初创公司尽调
尽调报告 Space infrastructure / orbital data centers Series A 2026-10-21

Starcloud

技术验证和大额融资先于商业规模运营证据。

Starcloud 已拿出可信的在轨计算验证,值得继续尽调;但在商业化、可靠性、监管和单位经济性证据仍缺失的情况下,$2.3 billion 融资估值偏贵。

覆盖范围与披露说明

公开来源审阅截至 2026-10-21。融资和估值为据报道的交易数据;当前收入、客户、流动性、所有权及规模化单位经济数据仍未披露。

封面要素

据报道员工数 04
25 employees [CO010]
首次商业任务目标 05
2027 year [CO014, CV007]
当前商业收入 06
Not publicly disclosed [CI009, CV010]

公司概况

Starcloud 成立于 2024 年,总部位于 Redmond,正在开发基于卫星的数据中心,用于轨道处理、持久存储以及主权云或韧性云工作负载。Starcloud-1 已在轨道上验证 NVIDIA H100;Starcloud-2 被定位为首次商业任务,目标于 2027 年投入运行。

官网
starcloud.com
创始人
Philip Johnston, Ezra Feilden, Adi Oltean
创立地点
Redmond, Washington, USA
总部
Redmond, Washington, USA
产品
Starcloud 构建轨道计算卫星,将 GPU、存储、通信、热管理和电力系统整合在一起。计划提供的服务包括在轨数据处理、持久存储、备份和主权云容量。
客户
面向寻求低延迟轨道处理,或韧性与主权计算和存储的航天器及空间站运营商、政府和企业。
商业模式
近期销售高溢价轨道处理、存储和主权云容量;若发射和规模化系统经济性具备竞争力,再拓展到更广泛的数据中心工作负载。
阶段
Series A
融资情况
2026 年 8 月宣布完成 $250 million A 轮追加融资,投后估值为 $2.3 billion,使公司披露的累计融资额达到 $450 million。
[CO001, CO002, CO003, CO004, CO005, CO008, CO012, CO014, CI002]

执行摘要

主要优势

  • Starcloud-1 已将 NVIDIA H100 送入轨道,并据报道跑通了 AI 工作负载,公司不再只是纸面概念。
  • 公司已募集大笔资金,也吸引了基础设施合作伙伴和投资者的战略参与。
  • Starcloud-2 勾勒出明确的商业产品,覆盖在轨处理、持久化存储和主权计算。

主要风险

  • 付费客户、当前收入、定价、留存、利用率和客户集中度,尚无公开依据确认。
  • 商业规模化取决于尚未证明的热控和可靠性表现、更低的发射成本、发射服务商的可用性,以及监管批准。
  • 在规模化经济性尚未披露的情况下,$2.3 billion 投后估值快速上升;现金、烧钱速度、股权结构和优先权条款仍属非公开信息。

未决问题

  • 客户合同、发票、工作负载利用率、成果、续约、积压订单,以及按细分市场划分的集中度。
  • 可供独立审阅的任务遥测数据,包括在线率、热控余量、辐射错误、恢复行为和工作负载表现。
  • FCC 一手记录、频谱和碎片方案、审批状态,以及数据治理的司法辖区路径。
  • 当前收入、定价、毛利率、发射和保险成本、现金、烧钱速度、已承诺资本支出和资金跑道。
  • 完全摊薄后的股权结构表、清算优先权、债务、董事会权利和未来融资需求。
  • 面向主权工作负载的安全架构、认证、密钥托管、审计证据和事件响应准备度。

目录

Chapter 01

01公司概览

1.1 身份与运营概况

Starcloud 设计卫星,目标是提供 GPU 计算、持久存储和轨道数据中心服务。公司总部位于 Washington 州 Redmond,一家独立行业媒体将其成立时间追溯至 2024 年。运营进展虽已落地,但仍处早期:Starcloud-1 于 2025 年 11 月将 NVIDIA H100 送入轨道;Starcloud-2 被描述为首次商业任务,计划于 2027 年投入运行。公开披露尚不足以支持当前收入、ARR、客户数或合同金额数据。Tech Insider 报道截至 2026 年 4 月未披露商业收入,但这一有明确日期的观察不能证明 Starcloud 在 2026 年 10 月的收入状况。[CO001, CO002, CO012, CO014, CO016]

KPI 概览表
指标数值或状态日期或期间置信度缺口
总部Redmond, Washington, 美国当前官网与 2026 年报道高法律实体结构未披露
成立时间2024公司历史中具体注册日期未披露
当前阶段A 轮追加融资2026-08-21高轮次完成情况和证券条款未披露
最新融资$250 million 追加融资2026-08-21中已筹资金不等于可用现金
最新据报道估值$2.3 billion2026-08-21中估值基础和稀释情况未披露
此前融资$170 million A 轮融资,估值 $1.1 billion2026-03-30高批次经济条款未披露
收入或 ARR截至 2026-10-21低2026 年 4 月报道称未披露商业收入;没有当前经审计数字
客户被列为 Starcloud-2 客户类别的美国政府机构2026-08-21中数量、合同金额、续约和集中度未披露
员工数25 名员工2026-08-21中没有独立验证的当前员工名册
运营地点Redmond 总部;100,000 平方英尺的 Woodinville 设施2026-08-21中所有权、租赁条款和利用率未披露

值保留各来源的日期和指标标签;null 表示指定证据不支持当前公开数字,且已筹资金不视为可用现金。

[CO002, CO006, CO008, CO010, CO011, CO014, CO016]

1.2 管理层、创始团队匹配度与治理

已披露的创始团队覆盖商业战略、航天器工程和大规模软件基础设施。Philip Johnston 是联合创始人兼 CEO,曾参与卫星项目并接受金融与政策培训;Ezra Feilden 是联合创始人兼 CTO,拥有可展开空间结构经验;Adi Oltean 是联合创始人兼首席工程师,曾参与 Starlink 相关工作并负责生产环境 GPU 集群。这一组合与既定任务高度匹配,但也把外部融资、航天器架构和系统执行集中在三位明确的创始人身上。现有证据没有披露完整的高管梯队、董事会构成、继任计划、所有权或创始人投票控制权,因此治理韧性尚无法评估。[CO003, CO004, CO005]

管理层与创始人表
人员职务相关背景职能覆盖关键人员依赖
Philip Johnston联合创始人兼 CEO二次创业者;McKinsey 卫星项目;国家安全、商业、数学和金融培训战略、融资、商业叙事在商业领导和继任覆盖披露前为高
Ezra Feilden联合创始人兼 CTO卫星设计;可展开太阳能阵列和结构;Airbus Defence & Space、Oxford Space Systems 及 NASA Lunar Pathfinder 工作航天器架构与可展开系统在更广泛的技术领导梯队披露前为高
Adi Oltean联合创始人兼首席工程师曾任 SpaceX 跟踪波束首席工程师;拥有 Microsoft 生产环境 GPU 集群经验及超过 25 项专利太空网络、软件与 GPU 基础设施在系统职责归属和继任覆盖披露前为高

覆盖范围仅限 Starcloud 官网点名的创始人;依赖判断是基于已披露职责范围的分析师判断,不是对人员事件的报道。

[CO003, CO004, CO005]

1.3 资本形成与利益相关方结构

Starcloud 在 2026 年融资估值迅速重定价。多家媒体报道,公司 3 月以 $1.1 billion 估值完成 $170 million A 轮融资,当时累计融资超过 $200 million。随后 TechCrunch 报道,公司 8 月完成 $250 million 追加融资,估值达到 $2.3 billion。追加融资由 Manhattan West Ventures 领投,NVIDIA 和 Cisco 参与;此前披露的投资方还包括 Benchmark、EQT Ventures、NFX、776 Ventures、Y Combinator 等。这些金额是已筹资金,不等于可用现金或资金续航。本文审阅的公开记录没有披露各批次条款、清算优先权、债务、老股转让、当前现金、烧钱速度或备考所有权结构。[CO006, CO007, CO008, CO009]

利益相关方或投资方图谱
利益相关方已披露角色控制权或经济重要性尽调要求
BenchmarkA 轮联合领投主要融资利益相关方;保留证据未确认董事会权利董事席位、所有权、优先权及保护性条款清单
EQT VenturesA 轮联合领投主要融资利益相关方;治理权利未披露所有权、信息权和后续投资承诺
Manhattan West Ventures领投 8 月追加融资最新披露融资的领投方领投条款、批次条件及董事或观察员权利
NVIDIA追加融资参与方兼技术合作伙伴资本加关键加速器路线图关系投资金额、供应承诺、数据权利和排他性
Cisco追加融资参与方战略投资方;商业范围未披露投资金额及任何网络或市场进入义务
NFX已披露融资参与方所有权未披露的财务利益相关方当前所有权和按比例跟投权
776 Ventures已披露融资参与方未披露持股的财务投资者当前持股及按比例认购权
Y Combinator早期投资者及加速器早期股东;当前持仓未披露当前持股及任何持续有效的权利

这只是公开点名投资者的部分清单,不是股权结构表;任何一行都不代表控制权,在缺少持股和证券条款时也无法量化其经济重要性。

[CO006, CO007, CO008, CO009]

1.4 里程碑与执行依赖

Starcloud 已跨过重要技术门槛:首颗卫星于 2025 年 11 月搭载 NVIDIA H100 发射,并在 12 月运行 Gemini 的一个版本、训练 nanoGPT 模型。下一步难度明显更高。Starcloud-2 计划搭载 GPU 集群、存储和持续访问能力,为商业和政府工作负载服务;公司同时扩建 100,000 平方英尺的 Woodinville 设施,并于 2026 年 8 月披露员工数为 25 人。在星座规模上,TechCrunch 和 Via Satellite 报道,公司向 FCC 提交了最多覆盖 88,000 艘航天器的申请。因此,融资里程碑早于可重复制造、发射获取、监管授权、客户经济性和可靠商业运营的证明。[CO010, CO011, CO012, CO013, CO014, CO015]

FO001: 公司里程碑时间线

Starcloud 从成立和发射演示卫星起步,在商业指标公开前就拿到了数十亿美元级融资估值。

来源未给出具体日期的地方采用月份;2027 年是公司目标,并非已经实现的里程碑。

公司成立 [CO002]

Starcloud-1 发射 [CO012]

在轨 AI 演示 [CO013]

宣布 $170M A 轮融资 [CO006]

累计融资据报超过 $200M [CO007]

商业披露缺口仍在 [CO016]

据报完成 $250M 追加融资 [CO008]

Starcloud-2 运营目标 [CO014]

[CO002, CO006, CO007, CO008, CO012, CO013, CO014, CO016]

1.5 图表

Chapter 02

02市场分析

2.1 市场边界与替代方案

可投资市场应限定为轨道计算和存储服务,而不是整个地面数据中心市场。Starcloud 已披露的近期用例包括在下传前处理航天器生成的数据、持久存储、主权备份及部分高性能工作负载。因此,计入的支出应包括托管式轨道计算、在轨数据处理、韧性存储和任务专用基础设施溢价。地面超大规模云厂商、地面边缘站点、直接下传至地球、客户自有航天器处理器和传统备份仍是替代方案,不能自动视为 Starcloud 可捕获的收入。比特币挖矿更适合作为早期硬件的利用率过渡,而不是广泛云需求的证明。只有当轨道系统在商业规模上证明交付成本、可靠性和可用数据传输具备竞争力时,边界才会扩大。[CM001, CM003, CM004, CM005, CM006]

