初创公司尽调
尽调报告 Climate / ocean wave energy + offshore AI data centers private, Series B 2026-08-10

Panthalassa

Thiel 背书的海洋算力登月项目,投资团星光很强,但尚未产生收入,技术也未经过规模化验证

Panthalassa 同时具备 Thiel 领投的明星财团和真正新颖的海上计算概念;但公司仍未产生收入,技术尚未规模化验证,也没有可定价披露,因此更像一笔高风险的风险期权。

封面要素

Series B 轮融资额 01
140 USD M [CO012]
累计融资额 02
210 USD M [CO015]
隐含估值(FT 表述) 03
1000 USD M [CO016]
当前员工数 04
120 employees [CO010]
成立时间 05
2016 year [CO004]

公司概况

Panthalassa 是一家位于俄勒冈州波特兰的公益公司,成立于 2016 年。公司制造自主漂浮节点,把海浪能转成电力,在海上运行 AI 数据中心。2026 年 5 月,公司完成由 Peter Thiel 领投的 $140M Series B 轮融资,投资团可信,概念也有差异化;但公司仍未产生收入,没有披露客户,商业规模海上运营也尚未验证。

官网
panthalassa.energy
成立时间
2016-01-01
创始人
Garth Sheldon-Coulson, Brian Moffat
创立地点
Portland, Oregon
总部
Portland, Oregon
产品
自主、自推进的漂浮节点,借助海浪驱动的涡轮在板上发电,为密封、海水冷却的 AI 算力供电,并通过低地球轨道卫星回传数据。
客户
需要批处理 / 推理算力容量的超大规模云厂商、新云服务商和 AI 实验室。
商业模式
销售在海上生产的 AI 算力容量;电力不输送上岸。
阶段
private, Series B
融资情况
私募融资;公开来源支持其截至 2026 年 5 月由 Peter Thiel 领投的 $140M Series B 轮已累计融资约 $210M,投资团包括 John Doerr、TIME Ventures、SciFi Ventures 以及回投的气候投资者。
[CO001, CO002, CO004, CO006, CO008, CO012, CO013, CO015]

执行摘要

主要优势

  • $140M Series B 由 Peter Thiel 领投,John Doerr、TIME Ventures、SciFi Ventures 和老气候投资人跟投,资本厚度和可信度光环都很强。
  • 纵向一体化节点把波浪发电和海上计算装进一个自主、自推进船体,概念确实差异化。
  • AI 计算电力需求形成强顺风;陆地和电网约束越紧,市场越会寻找新型选址。

主要风险

  • 公司仍处于收入前阶段,没有披露付费客户;商业规模下的海洋生存能力(腐蚀、生物污损、风暴)尚未验证。
  • 投后估值、收入、利润率和董事会构成均未披露,公开证据无法支撑入场价格纪律。
  • 执行集中在两位联合创始人身上;Starlink 延迟和波浪 LCOE 经济性把近期可行场景限制在批处理工作负载。

未决问题

  • Series B 准确投后估值、每股价格,以及优先权 / 稀释条款。
  • 已签客户、意向书或包销协议,以及商业管线。
  • Ocean-3 试点结果,以及商业规模的生存能力、正常运行时间和单位经济性数据。
  • 当前董事会构成、治理安排和控制权。

目录

Chapter 01

01公司概览

1.1 身份、产品模式和运营足迹

Panthalassa 将自己定义为一家位于俄勒冈州波特兰、正在打造全新能源与算力基础设施的公司:自主漂浮平台,公司称为节点,捕获海洋波浪能并转成电力,直接在海上运行人工智能数据中心。公司并不把电送上岸,而是主张把发电和算力放在同一个海上结构上,销售 AI 算力容量,从而绕开电网并网排队和土地约束。多家 2026 年来源显示,公司总部在波特兰,原型和海试活动分布在太平洋西北地区,包括 Juan de Fuca 海峡和 Puget Sound。公司以公益公司形式组织;2026 年 5 月融资被主流科技、气候和商业媒体广泛报道,说明它虽处早期,却已有较高公开能见度。落到实践层面,每个节点设计为可在海上自主运行,海浪驱动发电系统和算力载荷装在同一结构上,因此不需要海底电缆或电网并网。该架构就是公司叙事的核心:把算力放到海上能源丰富的地方,而不是在岸上争夺稀缺土地、电网容量和并网审批。[CO001, CO002, CO003, CO005, CO032, CO034]

FO002: 公司快照逻辑

Panthalassa 的逻辑从波浪能捕获延伸到板载供电、AI 算力和可销售服务;2026 年深厚财团提供支持,但披露缺口仍构成约束。

[CO001, CO005, CO012, CO017, CO029, CO038]

1.2 创始人、领导层和关键人物依赖

保留来源对创始信息的表述一致。Panthalassa 成立于 2016 年,由联合创始人兼 CEO Garth Sheldon-Coulson 领导;他曾任 Bridgewater Associates 高级投资助理和 AI 研究员。另一位联合创始人兼首席创新官 Brian Moffat 是海洋能源研究员,曾在 Spindrift Energy 研究波浪能,并拥有 UC Irvine 三个理学学士学位。两人的组合支撑创始人与市场的匹配:一端是资本市场与 AI 能力,另一端是海洋波浪能工程。公开披露突出这种工程导向的领导层,而不是庞大的具名高管团队或已披露董事会名单;这使治理能见度偏低,也把执行风险集中在两名关键人物身上。考虑到从原型到商业船队要走十年级别、资本密集的路径,关键人物集中是重要尽调事项。[CO004, CO006, CO007, CO008, CO009, CO039]

