investors in the Series F financing. SO007, SO014, SO016
CO031 Combining the roughly $381 million raised before Series F with the approximately $1 billion 2026 transaction implies total capital raised of about $1.38 billion. SO024, SO026, SO028, SO007
CO032 By April 2026, VAST said it had surpassed $4 billion in cumulative bookings and exited the prior fiscal year with more than $500 million in committed annual recurring revenue. SO007, SO014, SO016
CO033 Hallak told CRN that VAST was GAAP profitable and cash-flow positive. SO016
CO034 Reuters-reported coverage said VAST had reached $200 million in ARR by January 2025 and had internal projections to reach $600 million the following year. SO028
CO035 Official 2026 financing materials said thousands of organizations rely on VAST, including CoreWeave, Lowe's, the U.S. Air Force, Cursor, JPMorganChase, Crusoe, and Mistral AI. SO007, SO013
CO036 Official late-2023 materials listed Booking Holdings, the U.S. Air Force, the U.S. Department of Energy, Verizon, Boston Children's Hospital, Pixar, and Zoom as customer examples. SO011, SO025
CO037 TechCrunch reported that VAST's customers include Pixar, ServiceNow, xAI, CoreWeave, and Lambda. SO022
CO038 VAST's customer stories page highlights the NHL, SK Telecom, NVIDIA-linked cybersecurity work, and animated-feature production use cases. SO004
CO039 Fierce Network reported VAST cloud partnerships across AWS, Google Cloud, Microsoft Azure, CoreWeave, Crusoe, Lambda, and Core42. SO026
CO040 Supermicro markets the VAST platform as a high-performance, scalable data management solution optimized for AI and GPU cloud markets. SO018
CO041 StorageReview reported that VAST planned to deliver CNode-X servers through OEM partners such as Cisco and Supermicro. SO017
CO042 VAST launched the Amplify program in January 2026 and said it can deliver up to six times more effective capacity from existing SSD estates depending on workload. SO008
CO043 At VAST Forward in February 2026, VAST launched Polaris as a global control plane for distributed AI infrastructure across cloud and on-premises environments. SO009, SO017
CO044 StorageReview said the PolicyEngine and TuningEngine were slated for release by the end of 2026. SO017
CO045 The NetApp and Red Stapler dispute created reputational and integration risk for VAST even though VAST itself was not named as a defendant in the Florida complaint. SO020, SO021
CO046 NetApp's Florida lawsuit against Stefansson was dismissed on jurisdictional grounds, but NetApp said it had appealed and was pursuing related action in Iceland. SO020
CO047 Calcalist reported that the Red Stapler dispute arrived while VAST was trying to raise a new round at roughly $28 billion to $30 billion. SO021
CO048 Hallak said in April 2026 that VAST had not started an IPO process even though internal preparation was underway. SO016, SO028
CO049 The retained source pack does not disclose an exact current customer count despite company language about thousands of organizations and multiple named logos. SO007, SO013, SO022
CO050 The retained source pack does not disclose a precise 2026 employee count beyond the more-than-700 figure published in late 2023. SO011, SO025
CM001 VAST positions its AI Operating System as a unified layer for storage, database, and compute aimed at agentic and other data-intensive applications. SM001
CM002 VAST says AI adoption is constrained by exploding data volume, outdated compute architectures, and the high cost of effective AI. SM002
CM003 VAST markets the platform as enterprise scale-out all-flash infrastructure for both structured and unstructured workloads that removes storage tiering and HDD dependence. SM015
CM004 VAST says modern AI workloads require storage, networking, and compute to operate as a tightly coupled data-centric fabric across datacenters. SM016
CM005 VAST says its platform covers file, object, block, tabular, vector, and streaming data through the DataStore, DataBase, and DataSpace services. SM016
CM006 VAST says its DASE architecture disaggregates compute from physical storage and scales from terabytes to exabytes over NVMe-over-Fabrics. SM016, SM018
CM007 VAST argues that a unified control plane can replace separate object stores, analytics clusters, orchestration layers, and message brokers in large AI environments. SM016
CM008 VAST explicitly frames AI, compliance, containers, data analytics, backup and recovery, and HPC as solution areas around the platform. SM015
CM009 IDC projected the external OEM enterprise storage market to reach about $37.7 billion in 2026 after roughly $35.4 billion in 2025 and $33.6 billion in 2024. SM004
CM010 IDC highlighted all-flash arrays as a key growth driver because enterprises are shifting toward high-performance flash for AI, analytics, and high-throughput processing. SM004
CM011 Gartner forecast total AI spending of $2.528 trillion in 2026, including $1.366 trillion for AI infrastructure and $3.119 billion for AI Data. SM003
CM012 Gartner said AI would sit in the Trough of Disillusionment through 2026 and that enterprise scale depends on improved predictability of ROI. SM003
CM013 Fortune Business Insights sized the AI-powered storage market at $44.94 billion in 2026 after valuing it at $35.90 billion in 2025. SM005
CM014 Fortune Business Insights attributed AI-powered storage growth to high-throughput data access, large datasets, HPC demand, and cloud-service adoption. SM005
CM015 Fortune Business Insights said file-based architectures historically dominated the category while object-based storage is expected to grow faster for AI and ML data sets. SM005
CM016 Fortune Business Insights grouped AI-powered-storage buyers into enterprises, telecom companies, cloud service providers, and government bodies, with enterprises accounting for 30% share in 2026. SM005
CM017 Research and Markets valued the broader AI infrastructure market at $90.91 billion in 2026 and structured it by offerings, function, technology, deployment type, and end user. SM006
CM018 Coherent Market Insights valued AI infrastructure at $90 billion in 2026 and said the category is 54% hardware, 46% on-prem, 48% enterprise, and 40% North America. SM007
CM019 Public 2026 market estimates conflict mainly because they measure incompatible scopes ranging from OEM storage systems to AI-powered storage to full AI infrastructure. SM003, SM004, SM005, SM006, SM007
CM020 No public source cleanly isolates a VAST-specific SAM or SOM, so the defensible public framing is a multi-lens range rather than a single TAM headline. SM004, SM005, SM006, SM007
CM021 VAST’s closest public-market fit is the shared data-infrastructure layer inside enterprise AI factories rather than total AI infrastructure or general AI software spend. SM001, SM016, SM018
CM022 VAST and NVIDIA position the platform for accelerated analytics, generative AI, agentic AI, and small-scale HPC inside enterprise AI factories. SM014, SM018
CM023 The VAST reference architecture says enterprises can begin with four-node scalable units and expand as AI demand grows. SM018
CM024 The VAST reference architecture lists faster time-to-token, resource efficiency, lower TCO, and reduced deployment risk as core business benefits. SM018
CM025 The VAST reference architecture says DASE can deliver enough parallelism to challenge the assumption that NFS is inadequate for AI and HPC workloads. SM018
CM026 Supermicro and VAST market CNode-X as a turnkey enterprise AI data platform that bundles compute, storage, and software for AI factory deployment. SM017
CM027 Supermicro says the joint platform is designed to keep GPUs fed with data and move teams faster from deployment to first token. SM017
CM028 IDC buyer-shift research says 2026 buyers demand precise, measurable ROI and time-to-impact early in the buying process. SM019
CM029 IDC buyer-shift research says procurement, finance, and revenue operations have become central decision-makers rather than supporting stakeholders. SM019
CM030 IDC buyer-shift research says boards and budget owners increasingly expect scenario-based planning and dynamic TAM-SAM-SOM-style forecasting before qualifying vendors. SM019
CM031 Forrester says the typical 2026 business buying decision includes 13 internal stakeholders and nine external influencers. SM020
CM032 Forrester says procurement is a decision-maker in 53% of business buying cycles. SM020
CM033 Forrester says more than 60% of buyers use a trial and 78% of buyers spending more than $10 million engage in a trial first. SM020
CM034 Deloitte says worker access to AI rose 50% in 2025 and the share of companies with at least 40% of projects in production is set to double within six months. SM009
CM035 Deloitte says the AI skills gap remains the biggest barrier to integration and only one in five companies has mature governance for autonomous AI agents. SM009
CM036 Informatica says 57% cite data reliability as a top barrier, three out of four say governance has not kept pace, and 86% plan to increase data-management investment. SM012
CM037 Electronic Design and SNIA say storage now directly shapes AI performance, efficiency, power, cooling, rack density, and time-to-market. SM011
CM038 Avnet says enterprise SSD demand is expected to grow 41% and AI data centers could consume about 70% of high-end DRAM in 2026. SM008
CM039 Avnet says DRAM and NAND tightness could persist through 2027 as capacity expansion lags demand. SM008
CM040 NetApp says organizations are increasingly choosing AI-ready data infrastructure and that its all-flash business reached a $3.6 billion annualized run rate in Q1 FY26. SM021
CM041 Pure says AI-era competitive advantage depends on data accessibility and breaking data free from application silos. SM023
CM042 Pure’s subscription ARR rose from $1.7 billion in Q1 FY26 to $1.8 billion in Q3 FY26, showing incumbents are monetizing AI-era data-platform demand. SM022, SM023
CM043 Cloudera says 2026 is when enterprises move from experimentation to operational intelligence orchestration and need unified control planes, governance, and hybrid execution. SM024
CM044 MinIO argues legacy storage architectures can become the primary bottleneck for enterprise AI and that scalable object storage is well-suited to distributed AI workloads. SM013
CM045 INFUSE says buyer confidence continues to erode despite AI-assisted research and that vendors must demonstrate value quickly and precisely. SM025
CM046 Public evidence still lacks VAST pricing, customer count, conversion rates from trial to production, and workload-level revenue mix, which limits precise SOM modeling. SM019, SM020
CP001 VAST publicly presents itself as an AI Operating System that unifies storage, database, and compute for agentic and data-intensive applications. SP028
CP002 VAST’s platform overview says its architecture is meant for petabyte- and exabyte-scale datasets with high-performance random I/O. SP001