市场定义与替代边界
细分市场纳入的支出排除或替代的支出买方或付款方相关性
太空内数据处理可减少原始数据下传的托管计算客户自有的星载处理器;下传至地面云卫星运营商和任务项目目前证据最充分的近期工作流
主权轨道云不依赖地面站点的高端计算与存储国内主权云区域和强化型地面设施政府、国防及受监管企业潜在高端细分市场;缺少采购证据
韧性备份轨道上的关键数据复制与恢复能力多区域地面备份和离线存档业务连续性和基础设施团队差异化取决于恢复经济性和访问能力
通用 AI 计算仅包括交付后的轨道经济性具备竞争力的工作负载超大规模地面训练和推理云和 AI 基础设施买方长期邻近市场,不是当前可触达需求
比特币挖矿过渡将早期航天器的闲置容量变现利用低成本电力供能的地面挖矿挖矿运营商或航天器所有者利用率过渡方案,而非已验证的云计算 TAM

边界综合自 Starcloud 的产品主张和独立报道;付款方归属、定价和转化均未验证。

[CM003, CM004, CM005, CM006, CM007]

2.2 避免虚假精确的市场测算

现有证据不足以支持严谨的 TAM/SAM/SOM 测算。一篇二手文章给出 2030 年 $3–5 billion 的细分市场区间、将 2029 年 $1.77 billion 推至 2035 年 $39.09 billion 的据引述预测,以及激进的 2035 年 $50–100 billion 情景。文章没有提供底层模型、地理范围、收入定义或一手预测原文,因此这些只是情景标记,并非经过独立验证的市场规模。Starcloud 自有公开材料给出的是产品和容量目标,而不是可审计的收入桥接。可投资的 SAM 应限定为能够承担发射、热管理、通信、保险和更换成本的高溢价工作负载子集。在价格、签约容量、正常运行时间和发射节奏披露前,SOM 无法负责任地量化。[CM007, CM014, CM015]

受证据约束的市场规模测算视角
发布方或视角预测年份地理范围数值CAGR方法论置信度与局限
Tech Insider 专业细分场景2030未披露$3–5B未披露边缘推理和国防专业细分市场低:二手文章没有提供底层模型
Tech Insider 引述的 Introl 预测2035未披露2029 年从 $1.77B 增至 $39.09B67.4%轨道数据中心市场预测低:原始预测不在指定证据池中
Tech Insider 激进场景2035未披露$50–100B未披露假设发射成本下降且超大规模云厂商迁移低:场景区间没有披露推导过程
Starcloud 可服务市场当前未披露能够承受轨道交付成本的高端工作负载未量化:缺少定价、签约容量和单位成本

所有货币区间都只是二手场景标记,并非独立验证的 TAM;null 表示指定证据无法支持该数值。

[CM014, CM015]
FM001: 轨道计算市场证据金字塔

证据从宽泛情景逐步收窄到已展示或近期可商业化的较小能力集合。

各层宽度仅表示层级,不代表数值比例。

各层按主张覆盖范围排序,并非按规模排序;预测值不能与任务数量比较。

[CM001, CM003, CM012, CM014, CM015]

2.3 买方、用户、付款方与采用路径

当前可见两类买方体系。航天器和空间站运营商可能用轨道处理减少原始数据下传、缩短获得洞见的时间;任务运营和载荷团队是用户,项目所有方或政府任务预算可能是付款方。地面买方可能看重主权存储和韧性备份;安全、业务连续性和基础设施团队会使用该服务,采购预算可能来自云、国防或灾备。由于资料池没有公开价格、采购记录或合同金额,这些预算归属仍只是假设。采用路径应是先付费演示,再在 Starcloud-2 上重复运行工作负载,最后只有在独立证明在轨热性能、可用性、数据传输经济性和合规性后,才进入规模化承诺。[CM004, CM005, CM006, CM012]

细分市场、买方与采用路径
细分市场买方用户付款方或预算负责人工作流采用触发因素
地球观测推理卫星运营商载荷和分析团队任务或影像项目预算下传前处理原始传感器数据可测量的延迟和带宽节省
国防任务计算国防项目或主承包商任务运营和情报团队项目或韧性计算预算在轨道传感器附近运行推理安全审批和运行可靠性
主权存储政府或受监管企业基础设施和安全团队主权云或连续性预算将受保护副本存储在地面基础设施之外经审计的安全、访问和恢复控制
关键备份企业或公共部门运营商业务连续性团队灾难恢复预算复制并恢复优先数据集具竞争力的恢复时间、带宽和成本
通用 AI 容量云或 AI 基础设施买方模型和平台团队计算基础设施预算在轨道上训练或服务模型交付成本和正常运行时间达到地面方案水平

用户和工作流依据已披露的产品定位;付款方和预算负责人字段属于尽调假设,不是公开采购证据。

[CM004, CM005, CM006, CM008, CM009]

2.4 驱动因素、约束与估值意义

最强驱动因素包括地面电力和许可压力、下传前处理传感器数据的价值、对主权韧性的需求,以及对非地面基础设施的战略兴趣。约束更为直接:辐射冷却必须持续有效,芯片可靠性不能吞掉节能收益,发射运力稀缺,成本持平路线又依赖 Starship 变得常态化。独立跟踪分析还认为,Starcloud-1 的轨道使其每次绕行约有 38% 时间处于地球阴影中,因此近乎持续日照的前提在商业功率水平上仍未验证。随着星座规模扩大,监管和碎片风险会急剧上升。因此估值应按里程碑加权:经验证的付费利用率、热性能和正常运行时间遥测、发射合同及交付单位成本,应比宽泛市场预测或远期吉瓦级愿景占据更大权重。[CM008, CM009, CM010, CM011, CM013, CM016]

2.5 图表

Chapter 03

03竞争格局

3.1 竞争范围覆盖轨道、发射平台与地面云

竞争格局比其他轨道数据中心初创公司更广。SpaceX 和 Blue Origin 可以在一次采购路径中整合发射、通信、托管和计算;Lonestar 与 Axiom 则覆盖月球存储或托管式轨道计算等相邻用例。地面 GPU 云仍是现实中的主流,因为它们已经销售 AI 基础设施,不受发射、辐射或在轨维护约束。Sacra 将 CoreWeave、Crusoe 和 Lambda 视为争夺同一 AI 基础设施支出的替代方案;Reuters 报道 SpaceX 和 Blue Origin 已披露轨道数据中心雄心。超大规模云厂商自建也具有可信度,因为 Starcloud 点名的生态伙伴包括 AWS、Google 和 Nvidia,这些组织拥有自行开发系统所需的资金和技术能力。因此,竞争边界包括直接同行、可能转为入场者的赋能伙伴、现有巨头以及持续的地面部署。[CP001, CP002, CP004, CP005, CP012, CP013]

竞争者与替代方案画像
竞争者或类别类别已披露规模或融资目标客户或工作负载差异化主要局限或未知项
Starcloud纯轨道计算公司路透社于 2026 年 3 月报道的累计融资 $200M太空原生处理、主权存储及未来通用 AI 计算H100 已送入轨道;商业任务描述了 GPU 集群和持久存储商业正常运行时间、定价、规模化可靠性和合同经济性均未披露
SpaceX垂直整合的发射与轨道平台留存证据未量化公司规模;计划被描述为最多部署一百万颗轨道数据中心卫星整合发射、网络和未来轨道计算掌握发射能力并运营星座留存证据未描述当前在售的独立计算服务
Blue Origin垂直整合的太空平台留存证据未量化融资和估值托管载荷、通信、存储和边缘计算可打包提供轨道托管和基础设施本文未确立其计算产品成熟度、定价和运营规模
Lonestar Data Holdings专业轨道 / 月球数据基础设施Tech Insider 引用过往报道称约 $10M主权存储和归档工作负载月球数据保险库定位部署路径更长、更专业;当前商业规模不清晰
Axiom Space商业空间站平台Tech Insider 引用的融资额超过 $500M托管载荷和基于空间站的计算既有载人航天和空间站平台战略计算是其主营平台的邻近业务,而非唯一业务
CoreWeave、Crusoe 与 Lambda地面 GPU 云替代方案Sacra 报道 CoreWeave 拥有超过 250,000 个 GPU 和 43 座数据中心;其他规模数据未留存通用 AI 训练和推理工作负载可用的地面容量和熟悉的云采购方式不提供太空原生数据本地性或离地韧性价值主张

规模数字是发布方报道的截面数据,未经标准化或独立审计;“未量化”表示留存的章节证据没有确立可比数值。

[CP001, CP004, CP005, CP009, CP012, CP013]

3.2 飞行验证形成差异化,但分销并不独占

Starcloud 当前最强的差异化在于可见的技术验证:官方 Starcloud-1 页面称,2025 年 11 月任务搭载 Nvidia H100,随后在轨运行 Gemini 并训练 nanoGPT 模型。计划中的商业系统还将加入 GPU 集群、持久存储、持续访问以及专有热管理和电力系统。这些证据让 Starcloud 区别于只有概念的入场者,但不能证明规模化可靠性、独家供应或持久客户锁定。Reuters 描述了 Starcloud 与 Nvidia 以及 Amazon、Google 云业务的关系,但这些关系既能扩大分销,也会让架构暴露给有能力的未来竞争者。多云并用在结构上可行,因为买方可以继续使用地面计算,同时测试轨道工作负载;而对发射的依赖又赋予垂直整合供应商议价能力。因此,护城河必须从一次性“首个”转移到运营遥测、可重复制造、预留容量和整合式客户工作流。[CP002, CP006, CP007, CP008, CP016]

能力与分销证据矩阵
购买标准StarcloudSpaceXBlue Origin地面 GPU 云其他轨道专业公司
已飞行的轨道计算证明声称已完成 H100 任务和在轨模型演示有轨道数据中心计划报道;本文未确立产品证明报道了类似雄心;本文未确立产品证明非轨道;运营地面计算Tech Insider 报道了 Axiom 节点和 Lonestar 载荷活动
发射垂直整合否;依赖发射服务商是,结构上已整合是,结构上已整合不适用没有自有发射能力的留存证据
当前商业 GPU 容量没有普遍可用容量的公开证据轨道计算未知轨道计算未知地面供应商整体上具备未知
太空原生处理价值主张明确瞄准太空内数据分析可能包含在更广泛的平台内托管载荷和边缘计算价值主张否是,采用月球或空间站专用方案
公开定价与产品打包未知轨道计算未知轨道计算未知留存证据未作比较未知
分销杠杆已点名云和芯片合作伙伴,但未显示排他性发射加星座控制发射和托管打包既有云采购和已部署容量专业渠道;规模未知

单元格只总结留存的指定证据;“unknown”是有意保留的状态,不代表能力不存在。

[CP001, CP002, CP004, CP005, CP006, CP007, CP012, CP013, CP015, CP016]
FP001: 按成熟度与轨道计算重点划分的有证据竞争定位

序数评分区分当前部署成熟度与轨道计算专业化程度;这些是分析性定位,不是实测市场份额。

条目部署成熟度 (1–5)轨道计算专业化 (1–5)说明
1. Starcloud 3 5 已飞行的 H100 演示器和商业任务计划,但未披露规模化服务。
2. SpaceX 4 4 成熟的发射与星座平台;留存证据中的轨道数据中心计划仍属预期。
3. Blue Origin 2 3 具备一体化轨道托管野心,但留存证据未能证明计算成熟度。
4. Lonestar 2 5 高度专业化的月球存储方案,已披露规模有限。
5. Axiom Space 3 4 拥有空间站和托管载荷平台,据报配备轨道计算节点。
6. 地面 GPU 云 5 1 已实现规模化运营的 AI 基础设施,但不在轨道上。

评分是根据所引证据作出的序数判断:1 表示低,5 表示高。它们不是量化基准,未知字段没有按零分处理。

[CP004, CP007, CP012, CP013]