领导层与创始人表
人物职务背景置信度
Garth Sheldon-Coulson联合创始人兼 CEO前 Bridgewater 高级投资助理和 AI 研究员。
Brian Moffat联合创始人兼首席创新官海洋能源研究员;曾任职 Spindrift Energy;拥有 UC Irvine 三个理学学士学位。
更广泛高管梯队未公开列举工程主导团队;本次留存来源未披露完整高管名单。

创始人归属一致;更广泛领导层和董事会披露较薄。

[CO006, CO007, CO008, CO009, CO039]

1.3 融资、估值和投资者基础

Panthalassa 于 2026 年 5 月 4 日宣布完成 $140M Series B 轮融资,由 Peter Thiel 领投,已披露累计融资约 $210M。Financial Times 将公司称为「估值 $1bn 的海洋数据中心初创公司」,暗示接近独角兽估值,但保留来源没有披露精确的投后估值。投资团混合了回投方,包括 Founders Fund、Gigascale Capital、Lowercarbon Capital、Unless 和 WovenEarth,也有一批新的深口袋投资者,包括 John Doerr、Marc Benioff 的 TIME Ventures、Max Levchin 的 SciFi Ventures、Susquehanna Sustainable Investments、Hanwha Group、Fortescue Ventures、Super Micro Computer、Sozo Ventures 和俄勒冈本地基金。相关特殊目的载体在 2026 年提交了 SEC Form D 通知,披露的是较小的配售池,而非完整轮次。相较估算的此前融资,Series B 轮是资本台阶式上行,支撑公司从海试原型迈向首个商业试点。[CO012, CO013, CO015, CO016, CO017, CO018]

快照 KPI 表
指标数值 / 状态日期置信度缺口 / 备注
成立20162016 年公开记录二手来源中的成立年份一致;未保留准确注册日期。
总部Portland, Oregon2026 年公开状态总部位于 Portland,并在太平洋西北地区开展海试活动。
法律形式公益公司2026二手报道将其描述为 PBC。
阶段未上市,Series B2026-05-042026 年 5 月宣布 Series B。
累计融资(USD M)2102026-05-04Series B 后披露的近似累计融资额。
Series B(USD M)1402026-05-04Peter Thiel 领投。
最新估值(USD M)FT 称其为「$1bn 海洋数据中心初创公司」;未披露准确投后估值。
员工人数1202026当前员工数为近似值;高于 2024 年初约 70 人。
收入 / 年化收入(USD M)收入前;未披露收入。
客户数量未公开披露付费客户。

后续章节复用规范身份和规模事实;未获支持的估值、收入和客户单元格保持 null,而不作推断。

[CO003, CO004, CO010, CO012, CO015, CO016]
利益相关方或投资者图谱
投资者轮次角色类型置信度
Peter Thiel领投方个人 / Founders Fund 圈层
Founders Fund返投风险投资
Gigascale Capital返投气候风投
Lowercarbon Capital返投气候风投
John Doerr新进个人
TIME Ventures (Marc Benioff)新进风险投资
SciFi Ventures (Max Levchin)新进风险投资
Hanwha Group新进战略 / 工业
Super Micro Computer新进战略 / 硬件
Portland Seed Fund / Intrepid Oregon Fund 早期投资方新进区域风投

返投气候投资方与广泛新财团的组合;角色根据公告报道推断。

[CO013, CO018, CO019, CO020, CO021, CO036]
FO003: 快照 KPI

公开 KPI 栈显示,公司资本雄厚、工程驱动、尚无收入,且最近一轮融资规模很大。

员工数和原型输出是公开近似数,不是经审计的点估计。

[CO010, CO012, CO015, CO016, CO025]

1.4 里程碑、轨迹和未解问题

Panthalassa 的公开时间线从 2016 年成立开始,经过 2021 年在 Juan de Fuca 海峡部署 Ocean-1 原型、后续原型到 2025 年在 Puget Sound 产生约 50 千瓦电力、2026 年 5 月 Series B 轮融资、计划在 2026 年 8 月前后于北太平洋部署 Ocean-3 试点节点,并指向约 2027 年的商业部署目标。公司早期探索过制氢或清洁燃料,后来把重点转向 AI 算力,并提出长期愿景:部署数千个自主节点。与这种势头相对,公司仍未产生收入,也没有公开披露付费客户;独立评论者提醒,腐蚀性强、机械环境恶劣的海洋条件会带来真实执行风险,Panthalassa 至少未回应一项持怀疑态度的媒体询问。估值精度、收入、客户证明和董事会构成,仍是后续章节要继续追踪的主要未解尽调事项。[CO024, CO025, CO026, CO027, CO028, CO029]

里程碑表
日期里程碑类别置信度
2016公司在 Portland, Oregon 成立创立
2021Ocean-1 原型部署于 Strait of Juan de Fuca产品
2024团队约 70 人;继续推进原型工作规模
2025~50 kW 原型在 Puget Sound 测试产品
2026-03SEC Form D SPV 申报($319K,14 名投资者)融资
2026-05-04宣布 $140M Series B,由 Peter Thiel 领投融资
2026-07相关 SPV 申报(~$1.94M,31 名投资者)融资
2026-08计划在北太平洋部署 Ocean-3 试点节点产品
2027目标商业部署产品