CP003 Pure FlashBlade//S is marketed as a scale-out unstructured storage platform with native file and object protocols for analytics, AI/ML, and HPC workloads. SP002
CP004 Pure’s Evergreen//One commercial model is consumption-based, SLA-backed, and built around paying for used capacity rather than provisioned capacity. SP003
CP005 Pure said fiscal 2025 revenue surpassed $3 billion and grew 12% year over year, giving it public-company scale beyond most private specialists. SP004
CP006 NetApp markets a unified storage environment spanning on premises, public clouds, and hybrid cloud under an ONTAP-led portfolio. SP005
CP007 NetApp Keystone sells block, file, and object storage through pay-as-you-go terms and cloud bursting rather than only fixed hardware purchases. SP006
CP008 WEKA markets NeuralMesh as software-defined AI infrastructure built on containerized microservices, true multi-tenancy, and independent service scaling. SP007
CP009 WEKA’s 2026 newsroom highlights NVIDIA-linked AI data platform solutions and regional partners, implying strategic emphasis on AI factory delivery and channel expansion. SP008
CP010 Globes reported that WEKA raised $140 million at a $1.6 billion valuation in 2024 and had raised about $400 million in total. SP009
CP011 DDN positions itself as a data intelligence platform for AI factories, inference, sovereign AI, hyperscalers, and regulated workloads, not only classic HPC storage. SP010
CP012 DDN says 8 of 10 leading automotive companies and 7 of 10 top banking and securities firms trust it, signaling enterprise reach in demanding verticals. SP010
CP013 DDN markets Infinia as an end-to-end AI data platform across cloud, core, and edge and says it delivers five nines of availability. SP011
CP014 DDN says Infinia can fit 100 PB of storage in a single rack, which is a scale and density claim aimed squarely at AI factory economics. SP011
CP015 IBM Storage Scale is software-defined storage for AI, HPC, and analytics and unifies file and object data across data centers, cloud, and edge. SP012
CP016 IBM’s Storage Scale System 6000 page lists NVIDIA-certified positioning and up to 340 GB/s throughput, reinforcing IBM’s enterprise-AI performance story. SP013
CP017 Hammerspace Tier 0 uses GPU server-local NVMe as shared storage and says it can be activated without proprietary clients or agents. SP014
CP018 Hammerspace’s own competitive brief says VAST has no Tier 0 equivalent and that VAST cloud clusters remain limited on performance and scale. SP014
CP019 TechCrunch reported that Hammerspace raised $100 million at a valuation above $500 million and named Meta and the U.S. Department of Defense as customers. SP015
CP020 Qumulo describes itself as a modern file and object data platform spanning edge, data center, and cloud with a global namespace and AI-assisted visibility. SP016
CP021 Qumulo’s cloud packaging mixes a fully managed Azure Native offer with self-hosted AWS and Google Cloud deployments. SP017
CP022 AWS FSx for Lustre uses usage-based pricing with separate charges for storage, throughput, metadata IOPS, backups, and some data transfer. SP018
CP023 AWS S3 exposes explicit usage-based pricing for storage, requests, and related data-movement line items, making it a transparent substitute for colder tiers. SP019
CP024 Azure Managed Lustre is a pay-as-you-go managed filesystem for HPC and AI workloads with per-GiB and per-hour pricing constructs. SP020
CP025 Azure’s pricing page explicitly notes Azure Government procurement eligibility, showing that trust and regulated purchasing rails are part of the substitute story. SP020
CP026 Google Filestore bills on provisioned capacity and, for custom performance, on provisioned IOPS, so cloud simplicity can still come with idle-capacity cost. SP021
CP027 Ceph documentation shows that internal-build paths can combine object, block, and file storage with POSIX semantics and S3- or Swift-compatible APIs in one system. SP022
CP028 MinIO AIStor markets exabyte-scale AI storage with hyperscaler economics and no vendor lock-in, directly challenging proprietary AI storage vendors on TCO messaging. SP023
CP029 Coldago’s 2025 map places VAST, Pure, NetApp, IBM, DDN, Qumulo, Hammerspace, and WEKA inside the same enterprise, high-performance, or cloud file-storage competitive frame. SP024
CP030 StorageNewsletter says Coldago ranks companies on portfolio breadth, installed base, recent performance, and product availability rather than on a single benchmark result. SP025
CP031 theCUBE Research argues that VAST’s AI OS ambition is ahead of its current maturity because it is not yet equivalent to a hyperscaler data platform, Databricks lakehouse, or Snowflake-grade database. SP026
CP032 theCUBE says VAST is trying to extend from storage into distributed scheduling, event streaming, vector search, database indexing, and agent execution. SP026, SP028
CP033 StorageMath argues that VAST’s Amplify and Flash Reclaim claims overstate effective capacity and turn data reduction into a lock-in narrative. SP027
CP034 Most specialist vendor pages in the retained set route buyers toward sales-led or calculator-led procurement rather than durable public list prices. SP002, SP003, SP005, SP006, SP007, SP010, SP011, SP012, SP013, SP014, SP016, SP017
CP035 Hyperscaler substitutes have a procurement advantage because pricing, metering, and billing are already integrated into existing cloud contracts. SP018, SP019, SP020, SP021
CP036 Pure and NetApp both counter specialist vendors with as-a-service packaging, which narrows any specialist advantage that depends only on moving capex to opex. SP003, SP006
CP037 Hammerspace and Qumulo foreground global namespace, data movement, or multi-cloud placement more explicitly than VAST’s retained public platform pages do. SP014, SP016, SP017, SP001
CP038 DDN, IBM, Pure, Qumulo, WEKA, and VAST all publicly target AI, HPC, or large unstructured-data workloads, so VAST does not enjoy a narrow uncontested wedge. SP001, SP002, SP007, SP010, SP012, SP016
CP039 Internal-build substitutes reduce vendor lock-in but move integration, lifecycle management, and operational burden back onto the buyer’s platform team. SP022, SP023
CP040 Qumulo’s managed Azure route plus self-hosted AWS and Google Cloud variants show that some challengers can land through hyperscaler channels rather than only direct appliance sales. SP017
CP041 WEKA’s funding, ARR estimates, and 2026 partner-heavy newsroom indicate a better-capitalized and more ecosystem-oriented rival than an early-stage niche point product. SP008, SP009
CP042 Hammerspace’s plan to use fresh capital for sales and marketing implies differentiated architecture but smaller current distribution reach than public incumbents and hyperscalers. SP015
CP043 DDN’s sovereign-AI narrative and Azure Government procurement language show that trust and regulatory posture are a competitive dimension, not just a feature checklist. SP010, SP020
CP044 VAST’s homepage says the company tripled valuation to $30 billion in a new Series F financing round, which supports its scale but also raises the bar for moat durability. SP028
CP045 Multi-homing remains practical because buyers can combine a primary AI storage layer with Qumulo or Hammerspace orchestration, cloud filesystems, object stores, and open-source tiers. SP014, SP017, SP018, SP019, SP022, SP023
CP046 VAST’s switching costs appear to come more from data gravity, tuned pipelines, and operational familiarity than from uniquely transparent pricing or a uniquely dominant compliance surface. SP001, SP018, SP020, SP022, SP023
CI001 VAST positions the AI Operating System as a unified platform combining storage, database, and compute rather than a single-purpose storage product. SI002, SI004
CI002 Across the reviewed public VAST commercial surfaces, buyers are routed to contact sales or demos rather than a self-serve list price. SI001, SI002, SI004
CI003 VAST describes Gemini as a software consumption model that disaggregates hardware from software economics. SI001, SI011
CI004 Cisco made the VAST AI Operating System available through its Global Price List and supports the joint stack as part of Cisco procurement. SI025
CI005 Polaris is included at no extra cost, implying that control-plane functionality is bundled into the AI OS rather than separately monetized. SI007
CI006 VAST signed a commercial agreement valued at $1.17 billion with CoreWeave, making VAST the primary data foundation for CoreWeave's AI cloud. SI014, SI030
CI007 VAST said in 2021 that customers made initial investments averaging about $1 million, that expansion averaged 328 percent, and that several customers had already invested more than $10 million. SI011
CI008 In FY22 VAST said it more than doubled its customer roster, managed more than three exabytes, averaged roughly 12PB per customer, and posted net revenue retention above 300 percent. SI010
CI009 VAST's 2023 Series E release said the company had surpassed $1 billion in cumulative software bookings, grown 3.3x year over year, maintained positive cash flow for 12 quarters, and reached nearly 90 percent gross margin. SI009
CI010 VAST's 2024 Forbes Cloud 100 release said the company ended 2023 at $200 million of ARR, remained cash-flow positive for three years, and posted a Rule of X of 162 percent. SI012
CI011 VAST's 2025 Forbes Cloud 100 release said the company had remained cash-flow positive for more than four years and was experiencing 5x year-over-year sales growth in FY25 with a Rule of X of 551 percent. SI013
CI012 VAST's 2026 Series F announcement said the company had surpassed $4 billion in cumulative bookings and exited the prior fiscal year with more than $500 million in committed annual recurring revenue, plus positive operating margin and free cash flow. SI006, SI015
CI013 VAST says thousands of organizations rely on its platform and that it supports environments powering millions of GPUs globally. SI006, SI002
CI014 Independent 2025 reporting said VAST reached $200 million of ARR by January 2025 and projected around $600 million of ARR in the following year. SI021, SI026
CI015 Independent 2026 reporting estimated that VAST's ARR-like revenue could be roughly $2 billion by the end of 2025, far above the company's disclosed committed-ARR floor. SI022, SI024
CI016 Public recurring-revenue signals are not directly comparable because they mix committed ARR, projected ARR, and broader ARR-like or non-committed revenue constructs. SI006, SI021, SI022, SI024
CI017 TNW and Sacra reported that top-100 new customers spent more than $1.2 million on average and that contracts typically ran five to seven years, while Sacra added that three customers represented more than $100 million of total commitments. SI024, SI026
CI018 VAST's go-to-market combines direct enterprise selling with partner and OEM distribution, as shown by contact-sales surfaces, Cisco GPL availability, and large cloud-partner agreements. SI001, SI014, SI025