3.3 定价证据不足以证明切换价值

保留的证据没有披露 Starcloud 商业容量的公开标价、合同计价单位、服务级别承诺或折扣表。产品页面描述能力并邀请潜在客户联系;Sacra 将业务定性为商业化前阶段,并将有意义的经常性计算收入指向 2027 年。因此,买方无法在全成本口径下比较轨道容量与地面 GPU 小时。切换成本也不对称:把空间原生数据处理移到更靠近传感器的位置,可能减少下传负担;但通用 AI 买方可以保留地面容量,只把轨道计算用于特定工作负载。Starcloud 可能通过航天器集成、数据本地性或容量预留形成锁定,但保留的公开证据尚未证明其中任何一项。因此,定价尽调需要已签合同、工作负载层面的经济性、续约机制,以及通过云合作伙伴接入不会让 Starcloud 沦为其分销渠道附属品的证据。[CP004, CP006, CP015, CP016]

定价与产品打包透明度
产品或替代方案公开价格或合同单位有证据支持的包含能力折扣或承诺披露竞争含义
Starcloud-2未披露GPU 集群、持久存储、全天候访问、热管理和电力系统未披露技术范围可见,但无法比较买方经济性和服务水平
Starcloud 主权云与存储未披露全球数据存储和主权云定位未披露潜在高端用例,但没有支付意愿或续约条款的证据
地面 GPU 云替代方案本章未留存CoreWeave、Crusoe 和 Lambda 提供的通用 AI 基础设施未留存它们仍是现状方案,轨道系统测试期间仍可使用
垂直整合的太空平台轨道计算未披露发射、通信、托管或边缘计算打包方案未披露即使没有可直接比较的标价,打包也会给纯轨道计算供应商施压

没有留存来源提供标准化标价或实际成交价;每个缺失值都是尽调缺口,而不是零值。

[CP004, CP005, CP006, CP015, CP016]

3.4 护城河依赖执行,也容易被捆绑

Starcloud 的竞争论点暴露在同一条吸引入场者的成本曲线上。Reuters 报道发射成本仍是挑战,管理层预计轨道系统要到 2028 或 2029 年左右才具备成本竞争力。Data Center Dynamics 另称,早先宣称以 $8.2 million 部署 40 MW 的假设可能过于乐观。若发射成本下降,Starcloud 会受益,但 SpaceX 和其他发射平台也会更像可信的整合型竞争者;若成本维持高位,地面云则保留经济优势。CIO Business Views 将发射经济性、基础设施可扩展性、网络安全和运营可靠性列为成功条件,而非已经解决的优势。Starcloud 确实在飞行 GPU 验证上领先,但可防守资产不只是概念。持久优势需要商业正常运行时间、制造节奏、信任控制,以及能经受现有巨头捆绑或超大规模云厂商自建的合作条款。[CP003, CP011, CP013, CP014, CP016]

3.5 图表

Chapter 04

04财务

4.1 收入模型:基础设施规模化前先做高溢价轨道服务

Starcloud 描述的是两个变现阶段,而不是成熟且已披露的收入结构。近期模型是向航天器运营商和地面用户销售在轨处理、持久存储和主权云容量;长期模型则是在发射经济性允许更大系统后,争夺更广泛的数据中心工作负载。Starcloud-2 被定位为首次商业任务,具备 GPU 集群、存储和持续访问,并分别面向在太空分析和地面主权计算提出服务方案。这些产品形态构成了合理的可收费服务,但没有披露合同结构、承诺容量、利用率、确认收入或毛利。因此,收入桥接是机制图,而不是预测。[CI008, CI010, CI011, CI012]

收入来源与证据状态
收入来源机制计费单位当前公开状态收入质量尽调要求
太空内数据处理下传前处理航天器和空间站数据Starcloud-2 已提供;收入未披露可能按用量或预留容量计费提供已签合同、利用率、每项工作负载价格和收入确认政策
持久轨道存储在 Starcloud-2 上存储数据产品能力已披露;收入未披露可能按经常性容量计费提供签约容量、期限、冗余 SLA 和续约数据
主权云计算不依赖地面基础设施的高端计算和备份产品价值主张已披露;收入未披露可能面向企业或政府的高端服务提供买方管线、安全要求、合同金额和交付里程碑
未来地面工作负载基础设施发射成本下降后销售更大规模的计算管理层的长期判断取决于发射成本和利用率提供容量计划、客户承诺、发射假设和分阶段资本开支计划

公开描述只确立了服务机制;计费单位、实际成交价、利用率、收入和收入结构均不可得。

[CI008, CI010, CI011, CI012]
FI001: 从工作负载到毛利的收入模型桥

Starcloud 披露了运营流程,但定价、收入、利用率和毛利仍不可得。

节点 / 连接节点 / 起点终点说明
节点 1 客户工作负载 [workload]航天器数据处理,或地面主权计算与存储需求。
节点 2 轨道容量 [capacity]GPU 集群、持久存储、热控系统、电力系统和持续访问。
节点 3 交付服务 [service]在轨分析、存储、备份或高端云计算。
节点 4 计费机制 [billing]公开信息不可得;可能按用量、预留容量、里程碑或合同计费。
节点 5 确认收入 [revenue]公开信息不可得;未披露收入结构或确认政策。
节点 6 毛利 [gross-profit]公开信息不可得,且取决于发射、硬件、保险、网络、运营和利用率。

该流程是分析性构建,不代表公司披露了转化率、价格、收入或利润率。

[CI006, CI008, CI010, CI011, CI012]

4.2 定价与牵引力仍属私有信息

公开材料没有提供标价、实际成交价、最低承诺、折扣表或合同期限。官方任务页面反复引导潜在供应商和客户填写联系表;独立的 4 月分析则称当时没有披露商业收入。Starcloud-1 证明地面 H100 级 GPU 能在轨运行,CNN 也独立描述公司正在开发轨道数据中心,但两者都不能证明经常性收入、付费利用率、留存或客户集中度。因此,收入质量尚无法拆分为合同型、按使用量计费、里程碑型或合作伙伴试验活动;销售周期和渠道经济性也仍无公开依据。[CI009, CI013, CI014]

定价与变现披露
产品标价实际成交价合同模式公开证据投资判断含义
Starcloud-2 计算未披露官方页面邀请直接联系没有依据估算每 GPU 收入或利用率
持久存储未披露已披露能力,但没有价格或计费单位无法与地面或太空存储替代方案比较
主权云与备份未披露定位为高端服务高端支付意愿和服务成本尚未验证
未来大规模计算未披露管理层讨论了电力成本平价条件电力成本目标不能确立客户价格或毛利率

Null 表示截至 2026-10-21,在指定的预取证据中未找到公开标价或实际成交价。

[CI006, CI009, CI014]

4.3 单位经济性取决于发射节奏与系统利用率

最具体的公开经济目标是管理层预计未来 Starcloud-3 级系统的电价约为 $0.05/千瓦时,但明确以商业发射成本接近 $500/公斤 为前提。管理层还承认,轨道能源成本要等 Starship 频繁飞行后才具备竞争力,因此当前经济性依赖更小的任务和高溢价用例。这是电力成本目标,不是毛利率:其中没有纳入航天器制造、发射与保险、更换节奏、通信、地面运营、折旧、故障损失、支持和闲置容量的公开证据。缺少这些输入,CAC、回本期、贡献利润率、毛利率和投入资本回报率都无法负责任地计算。[CI006, CI007, CI016]

单位经济性与资本充足度
指标公开数值置信度重要性尽调要求
未来轨道电力成本目标$0.05 per kWh中等;TechCrunch 报道的管理层目标是未来实现成本平价的关键输入,但不是完整的成本结构将能源目标与每个计算单位的全包成本核对
发射成本假设$500 per kg中等;管理层模型中的明确条件经济性对外部控制的发射市场高度敏感提供当前及 Starship 发射节奏延迟情况下的敏感性分析
商业收入低;无公开金额无法评估业务进展和收入质量按产品提供月度收入,并区分已确认收入和合同收入
毛利率低;无公开金额无法检验高端服务是否能覆盖发射和硬件成本提供包括发射、硬件、通信、保险和运营在内的 COGS 桥接表
现金、消耗与资金续航低;融资已披露,但流动性未披露决定下一轮融资的时点和稀释风险提供现金余额、月度消耗、已承诺支出、债务和资金续航
CAC、回本周期与销售周期低;无公开数据决定商业推进的可扩展性提供销售线索转化率、销售周期长度、获客成本、合同期限和回本周期

只有功率和发射成本数据是公开的管理层假设;所有 null 都是证据缺口,不代表数值为零。

[CI006, CI007, CI009, CI015, CI016]

4.4 资本充足性:资金充裕,但无法评估资金续航

2026 年 8 月,Starcloud 宣布完成 $250 million A 轮追加融资,投后估值为 $2.3 billion,使公司披露的自 2024 年以来累计融资达到 $450 million。公司称新资金将用于扩充制造、与 NVIDIA 开展工程合作以及采购发射额度。这笔资本规模可观,但已筹资金不等于可用现金:公告没有披露手头现金、月度烧钱、营运资金承诺、债务、发射订金、保险义务或资金续航。因此财务结论是混合的。Starcloud 具备融资能力,也有可信的服务概念,但收入质量、利润率路径、资本密集度及下一轮融资触发点仍是重大尽调阻碍。[CI001, CI002, CI003, CI015]

4.5 图表

Chapter 05

05产品与技术

5.1 产品定义与成熟度

Starcloud 的产品更适合被理解为分阶段推进的轨道计算服务,而不是一个完成度高的单一数据中心 SKU。Starcloud-1 是验证器:公司称其于 2025 年 11 月搭载 NVIDIA H100 发射,在轨运行 Gemini 的一个版本并训练 nanoGPT 模型。这些事件构成有意义的功能验证,但引用材料没有提供工作负载持续时间、错误率、热遥测、利用率或客户验收标准。Starcloud-2 是商业化步骤。官方页面规定其搭载 GPU 集群、持久存储、持续访问及专有电力和热管理系统,目标是在 2027 年于太阳同步轨道运行。计划中的工作流服务两类任务:在下传前处理航天器数据,或为地面用户提供远程存储和主权云容量。公开的 Starcloud-4 页面尚不是产品规格,因此后续代际仍是路线图占位符,而不是可据此承保的资产。[CE001, CE002, CE003, CE004, CE005, CE016]

产品模块与资产矩阵
产品或模块买方或用户公开成熟度声称的差异化尽调缺口
Starcloud-1技术演示用户2025 年 11 月发射;公司报告的演示首颗进入轨道的 NVIDIA H100,以及轨道内模型工作任务遥测、工作负载时长和错误率
Starcloud-2 太空计算航天器与空间站运营商首次商业任务;目标是在 2027 年投入运行在下行传输前处理大批量原始数据已预订容量、集成测试和实测链路可用性
Starcloud-2 地面云主权数据与关键数据用户公司描述的未来服务存储和计算不依赖地面设施安全控制、租户隔离、恢复目标和服务级别
Starcloud-4未公开说明官方页面已命名,但没有技术规格现有留存证据无法支持配置、时间安排、客户和资质计划

截至 2026-10-21,状态和差异化仅限于留存的公开证据;null 或未说明项目属于尽调缺口,不代表能力为零。

[CE001, CE002, CE003, CE004, CE005, CE016]

5.2 架构与运营工作流

该架构把普通 GPU 工作负载接入要求极高的航天器基础设施。Starcloud 描述了位于激光链路接入层后的标准地面芯片、持久存储、软件辅助的辐射容错、浸没式板级冷却、可展开散热器和轨道电力系统。面向客户的流程对在太空用户更具体:摄取高容量传感器数据,在源头附近运行推理或过滤,只下传选定洞见,而不是全部原始数据。地面主权云方案成熟度更低,因为现有证据无法确认地面接入拓扑、恢复目标、租户隔离、密钥管理或可执行的服务级别。创始人访谈给出每秒数百吉比特的网络目标和低至 50 毫秒的延迟,但这些是设计主张,而非观察到的 Starcloud-2 性能。因此,图示映射的是声称的组件和输出,不应视为已验证的生产架构。[CE003, CE004, CE005, CE009, CE010, CE011, CE012]