时间线混合了融资和产品里程碑;仅有年份的日期采用可得粒度。

[CO004, CO012, CO022, CO023, CO024, CO025]
FO001: 公司里程碑时间线

Panthalassa 的路径从 2016 年创立、海试原型,走到 2026 年 Series B 轮和 2027 年商业化目标。

仅有年份的里程碑使用 1 月 1 日,以保留时间顺序,但不暗示精确日期。

[CO004, CO012, CO024, CO025, CO026, CO027]

1.5 图表

Chapter 02

02市场分析

2.1 市场边界和替代方案

Panthalassa 不能按整个波浪能市场或整个云计算市场来估算。可投资边界是由海洋波浪能供电、部署在海上的 AI 算力:公司在自主节点上发电,并在板上消耗电力来销售算力容量。因此,纳入的支出覆盖整套集成栈:波浪能转换、海上部署、冷却和电力调节、板载 AI 服务器、网络回传以及商业算力服务。排除的支出包括把电力输送上岸的传统波浪发电项目、一般可再生能源发电,以及普通数据中心建设,除非这些支出正被海上算力替代。相关现状是接入电网电力的陆上超大规模或托管容量;邻近替代方案包括海上风电、漂浮光伏、核电或 SMR 支撑的数据中心,以及普通电网采购。这个边界刻意收窄,因为市场命题依赖于用同址海洋发电解决 AI 电力瓶颈,而不是证明所有海洋能源都是大市场。[CM001, CM002, CM003, CM004, CM005, CM013]

市场定义表
类别纳入支出排除支出买方 / 付款方相关性
核心市场由板载波浪发电供能的海上 AI 算力容量并网输出的波浪能电场,以及不与海上电力绑定的通用云云基础设施和 AI 基础设施买方Panthalassa 的主要可投资边界。
需求侧 TAM受电力接入约束的 AI 数据中心用电和算力容量所有企业软件或非 AI 云工作负载超大规模云厂商、新云、AI 实验室解释买方为何可能为新型选址付费。
供给侧视角为有用电力部署的商业波浪和海洋能源不含波浪转换的水电、风电、太阳能或储能基础设施发起方和战略能源团队显示供给成熟度和成本约束。
现状替代方案陆地超大规模数据中心、托管、电网电力、PPA仅限海上的基础设施云容量规划者和数据中心地产团队买方今天熟悉并信任的基线。
相邻替代方案海上风电、漂浮式光伏、核能或 SMR 支撑的数据中心消费者互联网或无关能源市场战略能源采购和企业发展释放受电力约束算力的竞争路径。

市场边界把需求侧算力支出与供给侧波浪能硬件分开,避免本章重复计算广义云或可再生能源市场。

[CM001, CM002, CM003, CM004, CM005, CM013]

2.2 双重口径估算:巨大的 AI 需求与很小的波浪供给

最有用的估算方法是用两副镜头,并拒绝把单位混在一起。需求口径极大:IEA 预计,到 2030 年数据中心用电需求可能达到约 945 TWh / 年,SemiAnalysis 也把能源可得性视为 AI 基础设施的核心约束。这个口径支撑能源背书算力容量的超大 TAM,但不能证明 Panthalassa 能服务这个市场。供给口径小得多:Mordor 估计,波浪能部署量将从 2026 年约 10 MW 增至 2031 年约 125 MW,其他分析机构也把它描述为高增长但早期的市场。Panthalassa 的 SAM 是两副镜头的交集,即波浪供电的海上 AI 算力。可支撑的近期 SOM 更窄:2026–2027 年的试点和早期商业节点,而不是全球 AI 用电需求的船队级百分比。因此,本章分别保留 TWh、MW 和节点数单位,而不是产出一个膨胀的单一 TAM 数字。[CM006, CM007, CM008, CM010, CM011, CM012]

TAM/SAM/SOM 或规模测算视角表
视角发布方 / 依据年份 / 期限数值或单位方法置信度局限
AI 数据中心电力 TAMIEA / SemiAnalysis2030~945 TWh/yr 数据中心用电需求自上而下的需求侧电力视角Panthalassa 若没有已证明的节点规模和买方信任,无法触达。
波浪能供给视角Mordor Intelligence2026~10 MW 已安装波浪能分析师市场预测基线基数很小;并非专指海上算力。
波浪能增长视角Mordor Intelligence2031~125 MW 已安装;~65.7% CAGR小基数上的分析师预测高 CAGR 可能误导,因为绝对 MW 仍然很小。
交叉校验预测DataM / 360i / PW / R&M 来源组2026-2033快速增长的波浪能市场多家分析师估算发布方在期限、地域和付费墙内方法上各不相同。
Panthalassa SAM推导交集2026 年当前波浪供能的海上 AI 算力AI 算力需求与波浪供能海上供给的交集未披露独立市场类别或收入历史。
Panthalassa SOM试点里程碑2026-2027Ocean-3 和早期商业节点受证据约束的近期采用单位需要试点产出、节点数量、定价和合同。