CI019 VAST's white paper says OEMs such as SuperMicro, Lenovo, and Cisco wanted to sell VAST clusters on standard servers, reinforcing the company's partner-hardware model. SI005
CI020 Because Gemini separates software from hardware and OEMs can sell VAST on standard servers, VAST likely carries less inventory and manufacturing working capital than a vertically integrated appliance vendor. SI005, SI011, SI025
CI021 VAST repeatedly frames its economics as software-gross-margin oriented, and the last explicit public gross-margin datapoint was nearly 90 percent in 2023. SI009, SI010
CI022 VAST Amplify shows the company can monetize installed-base flash capacity and capacity optimization even when customers delay new hardware procurement. SI008
CI023 Bundled services such as Polaris imply that at least some service-delivery and control-plane costs sit inside platform margin rather than in separately disclosed SKUs. SI002, SI007
CI024 Cisco GPL availability and Cisco support materially lower procurement friction and improve the odds of channel-assisted expansion into enterprise AI budgets. SI025
CI025 Fierce reported that Google Cloud, CoreWeave, AWS, Azure, Crusoe, Lambda, and Core42 appear in VAST's cloud-partner ecosystem, suggesting broad partner-led distribution into AI clouds. SI026
CI026 VAST's 2026 financing valued the company at $30 billion, more than tripling the $9.1 billion Series E valuation from late 2023. SI006, SI015
CI027 The Series F was a mixed primary-and-secondary transaction, and outside coverage suggested more than $500 million and possibly most of the round went to secondary liquidity for employees and early investors. SI006, SI019, SI022, SI024
CI028 VAST said primary proceeds from the Series F would be used for global growth and strategic transactions that expand its technology footprint and partnerships. SI006
CI029 SiliconANGLE reported that Hallak said VAST has been cash-flow positive for several years and does not require external capital to operate. SI023
CI030 Because a meaningful portion of the headline Series F appears secondary, the gross $1 billion transaction value likely overstates net new balance-sheet cash available to fund operations. SI006, SI019, SI024
CI031 No reviewed source publicly disclosed VAST's cash on hand, monthly burn, cash runway, or a board-defined next-round trigger as of runDate. SI006, SI015, SI023, SI026
CI032 No reviewed source publicly disclosed debt facilities, project-finance obligations, or covenant packages for VAST. SI006, SI015, SI027, SI028, SI029
CI033 No reviewed source disclosed CAC, CAC payback, or a quantified sales cycle for VAST's enterprise and partner channels. SI001, SI006, SI015, SI026
CI034 No reviewed source disclosed realized ASPs, partner take-rates, discount schedules, or separate support and services gross margins. SI001, SI002, SI025
CI035 Fidelity's N-PORT filings marked the same basket of VAST preferred shares in one portfolio at roughly $1.08 million in August 2024, $1.34 million in February 2025, and $3.51 million in February 2026. SI027, SI028, SI029
CI036 Public traction supports strong revenue quality because the company reports positive free cash flow, positive operating margin, long-duration contracts, high average land sizes, and multi-year customer expansion. SI006, SI013, SI024, SI026
CI037 A full financial underwrite is blocked by unreconciled recurring-revenue definitions, no public customer-concentration schedule, and missing treasury and debt detail. SI006, SI021, SI022, SI023
CI038 Public customer references indicate meaningful exposure to AI-cloud and hyperscaler-style accounts such as CoreWeave and xAI, which can accelerate growth but create concentration and renewal-timing risk. SI014, SI024, SI026
CI039 VAST increasingly resembles an infrastructure-software company sold alongside partner hardware, but the public record does not reconcile what share of revenue is software, services, or system-linked economics. SI003, SI005, SI011, SI025
CI040 The financial verdict is positive on revenue quality and margin potential but incomplete for underwriting because cash, burn, CAC, pricing realization, and concentration remain private. SI006, SI023, SI024, SI026
CI041 The gap between bookings, committed ARR, and higher ARR-like estimates suggests that VAST's public metrics mix delivery-timed system economics with recurring software constructs that are not yet bridged publicly. SI006, SI011, SI024
CE001 VAST publicly defines the AI Operating System as software that unifies storage, database, and compute. SE001, SE017
CE002 VAST's 2026 product narrative frames the platform as a single layer for ingestion, retrieval, analytics, and inference rather than as a storage-only product. SE004, SE023
CE003 Public technical and partner materials present the platform as a set of named modules that includes DataStore, DataBase, DataSpace, and DataEngine. SE006, SE017
CE004 VAST says DataStore can present the same underlying data over NFS, SMB, and S3 simultaneously. SE006
CE005 VAST describes DataBase as an ACID-compliant high-speed data lake for transactional and analytical workloads at exabyte scale. SE006
CE006 VAST says DataBase supports query engines such as Trino and Spark with pushdown capabilities. SE006, SE010
CE007 VAST describes DataEngine as a distributed processing environment for event-driven AI workflows that initially writes streaming events directly into the VAST DataBase. SE006
CE008 VAST says DataSpace creates a global namespace across data centers, cloud, and edge using global lease management and intelligent data movement. SE006
CE009 VAST's public Kubernetes guidance supports both direct NFS-backed usage and CSI-based dynamic provisioning for persistent volumes. SE009, SE011, SE015
CE010 VAST distributes the CSI integration through an official GitHub repository and Helm chart instructions. SE011, SE015
CE011 Public operator tooling includes a RESTful VMS API, Swagger exposure, Python SDK, Terraform provider, and DataEngine CLI. SE019, SE012, SE013, SE014
CE012 The public VAST GitHub organization page showed 55 repositories and several repositories updated during May 2026 when reviewed for this run. SE010
CE013 The public vast-csi repository showed 844 commits when reviewed for this run. SE011
CE014 DASE is documented as separating computational resources from persistent data and system state. SE006, SE019
CE015 VAST says any CNode can access all data, metadata, and system state directly while persistent state resides on NVMe-backed storage enclosures. SE006
CE016 VAST's architecture is positioned as allowing compute and capacity to scale independently rather than being locked together per node. SE006, SE018, SE019
CE017 The retained public architecture shows workload-optimized deployments with CBoxes and DBoxes and a BlueField-optimized mode with CNodes on NVIDIA DPUs. SE006, SE008
CE018 Cisco and Supermicro both publicly package VAST AI OS and DASE-based services into partner infrastructure offers. SE017, SE018
CE019 Supermicro says its EBox runs CNode and DNode containers on the same server while preserving DASE-style independent scaling across the NVMe fabric. SE018
CE020 Juniper's validated design exposes concrete VMS GUI and CLI management, REST API usage, VIP rebalancing, and a specific AI-lab cluster topology. SE019
CE021 The retained corpus supports a multiprotocol platform spanning NFS, SMB, S3, SQL or table access, and native block via NVMe/TCP by 2025. SE006, SE021, SE022
CE022 The Next Platform reported that VAST expected iSCSI support later, implying the block story was newer and not yet protocol-complete versus legacy SAN expectations. SE022
CE023 The Next Platform reported that VAST block volumes inherit snapshots, clones, replication, and QoS from the common platform. SE022
CE024 DBTA and The Next Platform reported that Event Broker adds Kafka-compatible streaming so streamed data can land on the shared platform and become queryable through SQL. SE021, SE022
CE025 VAST's 2026 launch materials say CNode-X embeds NVIDIA technologies including cuDF through Sirius, cuVS, NIM microservices, and CMX support into core VAST services. SE004, SE023, SE024
CE026 VAST's 2026 materials cite early Sirius benchmarks of up to 44 percent lower query time and up to 80 percent lower query cost. SE004, SE023, SE024
CE027 Polaris is described as a Kubernetes-based global control plane for public cloud, neocloud, and on-prem VAST deployments. SE023, SE025
CE028 PolicyEngine is described as an inline zero-trust enforcement layer with tamper-proof audit logs for agentic workflows. SE023, SE025
CE029 TuningEngine is described as a closed-loop model-improvement service and both TuningEngine and PolicyEngine were presented as rolling out through 2026 rather than as long-proven defaults. SE023, SE025
CE030 VAST says CNode-X will come to market through OEM partners including Cisco and Supermicro. SE004, SE024
CE031 VAST says it is open-sourcing DataEngine blueprints for video search, enterprise document RAG, and genomics workflows. SE004, SE024
CE032 VAST's AI reference architecture and security guide describe zero-trust controls including RBAC, ABAC, encryption, auditing, network isolation, tenant isolation, and tenant-specific key management. SE006, SE007
CE033 The Security Configuration Guide says the platform is a STIG-hardened appliance and is listed on the DoDIN Approved Products List. SE007
CE034 The Security Configuration Guide documents HTTPS-based VMS management, CLI management, and predefined roles including Administrators, Read Only, Configuration, CSI, and Debug Metrics. SE007
CE035 The Security Configuration Guide says VAST supports Active Directory, LDAP, NIS, and local users and requires trusted TLS before configuring SSO and MFA. SE007
CE036 Commvault documents VAST as a validated NFS disk library or S3 cloud library target and says WORM mode can create immutable backup copies. SE020
CE037 CrowdStrike and VAST say their 2026 partnership extends detection and response across data ingestion, model training, runtime operation, and inference, but the release also warns that described features may not yet be generally available. SE026
CE038 VAST's compliance page markets the platform around governance, security, and regulated-data contexts including SEC, GDPR, and FINRA. SE005
CE039 This chapter's retained public corpus does not show a broad public trust-center style catalog of current audit reports or certifications for VAST Data beyond compliance marketing and deployment documentation. SE005, SE007
CE040 Cisco's public data sheet says VAST software licensing is subscription-based, portable, and sold across components that include DataStore, DataBase, DataEngine, and DataSpace. SE017
CE041 VASTPY publicly supports API-token authentication on VAST 5.3 and later and exposes protocol metrics for NFS, SMB, and S3. SE013