工作流与用例证据
用户任务当前瓶颈Starcloud 方案公开证据可支持的收益局限
分析航天器传感器数据原始数据下行量和传输窗口在轨运行 GPU 分析只下行选定洞察,而不是全部原始数据没有留存的吞吐量、延迟或准确率实测数据
存储关键地面数据依赖地面基础设施具备持续访问能力的持久化轨道存储物理位置远离地面的存储地点未披露恢复目标、安全审计或可用性记录
运行高性能云工作负载地面电力、冷却和许可限制由轨道电力和热控系统支持的 GPU 集群据报道,一项演示在轨运行了 H100 工作负载商业规模经济性和可重复的可靠性仍未得到验证

收益来自公司声称,或由所引来源限定了分析边界;没有任何一行能证明已签约的服务级别或生产客户结果。

[CE001, CE003, CE004, CE005, CE011, CE015]
FE001: Starcloud 声称的产品架构

六层有证据边界的架构,从工作负载接入连接到交付的轨道计算输出。

工作负载接入
航天器数据,或地面存储与计算请求
模块
  • 航天器传感器
  • 地面用户
输出
  • 原始数据或工作负载请求
轨道连接
声称可高速接入星座骨干网络
模块
  • 激光终端
  • 第三方卫星网络
输出
  • 数据进出
计算与存储
商用 GPU 硬件和持久存储
模块
  • NVIDIA H100 或计划中的 GPU 集群
  • 持久存储
输出
  • 推理、训练或存储数据
辐射可靠性
公司描述的面向地面电子设备的防护
模块
  • 屏蔽
  • 软件容错
输出
  • 声称具备的错误韧性
热控与电力
支撑航天器运行的基础设施
模块
  • 导电液体电路板冷却
  • 可展开散热器
  • 轨道电力系统
输出
  • 散热和电力
服务交付
面向太空内外用户的输出不同
模块
  • 在轨分析
  • 主权存储与云
输出
  • 精选洞察或远程数据访问

这是对公司描述组件的综合,不是已建成拓扑或经过验证的性能模型。

[CE004, CE005, CE009, CE010, CE011, CE013]

5.3 差异化与关键依赖

Starcloud 的差异化在于:既利用航天器数据在轨处理的近场优势,又让商业地面 GPU 适应太空运行。已报道的 H100 验证比纸面设计更能支撑这一点,Redmond 载荷设施也显示出一定的内部制造能力。但所谓护城河很大一部分仍停留在未经验证的实现细节上:屏蔽几何、软件恢复行为、整板液冷传热、可展开散热器质量、光链路可用性以及与其他星座的集成。公司称当前架构并不模块化,只有在单星规模大得多之后才会模块化。这种节奏把早期技术和运营风险集中到每艘航天器上。独立报道还提到拟采用 Blackwell 硬件、AWS Outposts 和激光互操作性,但保留的主要技术文件或合作伙伴确认均未证明这些集成。 产品护城河因此应被评估为已验证的计算能力加上公司声称的系统专业能力,目前尚不是已证明的轨道云平台。[CE001, CE007, CE009, CE010, CE011, CE012, CE013, CE014]

技术与运营架构
层级或组件角色依赖主要风险
地面 GPU 计算推理、训练和数据处理针对轨道环境改造的标准芯片辐射错误以及无法维修硬件
抗辐射能力保护并恢复计算状态屏蔽加软件容错没有公开的单粒子翻转率或恢复测试证据
热量排散将板卡热量排向太空导热液体和可展开散热器质量、展开、泄漏和散热余量未披露
光学连接将轨道计算连接到用户和网络激光终端和第三方星座骨干网互操作性和持续可用性未经独立确认

架构描述主要来自创始人访谈;落地细节和性能数据未经独立验证。

[CE009, CE010, CE011, CE012, CE013, CE015]

5.4 信任、可靠性与路线图风险

企业和主权买方要求最严苛的地方,公开证据反而最薄弱。保留的官方页面没有披露可用性历史、辐射事件表现、热裕量、冗余设计、事件响应、加密控制、认证、隐私治理、数据驻留解释或生命周期终止恢复。Starcloud 宣传其独立于地球、处于高度冗余环境,但这些是产品主张,不是经审计的控制。一篇信誉较低的从业者讨论有用地指出发射成本、辐射、碎片和升级约束仍是未解决的经济与运营风险;它是风险提示,不是基准。同样,官方宣传页描述了近期更大且盈利的后续任务,却没有提供经济性或鉴定数据。在把 Starcloud-2 视为商业级基础设施前,尽调应要求任务遥测、部件鉴定、失效模式测试、安全架构、第三方集成确认和合同服务目标。[CE005, CE006, CE008, CE009, CE015, CE016]

5.5 图表

Chapter 06

06客户

6.1 买方图谱合理,但需求尚未落地

Starcloud 的公开定位区分出两类潜在需求池。对在太空用户,Starcloud-2 被定位为航天器和空间站数据的在轨处理平台,尤其针对原本需要大规模下传的地球观测工作负载。对地面用户,公司推销独立于地面基础设施的主权计算、韧性存储和备份。这些方案对应不同买方、采购路径和支付意愿:卫星运营商看重延迟和避免带宽消耗,政府与企业则会购买韧性或主权能力。NVIDIA 更适合被归类为技术伙伴和投资方,而不是客户验证。公开来源没有分部客户数、签约容量、价格或收入贡献,因此各分部的市场匹配度仍是尽调假设,而非已观察到的商业结果。[CU001, CU002, CU003, CU007, CU008, CU013]

客户细分与验证成熟度
细分市场买方 / 用户 / 付款方用例公开规模证据收入或战略价值主要缺口
在轨数据运营商卫星运营商购买;任务和分析团队使用在下行前处理地球观测和航天器数据公司称每天生成数 TB 原始数据,但没有客户分母可能降低延迟和原始数据传输量没有具名运营商部署、工作负载量、价格或结果
主权与韧性计算政府或企业买方;安全和 IT 团队使用独立存储、备份和高价主权云计算公司称 Starcloud-2 具备产品能力;没有客户数量战略韧性可能支撑溢价没有采购记录、认证、合同或付费部署
云和 AI 基础设施合作方云平台或基础设施研发组织测试轨道服务器、加速器和推理工作负载一个次级来源点名 AWS、Google Cloud、NVIDIA 和 Crusoe工程参与可以验证集成商业条款和付款方身份未获交叉印证
AI 研究人员和模型开发者研究或 ML 团队;付款方仍未确定在轨训练或运行 AI 模型Starcloud-1 演示了模型执行和 NanoGPT 训练演示降低了技术采用风险没有第三方付费工作负载证据
NVIDIA 生态关系NVIDIA 同时是供应商、投资方和技术合作伙伴GPU 飞行验证以及获得 NVIDIA 专业支持H100 已搭载 Starcloud-1 飞行;未来卫星计划围绕 Space-1 建设改善硬件获取并增强技术可信度该关系不等于客户收入证据

细分市场依据截至 2026-10-21 的 Starcloud 和合作方描述整理;公开规模陈述是公司口径,null 商业指标不按零处理。

[CU001, CU002, CU003, CU005, CU006, CU007, CU008, CU010, CU013]
FU001: 从用例兴趣到持久收入的有证据路径

Starcloud 已跨过技术演示里程碑,但付费生产使用、续约和扩张仍未得到验证。

阶段顺序是分析性安排,不代表转化率、经过时间或共同客户群体。

[CU001, CU002, CU003, CU004, CU005, CU010, CU014]

6.2 技术采用先于商业验证

采用路径已从演示资产推进到计划中的商业平台。Starcloud 称 Starcloud-1 于 2025 年 11 月搭载 H100 发射,随后在轨运行模型;NVIDIA 独立描述了任务硬件及其 Inception 支持,GeekWire 后来报道了 NanoGPT 里程碑。Starcloud 称 Starcloud-2 是首次商业任务,并计划在 2027 年提供 GPU 集群、持久存储和持续访问。然而,唯一点名 AWS、Google Cloud、NVIDIA 和 Crusoe 作为 Starcloud-2 工作负载参与方的指定来源是 Tech Insider。这是有用的线索证据,但不足以证明已签合同、付款、生产部署或可量化客户结果。因此,发射和在轨运营里程碑先验证产品准备度,再谈可重复需求。[CU004, CU005, CU006, CU010, CU011, CU012]

与客户相关的采用轨迹
指标或里程碑数值日期或期间来源归属置信度含义缺失的分母或限定
已送入轨道的数据中心 GPUStarcloud-1 上的一块 NVIDIA H1002025 年 11 月Starcloud、NVIDIA 和 GeekWire中建立技术演示的可信度不衡量用户、付费工作负载或重复使用
轨道内模型里程碑据报道完成 NanoGPT 训练及其他模型执行2025 年 12 月里程碑Starcloud 和 GeekWire中证明工作负载可以在轨道资产上运行没有外部客户、工作负载时长、利用率或结果
商业平台可用性Starcloud-2 被描述为已在太阳同步轨道全面运行2027 年前,公司目标Starcloud低确定最早的公开生产服务里程碑前瞻性目标;发射成功、容量和服务级别未知
具名 Starcloud-2 合作方据报道有“四家机构:AWS、Google Cloud、NVIDIA 和 Crusoe”2026 年 4 月报道Tech Insider低暗示商业任务周边存在集成兴趣单一来源报道;付款方身份、合同和工作负载范围未经验证

里程碑混合了已完成的演示和面向未来的公司目标;它们不是可比的客户漏斗,也不能证明转化或收入。

[CU004, CU005, CU006, CU010, CU012]
具名客户验证表
组织细分市场报道的部署或用例验证阶段报道的结果局限
NVIDIA技术供应商和投资方Starcloud-1 搭载 H100,以及未来 Space-1 模块合作已演示的技术合作关系H100 已在轨运行并支持模型训练没有 NVIDIA 是付费客户的证据
AWS云基础设施合作方据报道计划为 Starcloud-2 提供服务器刀片或软件工作负载报道的发射前集成没有公开测量结果仅由一个留存的次级来源点名;合同和付款未经验证
Google Cloud云基础设施合作方据报道计划为 Starcloud-2 提供工作负载报道的发射前集成没有公开测量结果仅由一个留存的次级来源点名;合同和付款未经验证
CrusoeAI 云合作方据报道计划为 Starcloud-2 提供边缘推理工作负载报道的发射前集成没有公开测量结果仅由一个留存的次级来源点名;合同和付款未经验证

仅列出指定抓取来源中点名的组织;“合作方”不等于客户、生产部署或已确认收入。

[CU005, CU006, CU007, CU008, CU010]

6.3 持久性与集中度尚无法测量

保留的来源没有报告毛收入或净收入留存、流失、续约率、合同期限、队列表现、客户满意度、订单积压、最大客户占比或渠道集中度。审阅的官方页面大多邀请供应商和客户联系公司,这说明市场开发活跃,但不能证明需求已成交。Tech Insider 报道截至 2026 年 4 月没有披露商业收入,并将发射或在轨故障列为近期最大风险;其另称 Starcloud-2 实际上已售罄,但指定资料池没有合同层面的佐证。在管理层提供客户台账和合同证据前,尽调应把扩张视为尚未证明,把集中度视为未知,而不是低。下一项关键验证是成功任务后仍能续约的付费工作负载,并按分部披露容量、收入和客户依赖。[CU009, CU010, CU013, CU014, CU015]

6.4 图表

Chapter 07

07风险

7.1 运营与技术暴露

首次任务显著降低了可行性风险,但不能证明生产可靠性。Starcloud 和 CNBC 报道 Starcloud-1 搭载 NVIDIA H100 并在轨运行推理和训练工作负载;Data Center Frontier 则将该任务描述为对真空、辐射和热循环环境下运行能力的测试。商业规模带来另一套风险边界。Control Plane 估算,排出一兆瓦废热需要约 1,200 平方米散热器面积;CNBC 引述分析师对辐射、维护、碎片、数据治理和太空交通的担忧。Starcloud-2 仍被公司描述为未来商业任务,搭载 GPU 集群、存储以及专有热管理和电力系统。因此,在独立审阅的遥测显示持续热裕量、错误率、正常运行时间、恢复行为和工作负载性能之前,剩余暴露仍然很高,不能只凭一次成功演示降低判断。[CR002, CR003, CR004, CR005, CR006, CR007, CR015, CR016]