各行有意把 TWh、MW 和节点数量单位作为不同视角混合呈现;不应加总成一个 TAM 数字。

[CM006, CM010, CM011, CM012, CM017, CM018]
FM001: 市场规模口径

受约束的 TAM/SAM/SOM 口径把巨大的 AI 用电需求收窄到波浪能离岸算力,再收窄到 2026-2027 年试点节点。

该图把 TWh、MW 和节点数这些不可兼容单位保留为标签,而不是转换成虚假的共同单位。

[CM006, CM010, CM016, CM017, CM018, CM019]
FM002: 市场估算区间

市场证据从当前很低的波浪能部署量,到未来很高的 AI 用电需求,说明 Panthalassa 不能靠一个数字测算规模。

区间围绕报告点估计取整后上下界,用来显示不确定性;标签标明单位,因为各行不能相加。

[CM006, CM010, CM011, CM014, CM035, CM036]

2.3 买方、用户、付款方和采用路径

买方地图跟随云和 AI 基础设施预算,而不是公用事业采购。AWS、Azure 和 Google 等超大规模云厂商是最显然的原型,因为它们大规模购买数据中心容量、电力、网络和硬件。新云和 AI 算力专门服务商也合理,因为它们把稀缺 GPU 容量变现;只要经济性和可靠性成立,它们可能容忍非传统选址。AI 实验室更可能是用户而非唯一付款方,除非它们直接签约购买批处理或推理容量。采用应从试点、非关键任务批处理和容量预留开始;海上自主节点必须先证明正常运行时间、安全、远程可维护性和网络性能,生产工作负载才会迁移。卫星回传让低时延交互式服务不适合作为第一用例;批量推理、训练相关任务或可容忍延迟的工作负载更可信。预算归属可能横跨云基础设施、数据中心容量规划、AI 基础设施和战略能源采购。[CM009, CM020, CM021, CM022, CM023, CM024]

细分市场 / 买方图谱
细分市场买方用户预算负责人采用触发因素
超大规模云厂商类似 AWS / Azure / Google 的云平台团队云客户和内部 AI 服务云基础设施、数据中心容量、战略能源受电力约束的区域,或对清洁增量容量的需求。
新云类似 CoreWeave 或 Crusoe 的 GPU 容量提供商购买 GPU 时间的 AI 开发者AI 基础设施和容量采购需要有差异化能源背书的 GPU 容量。
AI 实验室前沿模型或推理团队研究人员和生产 ML 应用AI 基础设施、研究算力、财务可容忍海上延迟的批处理或推理工作负载。
战略工业买方大型能源或技术战略方内部算力工作负载企业发展和能源采购主权或战略清洁算力期权价值。
非主要买方公用事业或电网运营商电力客户公用事业采购除非购买算力,否则排除在外,因为 Panthalassa 不向岸上输出电力。

买方角色根据云和 AI 基础设施预算逻辑推断;本次留存市场来源中没有公开披露已签约 Panthalassa 客户或 LOI。

[CM009, CM020, CM021, CM022, CM023, CM024]
FM003: 买家 / 细分市场地图

买家意愿取决于谁控制云基础设施预算,以及哪些工作负载能容忍离岸交付。

这些细分是根据公有云和 AI 基础设施市场结构推断出的买方原型,并非 Panthalassa 已披露客户。

[CM020, CM021, CM022, CM023, CM024, CM025]
FM004: 采用漏斗或价值链图

采用路径从市场痛点出发,经过工作负载选择、试点验证、容量签约,最终到舰队扩张。

漏斗数值是示意性的采用指数分数,不是市场份额估计;它们反映的是证据受限下的阶段风险。

[CM024, CM025, CM026, CM031, CM033, CM034]

2.4 增长驱动、约束和尽调缺口

多头逻辑由 AI 电力稀缺、并网延迟、土地约束、脱碳压力以及对主权算力容量的战略兴趣驱动。这些因素影响估值,因为如果替代方案是等几年拿土地、电网接入和购电协议,非传统选址就可能具备经济相关性。约束同样重要。商业波浪能仍处早期;一般海洋能源 LCOE 约 $388–$618 / MWh,远高于主流太阳能或风能基准;海上结构和 GPU 载荷带来重大资本密集度;监管、海事、环境和信任门槛,即使没有岸上送电,也可能拖慢采用。Panthalassa 及其投资者引用的是愿景型经济性,包括低目标电力成本和工厂规模节点生产;但保留的公开来源没有披露已签客户、定价、节点级正常运行时间、已签容量或经验证的单位经济。尽调因此应保留相互矛盾的估计,而不是把它们抹平:巨大的 AI 用电需求是真实的,波浪能部署量很小也是真实的,Panthalassa 的实际 SAM / SOM 取决于试点结果和买方信任。[CM014, CM015, CM026, CM027, CM028, CM029]

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 算力电力稀缺驱动因素当前至 2030 年催生对非传统、电力背书算力容量的需求。梳理实际缺电且愿意签约的目标买方。
电网并网和土地瓶颈驱动因素当前如果部署周期更快,海上选址会更有吸引力。比较端到端许可和部署时间线,与陆地替代方案对照。
脱碳和能源主权驱动因素当前支撑清洁本土算力供给的战略溢价。验证买家是否愿意为波浪能驱动的海上算力付溢价。
波浪能装机基数很小制约因素当前抬高供给侧执行的规模和可靠性风险。将 Ocean-3 的发电输出和可用率与波浪能市场预测对标。
海洋能源 LCOE 偏高制约因素当前除非一体化算力经济性跑赢通用波浪能基准,否则 ROI 会被削弱。要求提供经审计的节点级成本、容量因子和维护数据。
资本密集度制约因素近期商业收入出现前需要大量融资。压测工厂资本开支、GPU 采购和海上运营预算。
时延和网络制约因素近期早期采用会先偏向批处理或可容忍延迟的工作负载。按工作负载逐项跑时延和吞吐试验。
监管、海事和信任门槛制约因素当前至中期即使不向电网外送,也可能拖慢试点和生产转化。获取许可路线图、保险、安全、可用率和客户验收标准。