CE042 The Terraform provider publicly supports registry installation and importing existing VAST resources. SE012
CE043 The retained corpus shows a visible public developer surface across CSI, SDKs, Terraform, CLI, and Go tooling, but it does not show usage telemetry strong enough to quantify practitioner adoption. SE010, SE011, SE012, SE013, SE014, SE016
CE044 Public 2026 materials describe Polaris, PolicyEngine, and TuningEngine in detail, but the retained product-tech corpus does not show broad public customer reference deployments or mature admin documentation for them yet. SE023, SE025, SE026
CE045 VAST's most supportable differentiation is broad shared-data-plane scope, while its most visible technical risk is making the AI Operating System claim feel as mature as the underlying storage core. SE006, SE022, SE023
CE046 The Security Configuration Guide says the product should be deployed inside a hardened customer environment, implying secure operation depends on customer-side hardening as well as product defaults. SE007
CE047 This chapter's retained public corpus does not provide a clear current public SLA, uptime dashboard, or incident-history source for VAST Data reliability. SE001, SE002, SE005, SE007, SE017
CU001 Official and customer-proof pages show VAST adoption across AI cloud, financial services, healthcare and life sciences, government research, media, and telecom segments. SU001, SU007, SU010, SU013, SU017
CU002 Retained customer-proof pages name NHL, SK Telecom, CoreWeave, HSBC, CACEIS, Jump Trading, HHS, Invitae, PacBio, LLNL, and Chan Zuckerberg Initiative. SU002, SU007, SU010, SU013, SU014, SU015, SU016, SU017, SU018, SU019, SU020
CU003 The official 2023 Series E release also lists Booking Holdings, U.S. Air Force, U.S. Department of Energy, Verizon, Boston Children’s Hospital, Pixar, and Zoom as customer-roster examples. SU022
CU004 VAST’s buyer-user-payer pattern is segment specific, with central infrastructure budgets sponsoring deployments for downstream researchers, creators, analysts, or cloud tenants. SU007, SU010, SU013, SU015, SU017
CU005 The NHL says it originally selected VAST as a production platform for media workflows rather than as a generic storage refresh. SU007
CU006 The NHL-VAST relationship has lasted roughly six years and expanded from archive management into live media operations. SU007, SU008
CU007 The NHL says it moved more than 20 PB of archive footage onto the VAST platform. SU007
CU008 By 2025 the NHL had implemented VAST in all 32 arenas plus headquarters to create a single logical cluster or namespace. SU007, SU008
CU009 In a 2026 outdoor-game pilot, the NHL’s portable VAST node replicated more than 2 TB of footage with stable operation and minimal turnaround time. SU009
CU010 The NHL says the next phase is to operationalize its distributed VAST model across major events and league-wide arenas. SU009
CU011 SK Telecom deployed VAST AI OS as a core component of its sovereign AI cloud in 2025. SU010, SU012
CU012 SK Telecom says VAST-backed GPU environments can be provisioned in roughly 5–10 minutes. SU010, SU011, SU012
CU013 SK Telecom positions the deployment as infrastructure for government, research, and enterprise AI customers inside South Korea. SU010, SU012
CU014 SK Telecom’s VAST-enabled sovereign AI stack uses 1,000 NVIDIA Blackwell GPUs and claims throughput approaching 14 GB/s via RDMA passthrough. SU011
CU015 VAST announced a $1.17 billion commercial agreement with CoreWeave in November 2025. SU003, SU004, SU005, SU006
CU016 Official and independent coverage describe CoreWeave as a long-standing or expanded partner using VAST as a primary data foundation for its AI cloud. SU002, SU003, SU004, SU006
CU017 VAST’s CoreWeave customer story claims 23 global data centers, more than 250,000 GPUs, more than 500 PB of capacity, up to 2.3 GB/s per GPU, and 99.9999% uptime. SU002
CU018 CoreWeave evidence is partner mediated because VAST’s growth is tied not only to CoreWeave as an account but also to workloads CoreWeave delivers to shared end customers. SU003, SU004, SU005
CU019 SDxCentral says VAST’s platform under CoreWeave supports AI training and inference for CoreWeave customers such as Meta, OpenAI, and Microsoft. SU005
CU020 In 2021 VAST said average initial customer investment was about $1 million. SU021
CU021 In 2021 VAST said net revenue within existing customers had expanded 328% on average. SU021
CU022 In 2021 VAST said several customers had already invested more than $10 million. SU021
CU023 Customer-proof freshness is uneven because NHL, SKT, and CoreWeave have 2025-2026 proof while many other named logos appear only in 2023 roster disclosures or title-level case pages. SU007, SU009, SU010, SU012, SU022
CU024 HSBC, CACEIS, and Jump Trading extend proof into financial services, but the retained public pages provide less quantified outcome detail than the AI-cloud and media references. SU016, SU017, SU018
CU025 Invitae says VAST improved data access 6x and IOPS 30x for genetics workflows. SU013
CU026 PacBio says it added another 2 PB to its VAST cluster with no issues and uses VAST to scale genomics workflows. SU014
CU027 HHS testimony on FeaturedCustomers says the department deployed scalable flash-based architectures with VAST to modernize science workflows and improve public health research. SU015, SU025
CU028 LLNL’s case page presents VAST as the platform underpinning HPC and machine-learning work for drug-efficacy research. SU019
CU029 FeaturedCustomers lists 65 testimonials, 37 case studies, 36 customer videos, and a 4.8/5 score based on 1,417 reference ratings for VAST. SU025
CU030 PeerSpot shows VAST at 5.0/5 from 2 reviews with 100% willing to recommend as of May 2026. SU026
CU031 PeerSpot also cites write-performance and read-write-ratio weaknesses and says VAST is not the cheapest option. SU026
CU032 Public durability proof is indirect because the retained set shows historical expansion, long-lived named relationships, and review sentiment rather than current cohort metrics. SU007, SU021, SU025, SU026
CU033 The 2023 official release says VAST customer partnerships with CoreWeave, Lambda, and Core42 are central to next-generation AI-cloud infrastructure. SU022
CU034 SK Telecom evidence shows VAST adoption routed through NVIDIA and Supermicro ecosystem partners rather than only through direct enterprise procurement. SU010, SU012
CU035 The retained NVIDIA Nasdaq customer story documents Nasdaq AI outcomes, but it does not mention VAST, so it is adjacent ecosystem evidence rather than retained VAST customer proof. SU027
CU036 The retained NVIDIA BMW case study documents BMW AI-production outcomes, but it does not mention VAST, so BMW remains unverified as a VAST customer in this chapter. SU028
CU037 Sacra identifies Goldman Sachs and Dell Technologies Capital as investors in VAST rather than as customer references. SU024
CU038 The retained public proof set skews toward large enterprises, public-sector research bodies, AI clouds, and infrastructure-heavy buyers rather than SMB or self-serve usage. SU001, SU022, SU026
CU039 No retained source disclosed current total customer count, channel mix, sales-cycle length, partner take-rates, or a public list-price schedule. SU001, SU022, SU024, SU026
CU040 No retained source disclosed current GRR, NRR, churn, or renewal rates by cohort as of the run date. SU021, SU022, SU023, SU024
CU041 Official proof shows use cases spanning media production, sovereign AI cloud, AI-cloud infrastructure, genomics, quantitative trading, and financial-services AI. SU007, SU010, SU013, SU014, SU016, SU017, SU018
CU042 The strongest production-grade proof in the retained set comes from NHL, SKT, CoreWeave, Invitae, and PacBio because those references include deployment detail or measured outcomes rather than just logos. SU002, SU007, SU009, SU010, SU013, SU014
CU043 VAST’s customer journey often includes evaluation or codesign, production deployment, multisite rollout, and later expansion into adjacent workflows. SU007, SU009, SU011, SU021
CU044 CoreWeave is simultaneously VAST’s strongest public expansion proof and clearest public concentration risk because one disclosed partner-linked agreement is worth $1.17 billion. SU003, SU004, SU005, SU006
CU045 PeerSpot’s review traffic shows interest concentrated in large enterprises and led by financial-services researchers, reinforcing that VAST is sold into heavyweight environments. SU026
CU046 The official 2023 customer roster is qualitative and broad, but it does not disclose a denominator for total customers or the revenue mix behind those logos. SU022
CU047 The main unresolved underwriting asks are current customer count, cohort retention, top-customer concentration, and partner-sourced revenue mix. SU022, SU023, SU024, SU026
CR001 VAST closed its Series F at a $30 billion valuation, more than tripling the $9.1 billion valuation disclosed for the 2023 Series E. SR001, SR002
CR002 The Series F included both primary and secondary capital and carried total transaction value of about $1 billion. SR001, SR002, SR003
CR003 VAST said primary proceeds from the Series F will fund global growth, technology expansion, and strategic transactions or partnerships. SR001
CR004 TechCrunch reported in June 2025 that VAST was already targeting a roughly $25 billion valuation before the 2026 round closed. SR003
CR005 VAST's strongest public operating proof still comes from company-claimed metrics such as more than $4 billion of cumulative bookings, more than $500 million of committed ARR, and positive free cash flow. SR001
CR006 At a $30 billion valuation, the public underwriting case still leans heavily on company-claimed scale metrics rather than audited public disclosure. SR001, SR002, SR003
CR007 VAST disclosed a $1.17 billion commercial partnership with CoreWeave. SR004, SR005, SR006
CR008 The CoreWeave agreement makes VAST's AI OS the primary data platform underneath CoreWeave's compute infrastructure. SR004, SR006
CR009 SDxCentral reported that the CoreWeave environment VAST supports feeds workloads for Meta, OpenAI, and Microsoft. SR006
CR010 Globes reported that VAST's customer base includes xAI, CoreWeave, AWS, and other neocloud or enterprise accounts. SR018
CR011 VAST's AI Operating System became available through Cisco's Global Price List with Cisco support for the joint stack. SR009, SR010
CR012 Cisco, NVIDIA, and VAST published a validated enterprise AI data platform architecture inside Cisco Secure AI Factory with NVIDIA. SR010
CR013 Supermicro and VAST launched the CNode-X integrated AI data platform with NVIDIA in February 2026. SR008
CR014 VAST says its AI OS runs directly on NVIDIA-powered servers and uses NVIDIA libraries to accelerate compute and data services. SR007, SR022