运营 / 质量 / 安全风险登记表
失效模式可能性严重性缓解成熟度剩余暴露未解决证据
下一次商业任务发射或早期在轨失败中严重早期:Starcloud-1 证明了规模更小的轨道 GPU 演示高独立资质、保险、发射准备和恢复证据
热量排散无法随计算功率扩展高严重早期:公司声称拥有专有热控系统;商业遥测未公开高散热器测试余量、热负荷结果、衰减和安全模式行为
辐射、碎片或维护限制拖累可用性中高除演示任务外未经验证高辐射错误率、屏蔽余量、碰撞计划、可维修性和组件寿命数据
商业正常运行时间或工作负载性能未达到 2027 年运营计划中高计划中:GPU 集群、存储和全天候访问是公司口径高客户验收标准、正常运行时间 SLO、基准测试结果和持续遥测
安全或数据治理控制无法支撑主权工作负载中高公司声称拥有主权和冗余存储;独立控制未披露高架构审查、加密和密钥托管、审计日志、认证、DPA 和事故处理流程

严重性和可能性是结合所引任务证据及披露的风险机制作出的分析师评估;公开遥测、事故或控制文件缺失被记录为证据缺口,不视为失败已经或尚未发生的证明。

[CR002, CR003, CR004, CR005, CR006, CR007, CR014, CR015, CR016]
FR001: 风险热力图:影响、可能性与缓释成熟度

按分析判断对六个最重大的风险簇排序;标签反映有证据支持的判断,未知可能性或未经验证的缓释措施会明确标示。

序数标签是分析师截至 2026-10-21 的判断,不是实测概率;“未知”表示指定来源池未披露相关证据。

[CR001, CR002, CR003, CR008, CR010, CR012, CR013, CR014, CR015, CR016]

7.2 监管与法律暴露

公开证据证明的是监管雄心,而不是监管许可。Wikipedia 和 GeekWire 报道,FCC 申请最多覆盖 88,000 颗轨道数据中心卫星,但指定来源集中没有底层案卷、许可条件、频谱协调记录、碎片计划或审批状态。CNBC 另报道 Morgan Stanley 分析师提出的数据治理和太空交通担忧。历史法律挑战也暴露出执行摩擦:Wikipedia 报道,公司在 Lumen Technologies 发起挑战后,将原名 Lumen Orbit 改为 Starcloud,但保留文本没有披露案件条款。这些都不等于执法或拒批证据,而是需要通过监管机构和法律一手记录填补的尽调缺口;在此之前,不应把星座规模、辖区数据主张或经营自由视为已去风险。[CR003, CR008, CR009, CR014, CR016]

监管 / 法律风险登记表
规则、许可或案件管辖区公开状态可能性严重性缓解成熟度剩余暴露尽调路径
FCC 星座申请美国 / 轨道运营据报道申请最多 88,000 颗卫星;未留存主要案卷和批准状态中严重早期;分阶段任务展示了路线图,但不能证明星座获批高获取 FCC 申请、修订文件、协调记录、里程碑和当前处置结果
数据治理与空间交通义务美国及客户数据所在管辖区分析师通过 CNBC 识别出风险;适用批准和控制未列明中高在指定公开证据中未经验证高按服务梳理许可、隐私和本地化义务、避碰控制、保险及事故报告要求
Lumen Technologies 命名挑战美国Wikipedia 报道该挑战促成 2025 年品牌重塑;来源集中未留存条款和持续义务低中品牌重塑已完成;来源集中没有公开和解或案件条款低至中审阅诉状、和解条款、商标转让以及任何仍有效的限制

截至 2026-10-21 的部分公开来源登记表;可能性、严重性、缓解成熟度和剩余暴露是基于所引证据的分析师评估,不是监管机构结论;主要记录缺失被视为缺口,而非获批或违规的证明。

[CR003, CR004, CR007, CR008, CR009, CR014, CR016]

7.3 依赖、融资与论点失效触发因素

发射经济性是最明确的外部依赖。Control Plane 报道 Falcon 9 的交付成本约为 $3,600/公斤,而 Starcloud-3 的经济性要求约 $500/公斤;频繁的商业 Starship 服务目前尚不可用。同一来源提到,小型任务可用 Falcon 9 兜底,但这一方案无法填平成本差距。融资提供了推进路线图的能力,却不能证明路线图资金已经覆盖到盈亏平衡:3 月报道为 $170 million、$1.1 billion 估值;8 月报道则为追加 $250 million、累计融资 $450 million、投后估值 $2.3 billion。这些数字都没有披露可用现金、烧钱、已承诺发射义务或资金续航。因此,论点失效触发因素应包括 Starcloud-2 任务失败或重大延迟、热性能低于设计需求、无法取得必要批准、Starship 经济性长期远高于模型阈值,或融资无法跨过商业爬坡期。[CR001, CR007, CR010, CR011, CR012, CR013, CR015, CR016]

合作方 / 依赖风险登记表
依赖对手方角色集中度失败情景严重性缓解措施剩余暴露
低成本重型运载火箭SpaceX / Starship实现模型所示 Starcloud-3 发射经济性的必要路径高商业发射频次或降本进度延迟,或成本仍高于约 $500/kg 门槛严重继续执行规模更小的 Falcon 9 任务高,因为备用任务无法证明成本达到同等水平
GPU 平台与太空资质NVIDIA 及相关硬件生态已演示和计划任务的核心计算硬件高供应、资质或路线图变化延误商业载荷高H100 已在轨运行,提供初步技术证明中至高,直至多 GPU 可靠性得到独立测量
外部资本与发射承诺现有及未来投资方和发射服务商为制造、工程和未来发射采购提供资金高现金消耗或已承诺义务在商业规模化前耗尽可用资本严重据报道,2026 年融资扩大了资源高,因为现金、消耗、义务和资金续航均未披露

依赖集中度和剩余暴露是基于已披露架构、发射经济性和融资用途作出的分析判断;据报道筹得的资金不按可用现金或资金续航处理。

[CR001, CR004, CR005, CR010, CR011, CR012, CR013, CR016]

7.4 图表

Chapter 08

08估值

8.1 投资决策与价格纪律

最新披露的价格是 2026 年 8 月 $250 million A 轮追加融资对应的 $2.3 billion 投后估值;3 月融资据报道估值为 $1.1 billion。约 2.1 倍的加价发生在最有力的保留公开收入表述仍称截至 4 月未披露商业收入之时。融资显示投资者需求,也增加了筹资额,但不能证明当前可用现金、ARR 或站得住脚的收入倍数。建议为继续研究,信心中等,风险高,估值判断偏贵。应等待可独立审阅的运营证据或明显更好的条款再入场;泛泛的热情折价不是充分替代。[CV001, CV002, CV003, CV004, CV005, CV010, CV016]

建议摘要
维度评估证据基础决策含义
建议继续研究技术验证已经存在,但当前商业证据不完整。在完成主要尽调前,不要基于已披露估值做买入判断。
置信度中融资估值和任务主张均有记录,而收入和条款仍属私有信息。将结论视为有条件判断,而不是量化的公允价值意见。
风险评级高发射时点、轨道假设、规模化经济性和商业化仍未解决。要求按里程碑融资,或提供同等的下行保护。
估值态度昂贵8 月的投后估值约为 3 月估值的 2.1 倍,但公司未披露当前收入基数。等待经营证明,或等到入场条款明显改善。

这些判断是基于公开融资估值和不完整经营披露作出的分析性判断;融资额不等同于可用现金。

[CV001, CV003, CV005, CV008, CV009, CV010, CV016]
FV001: 推荐逻辑

当前融资估值只跨过了融资这一步;技术、商业、经济性和条款证明仍是投资前提。

节点 / 连接节点 / 起点终点说明
节点 1 $2.3B 投后融资估值 [disclosed-price]观察到的 2026 年 8 月交易参考值,不是公开市场估值模型。
节点 2 演示证明 [technical-proof]Starcloud-1 搭载 H100,并在轨道运行据报的 AI 负载。
节点 3 商业卫星待交付 [schedule-proof]Starcloud-2 目前被描述为计划于 2027 年运行,尚未入轨。
节点 4 收入证明缺失 [commercial-proof]留存证据未披露当前收入、利用率、积压订单和客户经济性。
节点 5 经济性与条款缺失 [economics-and-terms]规模化成本、现金、烧钱速度、股权表、稀释和优先权输入仍需尽调。
节点 6 继续研究;估值昂贵 [recommendation]只有在里程碑得到证明,或风险调整后条款明显改善时才投资。
连接 1 $2.3B 投后融资估值 [disclosed-price]演示证明 [technical-proof]
连接 2 演示证明 [technical-proof]商业卫星待交付 [schedule-proof]
连接 3 商业卫星待交付 [schedule-proof]收入证明缺失 [commercial-proof]
连接 4 收入证明缺失 [commercial-proof]经济性与条款缺失 [economics-and-terms]
连接 5 经济性与条款缺失 [economics-and-terms]继续研究;估值昂贵 [recommendation]

该流程是定性的决策顺序,不代表数值转化率或概率。

[CV004, CV005, CV006, CV008, CV010, CV015, CV016]

8.2 论点、反论点与证明负担

可投资论点是 Starcloud 已超越纸面概念:Starcloud-1 搭载 NVIDIA H100,公司和独立报道都描述了在轨 AI 工作负载演示。反论点是,演示卫星不等于商业数据中心。Starcloud 现在称 Starcloud-2 将于 2027 年运行;一名独立跟踪者记录其较早的 2026 年 10 月目标已推迟,并称航天器尚未入轨。同一跟踪者认为,Starcloud-1 的实际轨道倾角没有检验经济叙事核心的近乎持续日照假设。Microsoft 管理层公开表达的谨慎,为公司的预测提供了有用的需求侧制衡。[CV006, CV007, CV008, CV009, CV011, CV012, CV013, CV014]

正论与反论
视角有证据支持的论点哪些变化会改变判断
正论Starcloud-1 已将一块 H100 送入轨道,公司材料称其已在太空运行 AI 负载。独立遥测若能确认持续性能和可靠性,将强化这一论点。
正论最近一轮融资引入了战略基础设施投资者,并提高了累计融资额。若能清楚说明募集资金如何转化为已签约商业容量,就能把融资与价值创造连起来。
反论Starcloud-2 尚未入轨,公开时间表已推迟至 2027 年。成功进入所需轨道并随后开展付费运营,将降低执行风险。
反论留存证据没有披露当前商业收入,也未验证规模化单位经济性。经审计的收入、利用率、毛利率和发射成本敏感性数据,才足以支持模型化估值。

这是一套有选择性的决策框架,而非穷尽所有风险的清单;每一行都说明哪些证据足以改变投资判断。

[CV001, CV002, CV006, CV007, CV008, CV009, CV010, CV012, CV015]

8.3 情景方法与估值边界

里程碑方法比收入倍数更合适,因为保留证据没有提供当前收入、利润率、签约积压、利用率或经审计的单位经济性。乐观情景要求 Starcloud-2 到达目标轨道、可靠运行付费工作负载,并证明发射、热管理、辐射和通信成本能够支撑有吸引力的毛利率。基准情景假设技术继续进步,但商业验证延迟,使当前估值缺乏公开运营模型支撑。悲观情景则叠加再次延期、付费利用率疲弱或持续日照前提失效。资料池无法支持概率加权的公允价值区间;相比虚假的精确度,定性信号和明确触发条件更诚实。[CV005, CV007, CV008, CV009, CV010, CV015, CV016]