驱动因素和制约因素都挂钩采用节奏与尽调问题,因为只有买家把算力短缺认知转成海上合同,市场才算真实。

[CM014, CM015, CM026, CM027, CM028, CM029]

2.5 图表

Chapter 03

03竞争对手

3.1 格局:直接同行、替代方案和可能进入者

Panthalassa 很难干净地映射到单一竞争类别。最接近的发电侧同行是 CorPower Ocean、Oscilla Power、Eco Wave Power、C-Power、Marine Power Systems、Wave Swell Energy 和 Carnegie Clean Energy 等波浪能开发商,但它们公开的产品表述主要强调发电,而不是在海上销售 AI 算力。最接近的算力侧同行是 Aikido、NetworkOcean、Microsoft Project Natick、Subsea Cloud、Highlander Hailanyun 以及 Mitsui O.S.K. 船载研究等漂浮、水下或船载数据中心概念;但这些替代方案通常依赖海上风电、海底冷却、驳船或船舶,而不是 Panthalassa 式的自主波浪供电节点。现状仍是陆上超大规模数据中心、GPU 云新云、内部自建、电网电力,以及新兴的核电或面向数据中心的 SMR 方案。Starcloud 是有用的前沿类比,因为它在太空推进远程可再生能源供电算力,说明当地面约束看起来具有约束力时,投资者会为非陆地基础设施买单。[CP001, CP002, CP031, CP032, CP033, CP045]

竞争者画像表
竞争者 / 群组类别规模 / 融资信号目标客户产品范围定价 / 打包战略方向 / 限制
Panthalassa一体化波浪能驱动 AI 算力2026 Series B 轮 $140M;总融资约 $210M需要清洁远程容量的 AI 算力买家搭载算力的自主波浪能节点未公开标价;销售算力容量波浪发电与算力独特融合;尚未商业化、尚无收入。
CorPower Ocean波浪能开发商CorPack 集群被描述为 10-30MW 阵列公用事业公司和可再生能源项目开发商波浪能转换器和阵列留存来源未公开项目 / 设备经济性发电可信度较强,但没有海上 AI 算力产品。
Oscilla Power波浪能开发商Triton 有 16 项已授权专利支撑能源、国防、国土安全、海洋学Triton WEC 和传动系统技术留存来源未见公开标价有专利支撑的发电同类;不是算力提供商。
Eco Wave Power上市波浪能公司Nasdaq WAVE;披露项目管线 404.7 MW港口、沿海基础设施、电网 / 工业买家利用现有结构做岸上波浪能转换有上市公司披露,但无可比算力价格把波浪能与 AI 工厂相连,但模式以发电为先。
Aikido Technologies浮式数据中心平台声称可复用 GW 级海上风电;报道 2026 年挪威 100 kW 验证主权和 GPU 算力客户搭载 AI 级算力的浮式风电平台未公开算力单价可能从风电角度切入,并带有 NVIDIA 生态信号。
NetworkOcean浮式 / 水下数据中心创业公司公开信息仍处早期寻求海洋选址的云 / 托管机房买家驳船和水下舱体声称比陆上便宜,但未公开资费算力选址同类,但未声称使用波浪发电。
Microsoft Project Natick海底数据中心研发Phase 2 有 864 台服务器和 27.6 PB;到 2024 年已停用Microsoft 内部云研发密封海底服务器模块研发项目,不是商业打包强验证样本,也提醒商业化可能走不通。
Highlander / Hailanyun China商业水下数据中心报道 2.3 MW 演示正扩至 24 MW中国绿色算力和沿海工业用户与海上风电绑定的压力容器模块留存材料未见可比公开价格水下数据中心最可见的规模化样本。
Starcloud太空数据中心类比$170M Series A 轮;估值约 $1.1B受陆上电力制约的 AI 工作负载太阳能供电的在轨数据中心未来成本目标,不是可比商业定价显示资本愿意押注远程算力,但运营场景不同。
陆上超大云厂商 / 新云厂商现状 / 替代选项CoreWeave 和 Crusoe 代表成熟买方替代方案AI 实验室、企业、超大云厂商GPU 云、托管机房、自建、电网供电已有云合同和 SLA,但本处未做基准分销和信任上的既有玩家;陆上电力瓶颈仍在。

画像行结合竞争者官方公开信息和独立报道;没有留存来源披露可比资费时,定价保持为未知。

[CP001, CP003, CP006, CP008, CP009, CP011]
FP001: 竞争定位图

序数图:x = 计算整合强度,y = 海洋可再生能源自主性;Panthalassa 融合度最高,但商业验证最低。

分数为 0-5 的序数判断,来自留存证据,不是来源发布的数值排名。

[CP001, CP018, CP025, CP031, CP032, CP033]