CR015 VAST obtained NVIDIA-Certified Storage status in 2025, tightening its product story to NVIDIA's ecosystem and validation path. SR022
CR016 VAST's partner ecosystem page says deployments rely on a global network of resellers, integrators, service providers, technology leaders, and cloud providers. SR026
CR017 CRN reported that VAST's storage fabric is delivered on top of OEM hardware from Lenovo, Dell, HPE, SuperMicro, and Cisco. SR021
CR018 Dell publicly identified VAST as a direct rival in enterprise AI storage and claimed better power and rack-efficiency economics than VAST. SR021
CR019 VAST responded to Dell's comparison by calling it misrepresentation and marketing math, showing that benchmark credibility is now a live competitive battleground. SR021
CR020 StorageMath published a direct adverse critique arguing that VAST Amplify's six-times-capacity claim relied on fake math and lock-in framing. SR025
CR021 NetApp sued former CTO Jonsi Stefansson alleging he built Red Stapler while still employed at NetApp and that VAST later acquired the resulting company. SR011
CR022 A Florida court dismissed NetApp's complaint without prejudice on forum grounds and NetApp appealed rather than accepting a merits loss. SR012
CR023 The retained litigation reporting says VAST is not named as a defendant in the NetApp complaint. SR011, SR012
CR024 VAST's public privacy policy says its website processing includes IP addresses, device identifiers, contact data, job-candidate submissions, cookies, and named third-party tools, with storage in the United States and cloud infrastructure. SR015
CR025 Wilson Sonsini says the DOJ bulk-data rules effective April 8, 2025 can restrict vendor, investor, and employee access to covered datasets and can carry material civil and criminal penalties. SR013
CR026 White & Case says the 2025-2026 privacy landscape layers the bulk-data rule, new state laws, and more aggressive enforcement onto companies handling sensitive or cross-border data. SR014
CR027 VAST's September 2025 SDS announcement was issued from Tel Aviv and positioned VAST inside a sovereign AI cloud buildout for Israel. SR017
CR028 The SDS sovereign-cloud build relies on thousands of NVIDIA Blackwell GPUs and dozens of petabytes of VAST infrastructure. SR017
CR029 Globes reported in 2025 that VAST had about 1,000 employees, roughly half in Israel, with its main development center in Tel Aviv. SR018
CR030 The Israel Innovation Authority says Israeli high-tech output has been stagnant for two years, R&D roles fell 6.5% year over year, venture fundraising fell sharply, and foreign capital dominates deep-tech. SR019
CR031 JD Supra says Israeli export-control scrutiny, dual-use rules, and IIA grants can constrain IP transfer, manufacturing, and exit structuring. SR020
CR032 Goldman Sachs frames the AI build-out as assumption-sensitive rather than fixed, keeping open the possibility that spending enthusiasm outruns durable demand. SR023
CR033 Gartner says AI is in a 2026 trough of disillusionment and will often be sold by incumbent providers rather than attached to moonshot projects. SR024
CR034 Gartner still forecasts AI infrastructure spending of about $1.366 trillion in 2026, up roughly $401 billion year over year. SR024
CR035 Supermicro says CNode-X follows the NVIDIA AI Data Platform reference architecture and wraps validated hardware, networking, and management around VAST software. SR008
CR036 On residual severity, the most material underwriting risks are valuation reset, partner and customer concentration, execution breadth, legal or IP overhang, and Israel-linked continuity. SR001, SR004, SR011, SR019, SR021, SR023, SR024
CR037 VAST's public announcements still center founder and CEO Renen Hallak across financing, Cisco, and Israel sovereign-cloud launches, indicating meaningful leadership concentration. SR001, SR009, SR017
CR038 The Red Stapler acquisition was part of VAST's push into hyperscale cloud and control-plane capability, so any injunction or code-provenance issue would hit an active expansion vector rather than a legacy side project. SR011, SR012
CR039 Globes described VAST president Mike Wing as a former Dell executive, showing added commercial depth beyond the founders without removing founder dependence. SR018
CR040 VAST's public contract surface routes services through a public EULA plus separate authorized-distributor or written contracts rather than through website terms alone. SR016
CR041 Cisco's solution overview places VAST inside use cases spanning financial services, healthcare, retail, manufacturing, and enterprise IT, pushing the product into regulated or mission-critical workloads. SR010
CR042 The retained public source set does not disclose VAST's top-customer concentration, renewal timing, burn or runway, debt, or current security certifications. SR001, SR010, SR018
CR043 Gartner's incumbent-led 2026 AI adoption view increases go-to-market risk for challengers even if total infrastructure spend keeps rising. SR021, SR024
CR044 CRN says Dell remains a storage market leader while VAST is gaining traction mainly in AI and unstructured-data segments, underscoring competitive pressure from better-capitalized incumbents. SR021
CR045 VAST's current delivery stack is multi-layer dependent: NVIDIA for accelerated architecture, Cisco for enterprise channel, Supermicro and other OEMs for validated hardware, and CoreWeave or SDS for flagship cloud deployments. SR004, SR007, SR008, SR009, SR010, SR017, SR021, SR026
CR046 Because the Series F included major secondary liquidity, the headline $1 billion round overstates fresh balance-sheet cash available for execution. SR001, SR002, SR003
CR047 Because VAST's flagship disclosed wins are GPU-cluster and neocloud centric, demand remains exposed to AI capex timing and model-lab spending concentration rather than to a fully diversified enterprise base. SR004, SR006, SR018, SR023, SR024
CR048 Mitigation maturity is strongest where VAST has financing access and validated partner architectures, and weakest where investors still need private disclosure or external legal and compliance clearance. SR001, SR008, SR009, SR010, SR011, SR014, SR019
CR049 The cleanest thesis-break triggers are a down-round or failed IPO pricing event, VAST being pulled directly into Red Stapler litigation or an injunction, inability to evidence security certifications for regulated workloads, or diligence showing concentration materially worse than the public record suggests. SR001, SR011, SR012, SR014, SR018
CR050 The ranked risk stack is mitigated enough for continued diligence, but not enough for blind underwriting without concentration, legal provenance, and compliance evidence. SR001, SR006, SR011, SR014, SR019, SR021
CR051 Supermicro's solution page says the EBox platform is purpose-built for VAST's DASE architecture and optimized for AI and GPU cloud markets. SR027
CR052 VAST's public Terms of Use say website terms do not govern storage software or services, which instead fall under the EULA or separate written contracts. SR028, SR016
CR053 Israel's April 2025 national AI report frames AI as a national priority, which can support local infrastructure demand while increasing policy involvement in the ecosystem. SR029
CR054 Deloitte and F2 say the foundation layer of Israel's AI stack captured only about 13 percent of AI funding and that hardware, infrastructure, and cloud are only moderate-potential hubs, reinforcing the capital-intensity and ecosystem limits around infrastructure vendors. SR030
CR055 Partner breadth across Cisco, Supermicro, OEMs, and the wider ecosystem softens single-vendor dependence but does not remove system-level reliance on external channels and hardware partners. SR009, SR021, SR026, SR027
CV001 VAST Data said on 2026-04-22 that it closed a Series F financing at a $30 billion valuation with approximately $1 billion of total transaction value. SV001, SV003
CV002 The Series F valuation represented more than a threefold increase from VAST's $9.1 billion Series E valuation in late 2023. SV001, SV011, SV003
CV003 Drive Capital led the Series F with Access Industries as co-lead, and Fidelity, NEA, and NVIDIA also participated. SV001, SV003
CV004 March 2026 trade-press reports described the $30 billion raise before the company officially disclosed the closed round on 2026-04-22, so the public record contains both an early-reporting window and a later close date. SV004, SV006, SV001
CV005 The company said the financing mixed primary and secondary capital rather than being a purely primary growth round. SV001, SV003
CV006 StorageNewsletter reported that outside databases described the transaction as roughly half new financing and half secondary liquidity, but VAST did not publicly confirm an exact split. SV006, SV008
CV007 VAST said it exited the previous fiscal year with more than $500 million of committed annual recurring revenue. SV001, SV005
CV008 VAST said it has surpassed $4 billion in cumulative bookings. SV001, SV005
CV009 VAST said it exited that same fiscal year with positive operating margin and positive free cash flow. SV001, SV005
CV010 VAST said its most recent fiscal year produced a Rule of X score of 228 percent. SV001, SV005
CV011 VAST's CEO said the company often sells three-year and five-year licenses, which makes annual recurring revenue smaller than total multi-year contract value. SV016, SV010
CV012 Reuters reported in June 2025 that VAST had reached roughly $200 million of ARR by January 2025 and was projected to reach about $600 million the following year. SV012
CV013 CTech reported that VAST's ARR rose from about $350 million in early 2024 to above $500 million by July 2024, around $700 million by March 2025, and roughly $2 billion by the end of 2025. SV009
CV014 The public revenue record therefore spans at least three non-equivalent constructs: more than $500 million of committed ARR, about $600 million of sourced ARR projection, and roughly $2 billion of third-party ARR-like estimate. SV001, SV012, SV009
CV015 A Fidelity Select Tech Hardware filing for 2024-08-31 marked one VAST Series A preferred line at about $18.16 per share. SV015
CV016 A Fidelity filing for 2025-02-28 marked the same VAST Series A preferred line at about $22.54 per share. SV014
CV017 A Fidelity filing for 2026-02-28 marked that same VAST Series A preferred line at about $59.22 per share, roughly 3.3 times the August 2024 mark. SV013, SV015
CV018 Fidelity's filing progression independently corroborates a steep private-market step-up in VAST's mark ahead of the Series F announcement. SV013, SV014, SV015
CV019 A $30 billion price tag implies roughly 60 times VAST's official $500 million CARR floor. SV001
CV020 A $30 billion valuation implies roughly 50 times the Reuters-cited $600 million ARR projection. SV012
CV021 A $30 billion valuation implies roughly 15 times the CTech estimate of about $2 billion of ARR. SV009