乐观、基准与悲观情景
情景假设估值与回报逻辑关键风险概率信号
乐观Starcloud-2 进入所需轨道,可靠地运行付费负载,并证明可信的规模化经济性。只有在已签约收入和利润率证据足以建立模型后,溢价才可能站得住脚;目前没有公开区间可供合理辩护。热性能、发射成本、客户转化和后续融资。未量化;需要多个相互独立的里程碑得到确认。
基准技术进展延续,但商业证明和 Starcloud-3 的经济性仍然延迟披露,或仍属非公开信息。$2.3B 融资估值仍只是交易参考,而非已证明的公允价值;推迟入场。稀释、进度拖延、优先权压力,以及规模化前长期烧钱。与目前留存证据最一致。
悲观再次出现重大延误、部署到错误轨道、付费利用率疲弱,或成本假设落空。下行空间可能跌破 3 月 $1.1B 的融资参考值,但现有资料不足以支持精确的底价或回收价值。任务失败、资本稀缺、技术故障,以及战略买家转而自建。由触发条件决定,而非数值估算。

这些情景没有刻意按概率加权,因为当前收入、现金、烧钱速度、持股、优先权和可比交易数据都不可得。

[CV001, CV003, CV007, CV008, CV009, CV010, CV011, CV015, CV016]

8.4 退出准备度、尽调要求与论点失效触发因素

退出准备度低,因为证据基础显示的是融资和技术里程碑,而不是可重复的商业模型。在给投资定价前,尽调应取得 cap 股权结构表、清算优先权、债务和项目融资义务、当前现金、月度烧钱、客户合同、积压、工作负载利用率、毛利桥接、发射保险及卫星级遥测。主要论点失效触发因素包括 Starcloud-2 再次延期且没有可信补救计划、无法进入目标轨道、无法发布可靠的热性能和辐射表现,或商业使用仍停留在宣传而非付费。正面重估需要第三方对付费工作负载和经济性的证据,而不是又一份融资公告。[CV001, CV002, CV007, CV008, CV009, CV010, CV014, CV015, CV016]