3.2 公司画像、成熟度和功能广度

画像对比显示,在保留公司中,Panthalassa 是唯一声称把波浪发电和板载 AI 算力放进同一商业架构的公司。CorPower 通过 10–30MW CorPack 集群给出更清晰的公用事业级波浪阵列包装;Oscilla 拥有专利支撑的 Triton 波浪能转换器;Eco Wave Power 具备上市公司能见度和 404.7 MW 管线;Aikido、NetworkOcean、Natick、Highlander 和 Subsea Cloud 则从冷却、漂浮平台或海底角度切入海上数据中心问题。Microsoft Project Natick 是最成熟的公开证明,显示服务器可以在海底运行,包括一次 864 台服务器部署,故障率低于陆上对照组;但同一项目已不再活跃,这削弱了「技术成功意味着商业延续」的假设。因此,功能地图给 Panthalassa 高广度、低证明;陆上新云则相反,商业成熟度高,但没有自主海洋可再生能源护城河。[CP003, CP006, CP008, CP009, CP011, CP012]

功能 / 能力矩阵
采购标准Panthalassa波浪能同类海上数据中心同类太空数据中心类比陆上 / 自建现状
已验证的波浪发电已有原型 / 试点证据;商业验证仍待完成CorPower、Oscilla、Eco Wave 的核心产品品类通常不靠波浪能供电不适用不适用
海上 AI 算力集成核心差异点通常缺位Aikido、NetworkOcean、Natick、中国 UDC 的核心核心能力,但在轨道而非海洋核心算力已在陆上存在
自主自推进节点报道称采用无缆、自推进架构通常是系泊或按项目部署的发电设施多为驳船、海底模块或平台航天器自主性陆上设施
冷却 / 热管理优势公司声称采用海水冷却密封模块不以算力为重点海底 / 浮式同类的主要价值主张需要太空热管理系统液冷 / 风冷和用水约束
连接 / 时延状态卫星回传带来时延和带宽限制重点是向电网送电大概率靠光纤 / 海缆或沿海登陆点需要太空通信光纤资源充足,SLA 已验证
商业成熟度尚未商业化、尚无收入波浪能行业规模仍小,有试点和管线Natick 已验证但退役;中国在扩张;其他仍早期已获融资的前沿原型路径最成熟、最受信任
监管 / 环境信任海上许可和生态影响路径尚未验证海洋能源许可路径熟悉但难度高海洋热排放和海底运营会受环境审查太空许可和发射风险监管制度清楚,但地方反对和并网队列仍在
公开定价可见度留存材料未见标价留存材料未见可比资费留存材料未见可比资费只有未来成本目标云合同存在,但本处未做价格基准

矩阵单元格是基于留存公开来源的序位证据判断;缺乏支撑的定价和 SLA 单元格明确标为未知,而不是估算。

[CP001, CP002, CP012, CP014, CP023, CP024]
FP002: 功能广度 / 能力图

Panthalassa 在波浪、自主性和计算上覆盖较广;竞争对手要么在单一类别更深,要么在陆上更成熟。

低 / 中 / 高标签反映公开证据的覆盖广度和成熟度,不是工程性能评分。

[CP002, CP003, CP006, CP008, CP009, CP011]

3.3 能力、定价、GTM 和信任姿态

能力对比在架构集成上偏向 Panthalassa,但在证明、定价或信任上还不成立。公开来源支持节点概念、无电缆架构和海上处理路径;波浪能同行支撑可再生发电可信度,海上算力同行支撑冷却或选址可信度。相关前沿集合普遍缺少公开标价,买方还无法按 $/GPU-hour、$/kW、服务等级条款、数据出站经济性、时延或合同期限来比较 Panthalassa。GTM 权力也不对称:成熟超大规模云厂商、GPU 云提供商和硬件生态已经控制买方关系和采购信任;Panthalassa 则必须证明远程运营、带宽、正常运行时间、安全、环境合规和海上可维护性。关键投资判断是:集成节点可以具备战略差异化,但如果买方偏好已知云厂商、更清晰的 SLA 或更便宜的并网算力,它仍可能输掉。[CP002, CP024, CP025, CP034, CP035, CP036]

定价 / 打包对比
替代方案打包模式公开价格证据买家含义证据状态
Panthalassa海上节点自发电支撑的 AI 算力容量未披露标价或 SLA必须私下核验 $/GPU-hour、可用率、带宽和时延仅有私下证据的缺口
波浪能开发商项目、设备或发电部署无可比算力价格可用于发电基准,不适合直接对标算力采购产品证据公开,定价有限
Aikido浮式海上风电数据中心平台留存材料未见公开算力资费可打包为主权 / 海上 GPU 容量创业公司公开信息仍早期
NetworkOcean浮式驳船和水下舱体声称比陆上便宜;未留存资费价格说法需要对照陆上替代方案验证仅为供应商说法
Natick / 中国水下数据中心研发模块或海上风电供电的 UDC 部署Microsoft Natick 无商业报价;未留存中国价格更能验证技术可能性,而非采购可比性研发项目与第三方报道混合
Starcloud未来在轨 AI 算力基础设施未来成本目标无法与当前云合同可比可作估值类比,今天还不是采购可比样本已获融资的前沿类比样本
CoreWeave / Crusoe / 超大云厂商陆上 GPU 云、托管机房或自建数据中心留存材料之外存在公开云 / 合同价格买家可分散到多供应商,并要求已知 SLA现状替代项
借助电网 / 海上风电 / 核电自建自有数据中心加电力采购项目专属资本开支 / 运营开支,本处未留存如果 Panthalassa 证明需求,既有玩家可以垂直整合尽调需要拿到私有成本结构