CV022 Everpure, the rebranded Pure Storage, reported fiscal 2026 revenue above $3.6 billion and total cash, cash equivalents, and marketable securities of about $1.5 billion. SV017, SV019
CV023 Stock Analysis showed Everpure trading at about $29.35 billion of market capitalization and $28.02 billion of enterprise value on 2026-05-27. SV036
CV024 Everpure therefore traded at roughly 7.7 times EV to trailing revenue at runDate. SV017, SV019, SV036
CV025 NetApp reported $1.71 billion of Q3 FY2026 revenue and guided fiscal 2026 revenue to $6.772-$6.922 billion. SV020, SV022
CV026 Analyst-market-data sources showed NetApp at about $27.42 billion of market capitalization and $27.14 billion of enterprise value with $6.70 billion of trailing revenue on 2026-05-27. SV022, SV035
CV027 NetApp therefore traded at roughly 4.1 times EV to trailing revenue at runDate. SV022, SV035
CV028 CoreWeave reported Q1 2026 revenue of $2.078 billion, revenue backlog of $99.4 billion, cash of about $3.32 billion, and debt of about $24.9 billion. SV023
CV029 Stock Analysis showed CoreWeave at about $57.77 billion of market capitalization and $90.65 billion of enterprise value on 2026-05-27. SV034, SV025
CV030 CoreWeave's enterprise value equated to about 10.9 times annualized Q1 2026 revenue, illustrating the high end of public AI infrastructure multiples but with materially heavier leverage than VAST discloses. SV023, SV034
CV031 Lambda officially announced a $480 million Series D round, and Reuters said sources valued that round at about $2.5 billion post-money. SV026, SV037
CV032 Crusoe announced a $1.375 billion Series E at a valuation above $10 billion to scale its AI infrastructure platform. SV027
CV033 Together AI said its $305 million Series B valued the company at $3.3 billion to expand AI acceleration cloud capacity. SV028, SV029
CV034 Relative to these private AI infrastructure rounds, VAST's $30 billion price is a far larger absolute bet and only looks comparable if its true recurring revenue base is much closer to $2 billion than to the disclosed $500-$600 million band. SV009, SV027, SV029, SV037
CV035 VAST merits some premium to legacy storage peers because the public record still shows AI Operating System positioning, marquee AI and enterprise customers, rapid growth, and profitability. SV001, SV005, SV016
CV036 Public evidence does not yet prove enough revenue scale or metric consistency to justify paying a larger premium than the market already assigns to faster-growing AI infrastructure names. SV005, SV009, SV030, SV034
CV037 Because the round included material secondary liquidity, the headline $1 billion does not equal $1 billion of fresh operating cash for VAST. SV001, SV006, SV008
CV038 CNBC's January 2026 survey captured widespread concern that record AI valuations and deal activity could be vulnerable to bubble-like multiple compression. SV030
CV039 VAST's CEO said the company has not started an IPO process and has not hired bankers yet. SV016
CV040 The most supportable exit path is eventual IPO rather than strategic sale, but timing remains uncommitted and therefore cannot be priced with precision. SV016, SV004
CV041 A bear case that assumes only about $0.8 billion of durable recurring revenue, public-storage-style 8x-10x multiples, and weaker AI-capex sentiment implies roughly $6.4-$8.0 billion of value. SV001, SV017, SV020, SV030
CV042 A base case that assumes about $1.3-$1.5 billion of normalized recurring revenue power and 10x-14x AI infrastructure multiples implies roughly $13-$21 billion of value. SV001, SV009, SV034, SV036
CV043 A bull case that assumes the roughly $2 billion ARR estimate is directionally right and that VAST earns 14x-18x premium multiples implies roughly $28-$36 billion of value. SV001, SV009, SV034, SV036
CV044 The public record supports a track or research-more stance for fresh capital at $30 billion, with medium confidence, high risk, and an expensive valuation stance. SV009, SV030, SV034, SV036
CV045 The recommendation would improve only if private diligence proves at least about $1.5-$2.0 billion of durable recurring revenue, limited customer concentration, and a cleaner preference stack than public evidence reveals. SV006, SV009, SV016
CV046 Immediate thesis-break triggers are failure to reconcile ARR definitions, evidence of outsized customer concentration, or an IPO-ready revenue profile that still cannot support public-comp multiples. SV009, SV016, SV030
CV047 The company-quality case survives better than the entry-price case: product, customer, and demand signals remain strong even though fresh-money underwriting does not clear at the round price. SV001, SV005, SV016
CV048 The reviewed public sources do not disclose exact Series F liquidation preferences, participation rights, option-pool effects, or fully diluted ownership. SV001, SV006, SV016
CV049 The highest-value diligence asks are an audited revenue bridge, top-customer ARR and renewals, the post-Series-F cap table, and concrete IPO-readiness milestones. SV006, SV016, SV023
CV050 Existing low-basis investors can rationally hold for IPO optionality, but new investors at $30 billion need either a price reset or materially stronger evidence before adding capital. SV016, SV030, SV036
CV051 A $30 billion entry requires roughly a $60 billion exit for 2x gross and a $90 billion exit for 3x gross before dilution. SV001
CV052 At Pure-like 7.7x EV/revenue, a $60 billion exit would require roughly $7.8 billion of revenue, and even at CoreWeave-like 10.9x it would still require roughly $5.5 billion. SV017, SV019, SV023, SV034, SV036
来源
编号出版方标题引文
SO001 VAST Data About VAST Data - Revolutionary Data Platform - VAST Data
SO003 VAST Data VAST Data Contact - VAST Data
SO004 VAST Data VAST Data Customer Success Stories | AI Operating System - VAST Data
SO005 VAST Data VAST DataStore: Universal Data Store for the AI Era - VAST Data
SO006 VAST Data The VAST Platform White Paper
SO007 VAST Data VAST Data Valued at $30 Billion as AI Drives a New Infrastructure Stack The financing included primary and secondary capital, bringing the total transaction value to approximately $1 billion.
SO008 VAST Data VAST Data Launches VAST Amplify to Expand Flash Capacity During Shortages
SO009 VAST Data VAST Data Introduces Polaris for Hybrid Multicloud AI Orchestration
SO010 VAST Data VAST Data Triples Valuation to $3.7 Billion in One Year
SO011 VAST Data VAST Closes Series E Funding Round, Nearly Triples Valuation to $9.1 Billion Global Expansion: Now with more than 700 employees worldwide VAST Data is actively broadening its business footprint.
SO012 VAST Data 2026: The Year of AI Inference
SO013 CNBC Nvidia backs AI company Vast Data at $30 billion valuation
SO014 Data Center Dynamics Vast Data secures $1bn in Series F funding at $30bn valuation
SO015 Blocks & Files VAST Data raises $1B at $30B valuation as AI storage demand surges
SO016 CRN Vast Data CEO Renen Hallak On AI, Growth And A $30 Billion Valuation
SO017 StorageReview VAST Data Unveils Agentic AI OS and Advances Its Thinking Machine Vision
SO018 Supermicro VAST Data Platform for Supermicro | Supermicro
SO019 Hedgeweek Tiger Global Management leads USD83m Series D VAST Data funding round as company valuation triples in one year to USD3.7bn - Hedgeweek
SO020 SDxCentral Former NetApp CTO wins dismissal of lawsuit after Vast Data deal NetApp has appealed the ruling.
SO021 CTech Vast Data executive accused by NetApp of stealing company secrets | CTech Vast itself is not named as a defendant.
SO022 TechCrunch Exclusive: AI storage platform Vast Data aimed for $25B valuation in new round, sources say
SO023 VentureBeat AI infrastructure major VAST Data’s valuation surges to over $9B after fresh funding
SO024 Crunchbase News Vast Data Closes $118M Series E, More Than Doubles Valuation
SO025 Enterprise Times VAST Data gets a vast valuation and raises more funds
SO026 Fierce Network Here’s why Nvidia-backed startup VAST Data is all the rage
SO028 Yahoo Finance / Reuters Exclusive-Alphabet's CapitalG, Nvidia in talks to fund Vast Data at up to $30 billion valuation, sources say
SO029 VAST Data VAST AI Operating System: Powering the Agentic AI Revolution - VAST Data
SM001 VAST Data VAST AI Operating System: Powering the Agentic AI Revolution
SM002 VAST Data VAST Data Platform Services: AI-Powered Discovery Engine
SM003 Gartner Gartner Says Worldwide AI Spending Will Total $2.5 Trillion in 2026
SM004 InfotechLead Storage Market Forecast for Stronger Growth in 2026 as AI Data Demand Accelerates - InfotechLead
SM005 Fortune Business Insights AI Powered Storage Market Size, Growth | Industry Report, 2034
SM006 Research and Markets AI Infrastructure Market Report 2026 - Research and Markets
SM007 Coherent Market Insights AI Infrastructure Market Size and YoY Growth Rate, 2026-2033
SM008 Avnet Riding the AI Supercycle: Navigating the 2026 Memory & Storage Market
SM009 Deloitte The State of AI in the Enterprise - 2026 AI report
SM010 Goldman Sachs Tracking Trillions: The Assumptions Shaping the Scale of the AI Build-Out
SM011 Electronic Design / SNIA Data Center Storage in 2026: When Storage, AI, and Compute Converge
SM012 Informatica AI Adoption Trends 2026: Trust, Data Quality & Governance Challenges | Informatica CDO Insights
SM013 MinIO AI Storage Architecture: Overcoming the Bottleneck Limiting AI Scale in 2026
SM014 VAST Data VAST Data Introduces End-to-End Fully Accelerated AI Data Stack with NVIDIA
SM015 VAST Data Solutions
SM016 VAST Data The VAST Platform White Paper
SM017 Supermicro Supermicro and VAST Data Launch a New Enterprise AI Data Platform Solution with NVIDIA to Accelerate AI Factory Deployment
SM018 VAST Data / NVIDIA VAST Data Storage Reference Architecture
SM019 InfotechLead IDC Identifies Seven Critical Buyer Shifts Shaping Enterprise IT Strategy in 2026 - InfotechLead
SM020 Forrester Forrester’s 2026 Buyer Insights: GenAI Is Upending B2B Buying As Leaders Face Mounting Pressure To Justify Every Dollar Spent
SM021 NetApp NetApp Reports First Quarter of Fiscal Year 2026 Results
SM022 Pure Storage Pure Storage Announces First Quarter Fiscal 2026 Financial Results
SM023 PR Newswire / Pure Storage Pure Storage Announces Third Quarter Fiscal 2026 Financial Results
SM024 Cloudera 2026 Data Architecture, Data Governance, and AI Trends & Predictions | Cloudera
SM025 INFUSE INFUSE Insights: Voice of the Buyer 2026 | B2B AI Trends | INFUSE
SP001 VAST Data VAST Data Platform Services: AI-Powered Discovery Engine Unlock 90% of your untapped data with a breakthrough data architecture built to harness limitless data.
SP002 Everpure Everpure FlashBlade//S | Everpure FlashBlade//S provides a single solution for file and object workloads that’s easy to set up, manage, scale, and upgrade.
SP003 Everpure Evergreen//One: Storage-as-a-Service | Everpure You only pay for what you use. Get elastic capacity, non-disruptive upgrades, and built-in capacity planning.