8.5 图表

免责声明

本报告基于截至运行日期可获得的公开证据,不构成投资、法律、税务或技术建议。私营公司交易价值和公司主张在依赖前需要一手尽调。

证据索引

结论
编号陈述可信度来源
CO001 Starcloud is developing satellite-based data centers that combine GPU compute, persistent storage, and orbital data processing. 高 SO001, SO004, SO006
CO002 Starcloud is a Redmond, Washington-based company founded in 2024. 高 SO001, SO003, SO005
CO003 Philip Johnston is Starcloud's co-founder and CEO, with prior satellite-project work at McKinsey and graduate training spanning national security, business, mathematics, and finance. 高 SO001, SO002, SO008
CO004 Ezra Feilden is Starcloud's co-founder and CTO, with a decade of satellite-design experience focused on deployable solar arrays and structures. 中 SO001
CO005 Adi Oltean is Starcloud's co-founder and chief engineer, with prior SpaceX tracking-beam and Microsoft production GPU-cluster experience. 中 SO001
CO006 Starcloud announced a $170 million Series A on March 30, 2026 at a reported $1.1 billion valuation. 高 SO002, SO003, SO004, SO005
CO007 April 2026 reporting placed Starcloud's total capital raised at about or above $200 million after the Series A. 中 SO003, SO004, SO005
CO008 TechCrunch reported on August 21, 2026 that Starcloud added a $250 million extension that valued the company at $2.3 billion. 中 SO002
CO009 The August extension was led by Manhattan West Ventures and included NVIDIA, Cisco, Benchmark, EQT, Soma, NFX, 776, Cedar Capital, Goanna Capital, and Standard Capital. 中 SO002
CO010 TechCrunch reported that Starcloud had 25 employees in August 2026. 中 SO002
CO011 TechCrunch reported that Starcloud was developing production lines at a 100,000-square-foot facility in Woodinville, Washington. 中 SO002
CO012 Starcloud-1 launched in November 2025 carrying an NVIDIA H100 GPU. 高 SO003, SO004, SO007
CO013 Starcloud says Starcloud-1 ran a version of Gemini and trained a nanoGPT model in orbit in December 2025. 高 SO003, SO007
CO014 Starcloud describes Starcloud-2 as its first commercial mission, offering a GPU cluster, persistent storage, continuous access, and an operational target in 2027. 高 SO002, SO004, SO006
CO015 Starcloud has requested FCC permission to operate up to 88,000 spacecraft, leaving authorization and launch capacity as material dependencies for its constellation-scale plan. 高 SO002, SO005
CO016 Tech Insider reported that Starcloud had no disclosed commercial revenue as of April 2026. 中 SO003
CM001 Starcloud says Starcloud-1 launched in November 2025 carrying the first Nvidia H100 GPU in space. 中 SM003
CM002 Observer reported that Starcloud-1 was a 60-kilogram spacecraft carrying five Nvidia GPUs and orbiting Earth every 90 minutes. 中 SM001
CM003 Starcloud says Starcloud-2 will provide a GPU cluster, persistent storage, 24/7 access, and proprietary thermal and power systems in sun-synchronous orbit by 2027. 中 SM006
CM004 Starcloud positions in-orbit analysis of spacecraft data as a way to reduce raw-data downlink and deliver lower-latency insights. 中 SM006
CM005 Starcloud positions secure global storage and sovereign cloud computing independent of Earth as terrestrial-user offerings. 中 SM006
CM006 Observer reported that Starcloud was already running orbital edge-computing workloads for Earth-observation and military satellites, according to the company's CEO. 中 SM001
CM007 Bitcoin-mining hardware on Starcloud-2 is described as a revenue bridge for periods of low AI utilization rather than evidence of a mature cloud market. 中 SM004, SM005
CM008 Observer reported that cooling consumed roughly 70% of Starcloud's engineering attention and that the production cooling system had not yet been tested in space. 中 SM001
CM009 Observer reported that Starcloud's planned cost-competitive 200-kilowatt spacecraft depends on SpaceX's still-developing Starship. 中 SM001
CM010 The Silicon Review reported that constrained launch supply is becoming a major cost and scheduling issue for orbital data-center developers. 中 SM002
CM011 The Silicon Review reported that Starcloud's roadmap depends on Starship becoming a routine commercial launch vehicle. 中 SM002
CM012 Orbital Nodes reported that Starcloud-2 was not yet in orbit and was then expected to launch in 2027. 中 SM004
CM013 Orbital Nodes estimated that Starcloud-1's 45.4-degree rideshare orbit put it in Earth's shadow for roughly 38% of each revolution, leaving the continuous-solar premise untested. 中 SM004
CM014 Tech Insider reported a 2035 orbital-data-center forecast of $39.09 billion from $1.77 billion in 2029 at a 67.4% compound annual growth rate, attributing it to Introl. 低 SM005
CM015 The assigned evidence provides no primary market model, geography, pricing base, or contracted-capacity data sufficient to calculate Starcloud's SAM or SOM. 中 SM005, SM006, SM007, SM008
CM016 The fetched Starcloud-4 page provides no power, capacity, launch-date, or pricing specifications in its accessible text. 中 SM008
CP001 Reuters reported in March 2026 that SpaceX had revealed plans for a million-satellite orbital data-center network and that Blue Origin had expressed similar ambitions. 中 SP001
CP002 Reuters reported that Starcloud was working with Nvidia and the cloud units of Amazon and Google. 中 SP001
CP003 Starcloud management told Reuters that high launch costs remained a challenge and expected cost competitiveness with terrestrial facilities around 2028 or 2029. 中 SP001
CP004 Sacra identifies CoreWeave, Crusoe, and Lambda as terrestrial GPU-cloud substitutes competing for the same AI-infrastructure spend as Starcloud. 中 SP002
CP005 Sacra characterizes Blue Origin as a vertically integrated alternative able to bundle communications, power, storage, and edge computing. 中 SP002
CP006 Starcloud says Starcloud-2 will combine a GPU cluster, persistent storage, continuous access, and proprietary thermal and power systems. 中 SP004
CP007 Starcloud says its November 2025 Starcloud-1 mission carried an Nvidia H100 and later ran Gemini and trained a nanoGPT model in orbit. 中 SP003
CP008 As reviewed on the access date, Starcloud's Starcloud-4 page disclosed no compute specification, schedule, price, or packaging detail. 中 SP005
CP009 Reuters reported that Starcloud raised $170 million at a $1.1 billion valuation, bringing total funding to $200 million. 中 SP001
CP010 Data Center Dynamics reported that Starcloud's disclosed seed funding reached $21 million in 2025. 中 SP007
CP011 Data Center Dynamics described Starcloud's earlier assumptions for launching a 40 MW orbital data center for $8.2 million as potentially optimistic. 中 SP007
CP012 Tech Insider identifies Lonestar's lunar data-vault strategy and Axiom's station-based compute activity as direct or adjacent orbital alternatives. 低 SP008
CP013 Tech Insider reports that SpaceX filed to deploy as many as one million orbital-data-center satellites. 低 SP008
CP014 CIO Business Views identifies launch economics, infrastructure scalability, cybersecurity, and operational reliability as unresolved determinants of orbital-data-center success. 低 SP006
CP015 None of the three retained Starcloud product pages reviewed disclosed list pricing, a contract unit, discount terms, or a service-level commitment. 中 SP003, SP004, SP005
CP016 Starcloud's current moat appears more dependent on execution and partner integration than exclusive distribution because capable cloud and space-platform partners retain internal-build and bundling options. 中 SP001, SP002, SP003, SP004
CI001 Starcloud announced a $250 million Series A extension at a $2.3 billion post-money valuation in August 2026. 中 SI001
CI002 Starcloud said the August 2026 financing brought total capital raised since its 2024 founding to $450 million. 中 SI001
CI003 Starcloud said the new capital would fund manufacturing capacity, engineering work with NVIDIA, and future launch allocation. 中 SI001
CI004 TechCrunch reported a $170 million Series A at a $1.1 billion valuation in March 2026. 中 SI002
CI005 TechCrunch reported that Starcloud had raised $200 million in total as of March 2026. 中 SI002
CI006 Management's roughly $0.05-per-kilowatt-hour target for Starcloud-3 assumes commercial launch costs near $500 per kilogram. 中 SI002, SI003
CI007 Starcloud's CEO told TechCrunch the company would not be competitive on energy costs until Starship was flying frequently. 中 SI002
CI008 Management described a near-term business selling processing to spacecraft and a later model serving terrestrial data-center workloads after launch costs decline. 中 SI002
CI009 Tech Insider reported that Starcloud had no disclosed commercial revenue as of April 2026. 低 SI003
CI010 Starcloud presents Starcloud-2 as its first commercial mission and says it will provide a GPU cluster, persistent storage, and continuous access. 中 SI004
CI011 Starcloud says Starcloud-2 will process high-volume spacecraft data in orbit to reduce latency and downlink requirements. 中 SI004
CI012 Starcloud positions Starcloud-2 for secure storage, critical backup, and premium sovereign cloud computing for terrestrial users. 中 SI004
CI013 Starcloud says Starcloud-1 launched in November 2025 carrying an NVIDIA H100 GPU and subsequently ran a version of Gemini and trained a nano-GPT model in orbit. 中 SI005
CI014 The reviewed official mission and marketing pages solicit direct contact but disclose no list price, billing unit, discount, or realized contract price. 中 SI004, SI005, SI006, SI007
CI015 The August financing announcement discloses planned uses of funds but not cash on hand, burn, debt, launch commitments, or runway. 中 SI001
CI016 TechCrunch characterized Starcloud's model as dependent on unproven technology and significant capital expenditure. 中 SI002
CE001 Starcloud says Starcloud-1 launched in November 2025 carrying the first NVIDIA H100 GPU in space. 中 SE005
CE002 Starcloud says Starcloud-1 ran a version of Gemini and trained the nanoGPT model in orbit in December 2025. 中 SE005
CE003 Starcloud describes Starcloud-2 as its first commercial mission with a GPU cluster, persistent storage, continuous access, and proprietary thermal and power systems, targeted to be operational in 2027. 中 SE004
CE004 Starcloud says Starcloud-2 will process high-volume spacecraft data in orbit so users can downlink derived insights instead of massive raw datasets. 中 SE004
CE005 Starcloud markets Starcloud-2 to terrestrial users for secure global storage and sovereign cloud computing independent of Earth. 中 SE004
CE006 An official Starcloud promotional page described a near-term launch as the world's largest space data center and a larger follow-on as profitable without publishing supporting performance or economic data. 中 SE006
CE007 Starcloud's Y Combinator profile says the company established a payload manufacturing facility in Redmond, Washington. 中 SE001
CE008 Starcloud's Y Combinator profile reports high-value letters of intent for H100 compute time rather than disclosed production contracts. 中 SE001
CE009 Starcloud's founder describes radiation mitigation as a combination of shielding and software-level fault tolerance. 中 SE008
CE010 Starcloud's founder describes thermal management based on conductive-liquid immersion of the board and large lightweight deployable radiators. 中 SE008
CE011 Starcloud's founder claims a laser terminal can connect to satellite backbones at up to hundreds of gigabits per second with latency as low as 50 milliseconds. 低 SE008
CE012 Starcloud's founder says the current platform is not modular and that modular architecture becomes necessary only beyond a roughly 40-megawatt single-satellite scale. 中 SE008
CE013 Blockchain Council reports that the planned Starcloud-2 configuration includes H100 and Blackwell hardware, AWS Outposts capabilities, and laser links targeting other constellations. 低 SE002
CE014 Blockchain Council characterizes launch economics as the most consequential constraint on Starcloud's path to hyperscale deployment. 中 SE002
CE015 A Quora practitioner contribution says orbital-compute economics remain unproven at scale because launch expense, radiation, debris, and hardware upgrade constraints offset potential energy and location advantages. 低 SE003
CE016 The fetched Starcloud-4 page provides no configuration, timing, workload, or qualification specifications for that named product generation. 中 SE007
CU001 Starcloud describes Starcloud-2 as its first commercial mission with a GPU cluster, persistent storage, continuous access, and proprietary thermal and power systems. 中 SU006
CU002 Starcloud targets spacecraft and space-station operators that need to process high-volume raw data in orbit before downlink. 中 SU006
CU003 Starcloud targets terrestrial users seeking sovereign cloud compute, resilient storage, and backup independent of terrestrial infrastructure. 中 SU006
CU004 Starcloud says Starcloud-2 will be fully operational in sun-synchronous orbit by 2027. 低 SU006
CU005 Starcloud-1 launched in November 2025 carrying an NVIDIA H100 GPU. 中 SU002, SU003, SU005
CU006 Starcloud-1 was reported to have trained NanoGPT in orbit after launch. 中 SU003, SU005
CU007 NVIDIA provided Starcloud technical support, expert access, and GPUs through the NVIDIA Inception relationship. 中 SU002
CU008 GeekWire reported that NVIDIA invested in Starcloud and expanded its collaboration around future space-rated chips. 中 SU005
CU009 Tech Insider reported that Starcloud had no disclosed commercial revenue as of April 2026. 低 SU004
CU010 Tech Insider reported that AWS, Google Cloud, NVIDIA, and Crusoe planned hardware or software participation on Starcloud-2. 低 SU004
CU011 AInvest characterized the planned October 2026 launch as the first major integrated-platform test rather than as proven production adoption. 低 SU001
CU012 Starcloud's public Starcloud-1 page invites prospective suppliers and customers to contact the company. 中 SU003, SU008
CU013 The reviewed official product and marketing pages do not identify a paying customer, contract value, price, or production customer outcome. 中 SU003, SU006, SU007, SU008
CU014 None of the assigned retained sources discloses retention, renewal, churn, cohort, backlog, or customer-concentration metrics. 中 SU001, SU002, SU003, SU004, SU005, SU006, SU007, SU008
CU015 Tech Insider identified a Starcloud-2 launch or on-orbit failure as the company's largest near-term risk. 低 SU004
CR001 Control Plane reported in March 2026 that Falcon 9 delivery cost roughly $3,600 per kilogram while Starcloud-3 economics require launch costs around $500 per kilogram. 中 SR001
CR002 Control Plane estimated that rejecting one megawatt of waste heat in orbit requires approximately 1,200 square meters of radiator surface. 中 SR001
CR003 CNBC reported that Morgan Stanley analysts identified radiation, in-orbit maintenance, debris, data-governance, and space-traffic hurdles for orbital data centers. 中 SR002
CR004 Starcloud-1 launched in November 2025 carrying an NVIDIA H100 GPU. 高 SR002, SR003
CR005 Starcloud-1 subsequently ran Google Gemma inference and trained NanoGPT in orbit. 高 SR002, SR003
CR006 Data Center Frontier described Starcloud-1 as a test of whether a high-power GPU can operate through vacuum, radiation, and repeated thermal cycling. 中 SR005
CR007 Starcloud claims Starcloud-2 will provide a GPU cluster, persistent storage, 24/7 access, and proprietary thermal and power systems in sun-synchronous orbit by 2027. 中 SR006
CR008 Wikipedia and GeekWire reported that Starcloud filed with the FCC to operate as many as 88,000 orbital data-center satellites. 中 SR004, SR010
CR009 Wikipedia reports that a legal challenge from Lumen Technologies preceded the company's 2025 rebrand from Lumen Orbit to Starcloud. 低 SR004
CR010 Three April-or-earlier 2026 sources reported a $170 million Series A at a $1.1 billion valuation, with total funding then reported at about $200 million. 中 SR007, SR008, SR009
CR011 GeekWire reported on August 21, 2026 that Starcloud raised $250 million, bringing total capital raised to $450 million at a $2.3 billion post-money valuation. 中 SR004, SR010
CR012 The modeled path to cost-competitive large-scale orbital compute depends on frequent, low-cost Starship launches that were not yet commercially available in the retained evidence. 中 SR001, SR007, SR009
CR013 Reported 2026 financing is designated for manufacturing, engineering, and launch procurement, but the retained sources do not disclose cash available, burn, committed obligations, or runway. 中 SR008, SR010
CR014 Starcloud describes Starcloud-2 as offering sovereign, highly redundant storage, but the retained company page provides no independent security controls, certifications, or incident-response evidence. 中 SR006
CR015 The retained mission sources disclose a successful demonstration and a planned commercial system but do not disclose sustained uptime, component error rates, thermal margins, qualification results, or recovery performance. 中 SR002, SR003, SR005, SR006
CR016 Launch economics, thermal scaling, orbital reliability, provider dependence, regulatory scope, and financial runway are the highest-residual-risk clusters supported by the retained evidence. 中 SR001, SR002, SR007, SR008, SR010
CV001 Starcloud announced a $250 million Series A extension at a $2.3 billion post-money valuation on August 21, 2026. 中 SV001
CV002 Starcloud said the August extension brought total capital raised since 2024 to $450 million. 中 SV001