留存来源没有提供海上前沿组的同口径 $/GPU-hour 或 $/kW 定价,因此未知项保留为证据缺口。

[CP024, CP025, CP033, CP034, CP035, CP036]

3.4 切换成本、多云使用和供给获取

对目标 AI 算力买方而言,切换成本看起来取决于工作负载,而不是绝对刚性。批量推理、离线训练支持和可中断工作负载,比面向用户的低时延服务更容易跨提供商路由;这支持多云使用,但限制锁定。Panthalassa 如果控制稀缺清洁电力容量、能在能量密集的波浪区域稳定运行,并暴露熟悉的云接口,就可能形成锁定;如果时延、带宽或信任约束让客户把主要工作负载留在陆上,它也可能成为二级容量提供商。合作伙伴准入同样有希望但未验证。Super Micro Computer 出现在投资者名单中,对硬件供给是方向性利好;Aikido 和 Starcloud 也显示,与 NVIDIA 相关的生态正在主动接触远程算力概念。这些都不能证明配额、客户承诺或定价权,因此合作伙伴尽调应聚焦 GPU 采购权、服务协议、卫星带宽、保险和维护物流。[CP017, CP029, CP034, CP035, CP036, CP037]

FP003: 护城河 / 就绪度 KPI

核心 KPI 显示融资能力强、整合路径独特,但客户验证偏弱,海上计算先例也不利。

KPI 值混合了来源报告的融资、部署规模和公开披露数量;不应相加。

[CP010, CP012, CP015, CP018, CP027, CP028]

3.5 护城河耐久性和反向竞争证据

护城河在概念上真实存在,但耐久性尚未证明。Panthalassa 拥有一个有差异化的垂直整合故事:量产钢制节点、自主定点保持、波浪发电、板载算力和卫星回传。风险清单主要由一个事实主导:海洋系统和海上算力过去很难成为具备经济性的基础设施平台。Project Natick 给出了强技术结果,却已退出;New Scientist 强调盐雾和波浪条件严酷,也强调必须在经济性上打败传统数据中心;OTEC 等历史海洋能源实验则说明,天气和海浪可能在实现净发电之前摧毁基础设施。商品化风险同时压在产品两侧:波浪发电诀窍可在海洋能源同行间扩散,算力买方也能在超大规模云和新云之间多云使用。在 Ocean-3 和早期客户合同证明正常运行时间、单位成本和监管路径之前,Panthalassa 应被视为具备竞争差异化,但尚未受到护城河保护。[CP023, CP024, CP030, CP038, CP039, CP040]

护城河耐久度 / 竞争风险台账
护城河主张竞争威胁严重度缓释措施 / 尽调问题
波浪发电与节点内 AI 算力一体化海上算力同类可加入可再生能源供给,但不集成波浪能要求 Ocean-3 证明发电、冷却、算力、通信和自主性能够协同运转。
自主、无缆定位保持海洋环境、风暴、盐雾、腐蚀和维护物流审阅海试日志、故障模式、保险条款和维护计划。
绕开并网和土地瓶颈陆上超大云厂商可通过 PPA、核电、SMR 或海上风电锁定电力对照陆上替代方案,基准化交付成本和部署时间。
可制造的钢板节点设计一旦公开验证,波浪能技术诀窍和制造能力可能商品化核查知识产权、供应链、工厂资本开支和专有控制。
远程清洁算力容量客户可多供应商部署,并把关键工作负载留在可信云厂商验证已签客户、SLA 要求、工作负载匹配度和切换行为。
硬件生态访问GPU 供给仍由超大云厂商、新云厂商和硬件厂商掌控确认投资人名单之外的 GPU 配额权和战略供应商合同。
海底 / 浮式算力验证样本Project Natick 说明技术成功也可能退役,而不是商业化区分技术演示 KPI 与可复制商业运营、单位经济。
海洋前沿叙事和投资人光环Starcloud 和其他前沿类比公司会争夺资本和人才注意力跟踪太空和海上算力新进入者的融资、招聘和伙伴公告。

风险台账聚焦 Panthalassa 专属差异化能否持久,不覆盖后文讨论的一般公司执行风险。

[CP001, CP014, CP018, CP019, CP023, CP024]

3.6 图表

Chapter 04

04财务

4.1 收入模式:算力优先,燃料可选,收入缺席

Panthalassa 的财务故事从一个清晰区分开始:它没有公开向岸上售电。公司称,其自主节点在海上产生波浪电力,并在板上消耗这些电力来运行 AI 芯片,再通过卫星把推理 token 返回陆地。这让近期收入假设成为一种算力容量服务,而不是带有电网售电协议的传统电力项目。投资者和公司材料仍为清洁燃料或氢气作为丰富海洋电力的未来用途留出空间,但 2026 年融资叙事和 Ocean-3 里程碑集中在海上 AI 推理。没有保留来源披露确认收入、ARR、按 token、GPU 小时或预留计费的定价,也没有披露清洁燃料产出。正确的尽调姿态是把今天的收入建模为零,并把每一种变现机制都视为取决于 Ocean-3 证明和首批商业合同。[CI001, CI002, CI003, CI004, CI005, CI006]