SP004 Everpure Investor Relations Pure Storage Announces Fiscal Fourth Quarter and Full Year 2025 Financial Results Full year 2025 revenue surpasses $3 billion, representing growth of 12% year-over-year.
SP005 NetApp Data Storage – On-Prem, in the Cloud, Hybrid Cloud Environments | NetApp Store any data, anyplace, with a truly unified environment.
SP006 NetApp NetApp Keystone, Storage as a Service (STaas) | NetApp Seize your advantage with Keystone’s pay-as-you-go, hybrid cloud storage services.
SP007 WEKA How NeuralMesh™ by WEKA Powers Modular AI Infrastructure NeuralMesh redefines what storage means in the age of AI – a high-performance, software-defined system that powers HPC and AI workloads at any scale.
SP008 WEKA Weka Newsroom WEKA Accelerates AI Factory Deployment Times From Months to Minutes with Turnkey NVIDIA AI Data Platform Solution.
SP009 Globes Weka raises $140m at $1.6b valuation Weka has announced the completion of a $140 million oversubscribed Series E financing round, at a company valuation of $1.6 billion.
SP010 DDN DDN: Data Intelligence Platform Built for AI DDN is the data intelligence platform behind AI factories, sovereign AI, hyperscalers, and the world’s most demanding GPU workloads.
SP011 DDN DDN Infinia: Faster, More Cost-effective Alternative to Cloud Storage DDN Infinia is purpose-built to overcome today’s AI infrastructure challenges by providing a unified, high-performance AI data intelligence platform.
SP012 IBM IBM Storage Scale IBM Storage Scale unifies unstructured data across data centers, cloud, and edge into a single global platform.
SP013 IBM IBM Storage Scale System Designed for extreme performance and scalability to support data-intensive workloads. As NVIDIA-Certified Storage, it delivers GPU-optimized throughput.
SP014 Hammerspace Hammerspace Tier 0 Competitive Brief Vast clusters in the cloud lack performance and scale.
SP015 TechCrunch Hammerspace, an unstructured data wrangler used by Meta, raises $100M at $500M+ valuation | TechCrunch The funding is being described as a “strategic venture round,” and it values Hammerspace at over $500 million.
SP016 Qumulo What is Qumulo? — One platform for all data, edge-core-cloud The Qumulo Data Platform combines the world’s most advanced file system with a global namespace.
SP017 Qumulo Exabyte-Scale File Storage in the Cloud | Qumulo Deploy Qumulo in minutes as a fully managed service in Azure directly from the Azure portal.
SP018 Amazon Web Services Amazon FSx for Lustre Pricing With Amazon FSx for Lustre, you pay only for the resources you use and there are no minimum fees or set-up charges.
SP019 Amazon Web Services S3 Pricing Charges are incurred as follows.
SP020 Microsoft Azure Azure Managed Lustre - Pricing | Microsoft Azure Azure Managed Lustre is a managed, pay-as-you-go file system for high-performance computing (HPC) and AI workloads.
SP021 Google Cloud Pricing | Filestore You are charged based on the provisioned capacity, not based on the capacity used.
SP022 Ceph Documentation Welcome to Ceph — Ceph Documentation Ceph delivers object, block, and file storage in one unified system.
SP023 MinIO MinIO AIStor: Exabyte-Scale Storage Engineered for the AI Era Up to 40% lower TCO than proprietary “AI storage” vendors—no lock-in, no hardware markups, no hidden tiering costs.
SP024 Coldago Research Map 2025 for File Storage | Coldago Research High Performance File Storage ... leaders: DDN, Dell, Hammerspace, Huawei, IBM, Pure Storage, Qumulo, VAST Data and WEKA.
SP025 StorageNewsletter Coldago Map 2025 for File Storage Coldago evaluates and ranks companies rather than individual products, taking into account all offerings relevant to the specific Map.
SP026 theCUBE Research Special Breaking Analysis | Unpacking VAST Data’s Ambition to Become the Operating System for the Thinking Machine VAST has not (yet) achieved a Databricks-style lakehouse, a Snowflake-grade cloud database, nor a hyperscaler data platform.
SP027 StorageMath VAST Amplify's '6x Capacity' Claim: Exploiting a Real Crisis with Fake Math The SSD crisis is real. VAST’s math is not.
SP028 VAST Data VAST AI Operating System: Powering the Agentic AI Revolution - VAST Data The VAST AI Operating System unifies storage, database, and compute to transform data into action.
SI001 VAST Data About VAST Data - Revolutionary Data Platform - VAST Data
SI002 VAST Data VAST AI Operating System: Powering the Agentic AI Revolution
SI003 VAST Data VAST Data Customer Success Stories | AI Operating System - VAST Data
SI004 VAST Data VAST DataStore: Universal Data Store for the AI Era - VAST Data
SI005 VAST Data The VAST Platform White Paper
SI006 VAST Data VAST Data Valued at $30 Billion as AI Drives a New Infrastructure Stack
SI007 VAST Data VAST Data Introduces Polaris for Hybrid Multicloud AI Orchestration
SI008 VAST Data VAST Data Launches VAST Amplify to Expand Flash Capacity During Shortages
SI009 VAST Data VAST Closes Series E Funding Round, Nearly Triples Valuation to $9.1 Billion
SI010 VAST Data VAST Data: One of the Most Successful Enterprise Software Companies In History
SI011 VAST Data VAST Data Triples Valuation to $3.7 Billion in One Year
SI012 VAST Data VAST Data Named to the Prestigious 2024 Forbes Cloud 100
SI013 VAST Data VAST Climbs the 2025 Forbes Cloud 100
SI014 VAST Data VAST Data partners with CoreWeave for $1.17B to advance AI technology
SI015 CNBC Nvidia backs AI company Vast Data at $30 billion valuation
SI016 Data Center Dynamics Vast Data secures $1bn in Series F funding at $30bn valuation
SI017 Blocks & Files VAST Data raises $1B at $30B valuation as AI storage demand surges
SI018 CRN Vast Data CEO Renen Hallak On AI, Growth And A $30 Billion Valuation
SI019 CRN Vast Data Aims For $1B Round As Demand For AI Infrastructure Surges: Report
SI020 TechCrunch Exclusive: AI storage platform Vast Data aimed for $25B valuation in new round, sources say
SI021 Reuters / Yahoo Finance Exclusive-Alphabet's CapitalG, Nvidia in talks to fund Vast Data at up to $30 billion valuation, sources say
SI022 Calcalist / CTech Vast Data reaches $2 billion in ARR as valuation hits $30 billion | CTech
SI023 SiliconANGLE Vast Data raises $1B at $30B valuation as AI infrastructure demand accelerates - SiliconANGLE
SI024 The Next Web VAST Data’s $30 billion valuation is a bet that the data layer is the real bottleneck in AI
SI025 StorageNewsletter VAST Data and Cisco Expand Partnership
SI026 Sacra VAST Data revenue, funding & news
SI027 U.S. Securities and Exchange Commission SAMSUNG ELECTRONICS CO LTD
SI028 U.S. Securities and Exchange Commission SAMSUNG ELECTRONICS CO LTD
SI029 U.S. Securities and Exchange Commission ALPHABET INC CL A
SI030 CRN Vast Data And CoreWeave Knit $1.7 Billion AI Pact
SE001 VAST Data VAST AI Operating System: Powering the Agentic AI Revolution
SE002 VAST Data VAST Data Platform Services: AI-Powered Discovery Engine
SE003 VAST Data VAST Data Supported Platforms
SE004 VAST Data VAST Data Introduces End-to-End Fully Accelerated AI Data Stack with NVIDIA
SE005 VAST Data VAST Data for Compliance
SE006 VAST Data VAST Data AI Reference Architecture
SE007 VAST Data VAST Data Platform Security Configuration Guide
SE008 VAST Data VAST Data Storage Reference Architecture
SE009 VAST Data Knowledge Base VAST with Kubernetes
SE010 GitHub VAST Data
SE011 GitHub GitHub - vast-data/vast-csi: VAST's Container Storage Interface (CSI) Driver
SE012 GitHub GitHub - vast-data/terraform-provider-vastdata
SE013 GitHub GitHub - vast-data/vastpy: VASTPY is the official Python SDK for the VAST Management System
SE014 GitHub GitHub - vast-data/dataengine-cli: VAST DataEngine CLI - command-line interface for managing DataEngine serverless functions, triggers and engines
SE015 VAST Data GitHub Pages Usage | vast-csi
SE016 GitHub GitHub - vast-data/go-vast-client
SE017 Cisco Solutions - VAST Data on Cisco UCS Data Sheet
SE018 Supermicro VAST Data Platform for Supermicro | Supermicro
SE019 Juniper Networks VAST Storage Configuration | Juniper Networks
SE020 Commvault VAST Data Platform
SE021 Database Trends and Applications VAST DataStore Becomes Universal, Multiprotocol Storage Platform with Block Storage and Event-Processing
SE022 The Next Platform Vast Data Builds Out Data Platform With Block Storage And Kafka Streams
SE023 SiliconANGLE Vast Data expands AI Operating System with global control plane, zero-trust agent framework and deeper Nvidia integration
SE024 StorageNewsletter Vast Forward 2026: Vast Data Introduces End-to-End Fully Accelerated AI Data Stack with Nvidia
SE025 TechArena VAST Data Accelerates Agentic AI at VAST Forward 2026
SE026 CrowdStrike VAST Data and CrowdStrike Partner to Secure the AI Lifecycle
SU001 VAST Data VAST Data Customer Success Stories | AI Operating System
SU002 VAST Data CoreWeave and VAST: Powering the Future of AI
SU003 VAST Data VAST Data partners with CoreWeave for $1.17B to advance AI technology
SU004 CRN Vast Data And CoreWeave Knit $1.7 Billion AI Pact
SU005 SDxCentral Vast Data lands $1.17B CoreWeave deal for AI storage platform
SU006 Data Centre & Network News VAST Data, CoreWeave agree $1.17 billion partnership
SU007 VAST Data NHL Accelerating and Streamlining Media Production Operations
SU008 VAST Data How the NHL Turned a Century of Footage Into a Real-Time AI Engine
SU009 VAST Data The NHL Continues to Uplevel Its Video Production with VAST
SU010 VAST Data SK Telecom and VAST Data Power Sovereign AI
SU011 VAST Data How SK Telecom and VAST Data Rewrote the Rules of GPU Virtualization
SU012 VAST Data VAST, SK Telecom Optimize Korea’s Largest AI Infrastructure with NVIDIA
SU013 VAST Data Invitae Ushering in a New Era of Personalized Healthcare
SU014 VAST Data PacBio Accelerating Genomic Discovery with VAST Data
SU015 VAST Data U.S. Department of Health and Human Services Processing Genomic Data
SU016 VAST Data Jump Trading's Success with VAST Data
SU017 VAST Data Transforming Financial Services: AI-Driven Innovation at HSBC
SU018 VAST Data AI Agents Unlocked: CACEIS Redefines Client Conversations With VAST and NVIDIA
SU019 VAST Data Lawrence Livermore National Labs (LLNL)
SU020 VAST Data Chan Zuckerberg Initiative and AI’s Role in Democratizing Biomedical Innovation
SU021 VAST Data VAST Data Triples Valuation to $3.7 Billion in One Year
SU022 VAST Data VAST Closes Series E Funding Round, Nearly Triples Valuation to $9.1 Billion
SU023 CNBC Nvidia backs AI company Vast Data at $30 billion valuation
SU024 Sacra VAST Data revenue, funding & news
SU025 FeaturedCustomers 138 VAST Data Customer Reviews & References
SU026 PeerSpot VAST Data reviews 2026
SU027 NVIDIA How Nasdaq Is Driving Faster Insights and Smarter Investment Decisions with Scalable AI Innovation
SU028 NVIDIA Case Study: NVIDIA AI Enhances BMW Group's Production Efficiency
SU029 TahawulTech VAST Data partners with NVIDIA on industry first that is designed to eliminate complex AI deployments
SU030 VAST Data VAST Military Unique Deployment Guide
SR001 VAST Data VAST Data Valued at $30 Billion as AI Drives a New Infrastructure Stack The financing included primary and secondary capital, bringing the total transaction value to approximately $1 billion.