CV003 GeekWire reported a $170 million financing at a $1.1 billion valuation in March 2026. 中 SV002, SV003
CV004 GeekWire reported that Starcloud had raised $200 million in total after the March 2026 round. 中 SV002
CV005 The disclosed valuation increased by approximately 2.1 times from the March $1.1 billion mark to the August $2.3 billion post-money mark. 中 SV001, SV002
CV006 GeekWire reported that Starcloud-1 processed AI workloads in orbit after its November 2025 launch. 中 SV002
CV007 Starcloud says Starcloud-2 will provide a GPU cluster, persistent storage, and continuous access in sun-synchronous orbit by 2027. 中 SV005
CV008 Orbital Nodes reports that Starcloud-2 is not yet in orbit and that its launch target moved from October 2026 to rideshare flights in 2027. 中 SV004
CV009 Orbital Nodes reports that Starcloud-1 flew at 45.4 degrees rather than the near-97-degree dawn-dusk orbit required to test the continuous-solar premise. 中 SV004
CV010 Tech Insider reported that Starcloud had disclosed no commercial revenue as of April 2026. 低 SV003
CV011 GeekWire reported Microsoft President Brad Smith saying that Microsoft was not betting on the satellite data-center strategy. 中 SV002
CV012 Starcloud says Starcloud-1 carried the first NVIDIA H100 GPU into orbit. 中 SV006
CV013 Starcloud's Starcloud-4 page promotes a future involving gigawatts of compute deployed in space. 中 SV008
CV014 An official Starcloud tour description characterized a larger follow-on orbital data center as profitable without publishing the supporting economics. 低 SV007
CV015 Tech Insider reports that Starcloud's $0.05-per-kilowatt-hour target depends on a $500-per-kilogram Starship launch assumption that had not been demonstrated. 低 SV003
CV016 Given the rapid valuation step-up and missing current revenue, ownership, preference, and scaled-economics inputs, research-more is more supportable than a buy recommendation at the disclosed mark. 中 SV001, SV003, SV004
来源
编号出版方标题引文
SO001 Starcloud Data Centers in Space | Starcloud - The Future of AI We are a team of innovative deep-tech engineers with big ambitions to change the paradigm of hyperscale data centers, enabling the future of AI by deploying the largest training clusters on data centers in space.
SO002 TechCrunch Starcloud raises $250 million for orbital data centers as launch options dry up “Obviously if we can’t book any SpaceX launch capacity in 2029, that will be challenging for us,” Johnston said.
SO003 Tech Insider Starcloud's $170M Series A: The $1.1B Space Data Center Bet [2026] A $1.1 billion valuation for a company that has launched exactly one 130-pound satellite is, on any traditional metric, aggressive. Starcloud reported no disclosed commercial revenue as of April 2026.
SO004 Futurum Will Starcloud's Orbital Data Centers Solve NVIDIA's Terrestrial Energy Crisis? The funding brings the company’s total capital raised to $200 million and will be used to establish a dedicated manufacturing facility, expand headcount, and procure future launch contracts.
SO005 Via Satellite Starcloud Raises $170M to Fund Orbital Data Center Plans Starcloud, a space startup working to deploy satellites to serve as orbital data centers, raised $170 million in a Series A round announced Monday that values the company at $1.1 billion.
SO006 Starcloud Starcloud-2 | In-Space GPU Cluster & Cloud Computing Satellite Our first commercial mission, Starcloud-2, features a GPU cluster, persistent storage, 24/7 access, and proprietary thermal and power systems in a smallsat form factor. It will be fully operational in sun-synchronous orbit by 2027.
SO007 Starcloud Starcloud-1 Starсloud-1 launched in November 2025 with the first Nvidia H100 GPU in space - vastly more AI compute than has been in space before.
SO008 Starcloud The Birth of the Space Internet - Starcloud Launches AI into the Vacuum Touring Starcloud HQ in Redmond, WA with Founder and CEO Philip Johnston. Starcloud is building AI datacenters in space.
SM001 Observer Starcloud CEO Philip Johnston on Putting the First A.I. Data Center in Space Starcloud is already running edge computing workloads in space for Earth-observation and military satellites: Running inference—say, real-time ChatGPT-like requests that don’t need as much computing power or data transfer as training LLMs—in orbit.
SM002 The Silicon Review Starcloud Raises $250 Million for Orbital Data Centers as Launch Options Dry Up, Nvidia and Cisco Join $2.3 Billion Valuation Round The funding comes as launch options become a mounting concern for orbital data center startups.
SM003 Starcloud Starcloud-1 Starcloud-1 is a groundbreaking satellite that carries the first NVIDIA H100 GPU into orbit
SM004 Orbital Nodes Starcloud-2 — Bitcoin Mining in Space, Launch Tracker Starcloud-1 established that a space data center in miniature survives launch and runs in orbit: an H100 executing real workloads, including training a language model in space. It did not establish the energy model (wrong orbit), the disposal model (its eleven-month mission ended around 2 October 2026, and on 9 October it was still in a 499 km orbit with no sign of a deorbit burn and no passive backstop), or the economics.
SM005 Tech Insider Starcloud's $170M Series A: The $1.1B Space Data Center Bet [2026] Infrastructure research firm Introl offered a more granular midpoint in a February 2026 forecast, projecting the orbital data center market to grow from $1.77 billion in 2029 to $39.09 billion by 2035, a 67.4% compound annual growth rate.
SM006 Starcloud Starcloud-2 | In-Space GPU Cluster & Cloud Computing Satellite Our first commercial mission, Starcloud-2, features a GPU cluster, persistent storage, 24/7 access, and proprietary thermal and power systems in a smallsat form factor.
SM007 Starcloud The Birth of the Space Internet – Starcloud Launches AI into the Vacuum Starcloud is building AI datacenters in space.
SM008 Starcloud Starcloud-4 | Connect with the Future of Space Data Centers Build the future of data centers in space with us
SP001 Reuters Starcloud Reaches $1.1 Billion Valuation as AI Space Race Heats Up In February, Elon Musk's SpaceX acquired his AI startup xAI and revealed plans for a million-satellite orbital data center network.
SP002 Sacra Starcloud funding, news & analysis The largest commercial substitute for Starcloud in 2026 remains terrestrial GPU cloud.
SP003 Starcloud Starcloud-1 Starсloud-1 launched in November 2025 with the first Nvidia H100 GPU in space - vastly more AI compute than has been in space before.
SP004 Starcloud Starcloud-2 | In-Space GPU Cluster & Cloud Computing Satellite Our first commercial mission, Starcloud-2, features a GPU cluster, persistent storage, 24/7 access, and proprietary thermal and power systems in a smallsat form factor.
SP005 Starcloud Starcloud-4 | Connect with the Future of Space Data Centers The Future of Data Centers is Space
SP006 CIO Business Views Starcloud Expands the Future of Space-Based AI Infrastructure The success of this model will depend on launch economics, infrastructure scalability, cybersecurity, and operational reliability.
SP007 Data Center Dynamics Lumen Orbit rebrands to Starcloud, raises another $10m for in-orbit data centers As previously reported by DCD, Starcloud’s proposals make a number of potentially optimistic assumptions on the costs to build and launch large-scale modules into orbit. It claims it will be able to launch a 40MW data center into orbit for just $8.2m.
SP008 Tech Insider Starcloud's $170M Series A: The $1.1B Space Data Center Bet [2026] Starcloud is not alone. The orbital compute category has attracted a small but serious cohort of competitors, each approaching the problem from a different angle.
SI001 Starcloud via Business Wire Starcloud Raises $250 Million at $2.3 Billion Valuation to Scale AI with Orbital Data Centers Starcloud, the company pioneering data centers in space, today announced a $250 million Series A extension funding round at a $2.3 billion post-money valuation.
SI002 TechCrunch Starcloud raises $170 million Series A to build data centers in space It’s another sign of the interest in outsourcing data centers to orbit as resource and political obstacles slow their development on Earth, but the business model depends on unproven technology and significant capital expenditure.
SI003 Tech Insider Starcloud's $170M Series A: The $1.1B Space Data Center Bet [2026] A $1.1 billion valuation for a company that has launched exactly one 130-pound satellite is, on any traditional metric, aggressive.
SI004 Starcloud Starcloud-2 | In-Space GPU Cluster & Cloud Computing Satellite Our first commercial mission, Starcloud-2, features a GPU cluster, persistent storage, 24/7 access, and proprietary thermal and power systems in a smallsat form factor.
SI005 Starcloud Starcloud-1 Starсloud-1 launched in November 2025 with the first Nvidia H100 GPU in space - vastly more AI compute than has been in space before.
SI006 Starcloud The Birth of the Space Internet – Starcloud Launches AI into the Vacuum Starcloud is building AI datacenters in space.
SI007 Starcloud Starcloud-4 | Connect with the Future of Space Data Centers The Future of Data Centers is Space
SI008 CNN Why this billion-dollar tech company is sending data centers to space With AI straining Earth’s infrastructure, space-tech startup Starcloud is developing orbital data centers designed to meet the world’s growing computing needs.
SE001 Y Combinator Starcloud: Data centers in space Set up our payload manufacturing facility in Redmond, WA
SE002 Blockchain Council Starcloud: Data Centre Infrastructure in Space The most consequential constraint is launch economics.
SE003 Quora How do startup companies like Starcloud that build data centers in space make money? The economic viability of these models remains unproven at scale.
SE004 Starcloud In-Space GPU Cluster & Cloud Computing Satellite - Starcloud-2 Our first commercial mission, Starcloud-2, features a GPU cluster, persistent storage, 24/7 access, and proprietary thermal and power systems in a smallsat form factor.
SE005 Starcloud Starcloud-1 Starcloud-1 is a groundbreaking satellite that carries the first NVIDIA H100 GPU into orbit
SE006 Starcloud The Birth of the Space Internet - Starcloud Launches AI into the Vacuum Starcloud is launching the world's largest datacenter in space in just a few weeks, and will launch an even bigger one (that profitable) next year.
SE007 Starcloud Connect with the Future of Space Data Centers - Starcloud-4 Learn why gigawatts of compute will be deployed in space in the near future, and what architectures they will use.
SE008 Future of Computing Starcloud: Shaping the Future of Space-Based Data Centers Our approach to solving this combines shielding with software-level adaptations to build fault tolerance, enabling reliable operation in harsh environments.
SU001 AInvest Starcloud’s October 2026 Launch Could Validate the Orbital Data Center S-Curve—And Spark a New Infrastructure Race The company has already demonstrated the ability to run an AI model from space, but its next launch in October 2026 is the first major test of the integrated platform.
SU002 NVIDIA How Starcloud Is Bringing Data Centers to Outer Space The 60-kilogram Starcloud-1 satellite, about the size of a small fridge, is expected to offer 100x more powerful GPU compute than any previous space-based operation.
SU003 Starcloud Starcloud-1 Starcloud-1 is a groundbreaking satellite that carries the first NVIDIA H100 GPU into orbit
SU004 Tech Insider Starcloud's $170M Series A: The $1.1B Space Data Center Bet [2026] Starcloud reported no disclosed commercial revenue as of April 2026.
SU005 GeekWire Starcloud raises $250M to support the creation of data center satellite network in league with Nvidia Nvidia’s participation in the funding round brings Starcloud’s collaboration with the computer-chip titan to a new level.
SU006 Starcloud In-Space GPU Cluster & Cloud Computing Satellite - Starcloud-2 Our first commercial mission, Starcloud-2, features a GPU cluster, persistent storage, 24/7 access, and proprietary thermal and power systems in a smallsat form factor.
SU007 Starcloud The Birth of the Space Internet – Starcloud Launches AI into the Vacuum Touring Starcloud HQ in Redmond, WA with Founder and CEO Philip Johnston.
SU008 Starcloud Connect with the Future of Space Data Centers - Starcloud-4 Interested in working with us as a supplier or customer?
SR001 Control Plane Starcloud Reached Orbit. Now It Has a Harder Problem. If Starship is delayed, Starcloud will keep launching smaller versions on Falcon 9. Full cost-competitiveness is a 2030s story, contingent on a rocket that is not yet flying commercially and manufactured by a company that has publicly prioritized its own compute needs.
SR002 CNBC ‘Greetings, earthlings’: Nvidia-backed Starcloud trains first AI model in space as orbital data center race heats up Risks in operating orbital data centers remain, however. Analysts from Morgan Stanley have noted that orbital data centers could face hurdles such as harsh radiation, difficulty of in-orbit maintenance, debris hazards, and regulatory issues tied to data governance and space traffic.
SR003 Starcloud Starcloud-1 Starcloud-1 is a groundbreaking satellite that carries the first NVIDIA H100 GPU into orbit
SR004 Wikipedia Starcloud On February 3, 2026, Starcloud submitted a proposal to the FCC for a constellation of up to 88,000 satellites for orbital data centers.
SR005 Data Center Frontier Starcloud Launches Orbital AI Data Center With NVIDIA H100 GPU The mission’s primary objective is to determine whether a high-power GPU can operate reliably in the harsh conditions of space, including vacuum exposure, radiation, and repeated thermal cycling.
SR006 Starcloud Starcloud-2 | In-Space GPU Cluster & Cloud Computing Satellite Our first commercial mission, Starcloud-2, features a GPU cluster, persistent storage, 24/7 access, and proprietary thermal and power systems in a smallsat form factor. It will be fully operational in sun-synchronous orbit by 2027.
SR007 Tech Insider Starcloud's $170M Series A: The $1.1B Space Data Center Bet [2026] The near-term risk is a Starcloud-2 launch or on-orbit failure. If the satellite does not reach orbit, or if it reaches orbit but the cooling system cannot handle the 100-kilowatt heat load at design capacity, the Series B story becomes much harder.
SR008 Via Satellite Starcloud Raises $170M to Fund Orbital Data Center Plans Starcloud, a space startup working to deploy satellites to serve as orbital data centers, raised $170 million in a Series A round announced Monday that values the company at $1.1 billion.
SR009 Futurum Will Starcloud’s Orbital Data Centers Solve NVIDIA’s Terrestrial Energy Crisis? The launch cadence and payload capacity of SpaceX’s Starship which is the critical catalyst required to make the economics of orbital data centers viable at gigawatt scale.
SR010 GeekWire Starcloud raises $250M to support the creation of data center satellite network in league with Nvidia Starcloud has filed an application with the Federal Communications Commission to operate as many as 88,000 satellites as orbital data centers for AI and other applications.
SV001 Starcloud via Business Wire Starcloud Raises $250 Million at $2.3 Billion Valuation to Scale AI with Orbital Data Centers Starcloud... announced a $250 million Series A extension funding round at a $2.3 billion post-money valuation... The funding brings Starcloud's total capital raised to $450 million since its founding in 2024.
SV002 GeekWire Orbital AI: Seattle-area startup Starcloud hits $1.1B valuation to build space-based data centers Starcloud... announced $170 million in new funding Monday, vaulting it to unicorn status with a $1.1 billion valuation... The startup has raised a total of $200 million.
SV003 Tech Insider Starcloud's $170M Series A: The $1.1B Space Data Center Bet [2026] Starcloud reported no disclosed commercial revenue as of April 2026. The valuation is entirely forward-looking: a bet on the combined trajectories of SpaceX launch costs, Nvidia GPU efficiency, and the AI compute demand curve.
SV004 Orbital Nodes Starcloud-2 — Bitcoin Mining in Space, Launch Tracker Not yet in orbit... Launch2027, on rideshare flights — two Starcloud-2 satellites (company, August 2026; the earlier target was October 2026). No date or rocket named publicly
SV005 Starcloud Starcloud-2 | In-Space GPU Cluster & Cloud Computing Satellite Our first commercial mission, Starcloud-2, features a GPU cluster, persistent storage, 24/7 access, and proprietary thermal and power systems in a smallsat form factor. It will be fully operational in sun-synchronous orbit by 2027.
SV006 Starcloud Starcloud-1 Starcloud-1 is a groundbreaking satellite that carries the first NVIDIA H100 GPU into orbit
SV007 Starcloud The Birth of the Space Internet – Starcloud Launches AI into the Vacuum Starcloud is launching the world's largest datacenter in space in just a few weeks, and will launch an even bigger one (that profitable) next year.
SV008 Starcloud Starcloud-4 | Connect with the Future of Space Data Centers Learn why gigawatts of compute will be deployed in space in the near future, and what architectures they will use.