SR002 CNBC Nvidia backs AI company Vast Data at $30 billion valuation
SR003 TechCrunch Exclusive: AI storage platform Vast Data aimed for $25B valuation in new round, sources say
SR004 VAST Data VAST Data partners with CoreWeave for $1.17B to advance AI technology VAST Data secures commercial partnership deal with CoreWeave.
SR005 CRN Vast Data And CoreWeave Knit $1.7 Billion AI Pact
SR006 SDxCentral Vast Data lands $1.17B CoreWeave deal for AI storage platform The deal ... will see Vast’s AI OS serve as the primary data storage and management platform underneath the neocloud’s compute infrastructure.
SR007 VAST Data VAST Data Introduces End-to-End Fully Accelerated AI Data Stack with NVIDIA
SR008 Supermicro Supermicro and VAST Data Launch a New Enterprise AI Data Platform Solution with NVIDIA to Accelerate AI Factory Deployment | Supermicro
SR009 VAST Data VAST and Cisco Deliver Turnkey, Scalable Infrastructure for Enterprise AI The VAST AI Operating System is now available directly through Cisco’s Global Price List (GPL) and is fully supported by Cisco as part of the joint solution.
SR010 Cisco Solutions - Cisco, NVIDIA, and VAST Data Solution Overview
SR011 CRN NetApp Sues Former CTO, Alleges He Took Trade Secrets To Rival Vast Data NetApp’s lawsuit against Stefansson does not include Vast Data.
SR012 Blocks & Files Court dismisses NetApp complaint against ex-CTO now at VAST, but NetApp is appealing The complaint was dismissed without prejudice ... NetApp has already appealed.
SR013 Wilson Sonsini The Busy Lawyer’s Guide to the New “Data Export Control” Rules The Final Rules will apply to all transactions ... after the effective date of April 8, 2025.
SR014 White & Case Privacy and Cybersecurity 2025–2026: Insights, challenges, and trends ahead | White & Case LLP
SR015 VAST Data VAST Data Privacy Policy Information regarding the Users will be maintained, processed and stored ... in the United States, and as necessary, in secured cloud storage.
SR016 VAST Data VAST Data End User Services and License Agreement If you are accessing or using the Services ... you do so subject to VAST Data End User Services & License Agreement ... unless you entered into another written contract between Vast Data or its authorized distributor and You.
SR017 VAST Data VAST Data and SDS Partner to Build One of Israel’s Largest Sovereign AI Cloud
SR018 Globes Storage co Vast Data raising several billion dollars - report VAST Data currently has 1,000 employees, half of whom are in Israel, where the main development center is at the Tel Aviv Exhibition Center.
SR019 Israel Innovation Authority Israel Innovation Authority 2025 High-Tech Report: Record Year in Exits and Deep-Tech Growth; Stagnation in Output, Employment, and VC Fundraising
SR020 JD Supra Key Issues Shaping Israel’s Technology Sector in 2025 and Insights for 2026 | JD Supra
SR021 CRN Dell Targets Rivals Pure Storage And Vast Data As AI Race Heats Up Its next-generation storage fabric is delivered on top of hardware from OEMs including Lenovo, Dell, HPE, SuperMicro, and Cisco.
SR022 VAST Data VAST Data Advances AI Storage Leadership with NVIDIA-Certified Storage
SR023 Goldman Sachs Tracking Trillions: The Assumptions Shaping the Scale of the AI Build-Out
SR024 Gartner Gartner Says Worldwide AI Spending Will Total $2.5 Trillion in 2026 Because AI is in the Trough of Disillusionment throughout 2026, it will most often be sold to enterprises by their incumbent software provider rather than bought as part of a new moonshot project.
SR025 StorageMath VAST Amplify's '6x Capacity' Claim: Exploiting a Real Crisis with Fake Math The SSD crisis is real. VAST’s math is not.
SR026 VAST Data Meet Our Partners - AI Data Ecosystem Directory VAST Data customers are supported by a global network of Partners - Value Added Resellers, Global System Integrators, Service Providers, Technology leaders, and cloud providers.
SR027 Supermicro VAST Data Platform for Supermicro | Supermicro
SR028 VAST Data VAST Data Terms of Use
SR029 Government of Israel Artificial Intelligence - Government of Israel national priority report
SR030 Deloitte Israel and F2 Venture Capital Future Forward: Israel’s AI Expansion Blueprint
SV001 VAST Data VAST Data Valued at $30 Billion as AI Drives a New Infrastructure Stack The financing included primary and secondary capital, bringing the total transaction value to approximately $1 billion.
SV002 CNBC Nvidia backs AI company Vast Data at $30 billion valuation
SV003 Reuters via Yahoo Finance Nvidia-backed VAST Data valued at $30 billion in latest funding round The Series F round fetched it about $1 billion in primary and secondary capital.
SV004 HPCwire VAST Tops Off with Series F at a $30 Billion Valuation
SV005 Futurum Group VAST Data Valuation Triples. Can a Unified Platform Scale AI Globally? VAST Data reached a $30 billion valuation following its Series F round, highlighting its central role in AI infrastructure.
SV006 StorageNewsletter Confirmed, Vast Data Raised $1 Billion at $30 Billion Valuation
SV007 Data Center Dynamics Vast Data secures $1bn in Series F funding at $30bn valuation
SV008 Blocks & Files VAST Data raises $1B at $30B valuation as AI storage demand surges
SV009 CTech Vast Data reaches $2 billion in ARR as valuation hits $30 billion
SV010 VAST Data VAST Data: One of the Most Successful Enterprise Software Companies In History
SV011 VAST Data VAST Closes Series E Funding Round, Nearly Triples Valuation to $9.1 Billion
SV012 Reuters via Yahoo Finance Exclusive-Alphabet's CapitalG, Nvidia in talks to fund Vast Data at up to $30 billion valuation, sources say The company earned $200 million in annual recurring revenue (ARR) by January 2025, with projections to grow ARR to $600 million next year.
SV013 U.S. Securities and Exchange Commission Fidelity Select Portfolios N-PORT-P filing for 2026-02-28
SV014 U.S. Securities and Exchange Commission Fidelity Select Portfolios N-PORT-P filing for 2025-02-28
SV015 U.S. Securities and Exchange Commission Fidelity Select Portfolios N-PORT-P filing for 2024-08-31
SV016 CRN Vast Data CEO Renen Hallak On AI, Growth And A $30 Billion Valuation We do multi-year deals, and sell three-year licenses or five-year licenses, and so our annual recurring revenue is smaller as a result.
SV017 Everpure Investor Relations Everpure Announces Fiscal Fourth Quarter and Full Year 2026 Financial Results
SV019 CompaniesMarketCap Pure Storage (PSTG) - Revenue
SV020 NetApp Investor Relations NetApp Reports Third Quarter of Fiscal Year 2026 Results
SV022 CompaniesMarketCap NetApp (NTAP) - Revenue
SV023 CoreWeave Investor Relations CoreWeave Reports Strong First Quarter 2026 Results
SV025 CompaniesMarketCap CoreWeave (CRWV) - Market capitalization
SV026 Lambda Lambda Raises $480M to Expand AI Cloud Platform
SV027 Crusoe Crusoe, the AI factory company, raising $1.375 billion at a valuation above $10 billion to power the future of AI infrastructure
SV028 Together AI Together AI Announces $305M Series B to Scale AI Acceleration Cloud for Open Source and Enterprise AI
SV029 PR Newswire / Together AI Together AI Raises $305M Series B to Scale AI Acceleration Cloud for Open Source and Enterprise AI
SV030 CNBC Are we in an AI bubble? What 40 tech leaders and analysts are saying, in one chart
SV031 CompaniesMarketCap Pure Storage (PSTG) - Market capitalization
SV032 CompaniesMarketCap NetApp (NTAP) - Market capitalization
SV034 Stock Analysis CoreWeave (CRWV) Market Cap & Net Worth
SV035 Stock Analysis NetApp (NTAP) Market Cap & Net Worth
SV036 Stock Analysis Everpure (P) Market Cap & Net Worth
SV037 Reuters via Yahoo Finance AI cloud startup Lambda raises $480 million in new round; Nvidia among investors The company did not disclose its valuation, but sources said this round gave the firm a post-money valuation of $2.5 billion.