收入来源表
收入来源2026 年公开状态变现基础收入确认 / 质量问题证据立场
AI 推理算力容量规划中;未披露收入可能按 token、推理任务、GPU-hour 或容量合同收费;未公开计价单位收入确认取决于未来服务合同、可用率、可承接工作负载和计费条款证据支持其为主模型,但尚未公开商业化
向岸上售电不是该模式不向电网外送;电力在节点内消耗按所述架构,没有可确认的售电收入公司叙事明确排除
清洁燃料 / 氢可选未来用例可能有能源承购;未公开买家或价格没有承购和生产数据,无法评估收入确认投资人 / 背景材料提及,但次于算力
冷却 / 数据中心基础设施收益嵌入算力服务可能降低成本或延长芯片寿命,而非形成单独收入线除非单独签约,否则收益体现在利润率而不是收入公司已声称,未定价
工厂或节点授权未披露未公开授权、租赁或销售模式无可用收入处理方式推测项;排除在基准情形外

收入行把已陈述商业模式与可选用例分开;商业证据为空时保留为空,不做推断。

[CI001, CI002, CI003, CI004, CI005, CI030]
定价 / 变现表
定价要素公开数字 / 状态可能驱动因素披露缺口尽调含义
算力单位价格token、推理任务、GPU-hour 或预留容量未公开价目表或合同计价单位无法测算单节点收入
能源成本目标$0.02/kWh 目标波浪能转换和高利用率来自投资方的目标,不是已实现成本只作为上行情景敏感性
容量预留 / 承购超大云厂商或 AI 买家承诺未公开 LOI、在手订单或合同收入质量无法判断
清洁燃料产出价格若推进,则为氢气或燃料承购无产量或买方排除在近期预测之外
收入确认触发条件已交付算力、已验收任务或可用性 SLA未披露服务协议会计处理是尽调卡点

公开的唯一货币化数字是电力成本目标,不是客户价格,也不是已确认收入指标。

[CI005, CI006, CI007, CI030, CI035]
FI001: 收入模型桥

Panthalassa 的财务桥从波浪发电开始,把电力转换为板载计算;只有商业合同落地后,才可能转化为收入。

流程展示变现逻辑,不是收入预测或证明。

[CI001, CI002, CI003, CI004, CI005, CI030]

4.2 GTM 代理是需求侧和里程碑驱动,不是销售效率证明

从公开证据看,Panthalassa 的进入市场动作由基础设施驱动。公司必须先证明 Ocean-3 能在海上产出可用推理,再把证明转成与超大规模云厂商、AI 实验室、新云或能容忍卫星链路工作负载的企业算力买方签订的合同。需求侧论点清晰:陆上数据中心面临电网、冷却水、许可和社区约束,Panthalassa 则提出把能源和冷却问题移到海上。但销售效率无法衡量。公开信息没有 CAC、回本周期、销售周期、渠道利润率、管线转化、积压订单或客户集中度。最强牵引代理不是收入,而是投资者信念和部署进度:Thiel 领投的大额 Series B 轮、广泛的战略投资团,以及在 2027 年目标商业部署前计划推进的 2026 年 Ocean-3 试点。[CI016, CI017, CI018, CI019, CI029, CI031]

4.3 单位经济取决于愿景型电力成本和未定价的海上运营

公开单位经济故事很有吸引力,但仍是愿景型。Lowercarbon 重复约 $0.02/kWh 的目标,Gigascale 引用约 $1,500/kW 的制造 capex、约 90% 容量因子主张,以及一个情景:一座 $1B 工厂每年生产约 1 GW 节点容量。如果实现,这些数字会非常有吸引力;但它们是投资者和管理层来源的目标,不是商业船队经审计的运营结果。成本栈比能源转换更宽:钢板海洋结构、涂层、涡轮、电力电子、GPU、密封算力容器、卫星链路、拖航、维护船、保险、腐蚀管理、生物污损和更换周期都重要。DataDeep 的反向评估直接质疑,海上运营和维护能否让两美分电力成本叙事在真实部署中站住。[CI007, CI008, CI009, CI010, CI011, CI012]

单位经济模型表
驱动项公开数据 / 状态财务影响置信度缺口 / 风险
目标发电成本$0.02/kWh若其他成本不变,可支撑低成本算力海上未验证,反向来源质疑全成本可行性
制造资本开支目标~$1,500/kW把规模化节点资本开支框定为接近燃气电厂水平来自投资人材料;未计入实际物流和维护
容量因子据称 ~90%用更多发电小时摊薄固定资本开支需要 Ocean-3 扣除自耗电后、按季节拆分的数据
工厂产能$1B 工厂 -> ~1 GW/year规模化生产可能压低单位成本收入验证前就要砸下大额前期资本开支
钢制海上结构~85m 全钢节点显著推高制造、涂层、拖航和折旧实际物料清单未披露
GPU / AI 载荷大额资本开支与更新周期的主要驱动项载荷成本、供应商条款和折旧未披露
海上运维 / 保险可能吃掉毛利率腐蚀、生物污损、风暴、服务船和保险均未定价
卫星连接决定可用工作负载和交付成本回传定价和 SLA 经济性未披露

本表把单位经济模型说法视为目标;没有现场证据,不把它们转成预测毛利率。

[CI007, CI008, CI009, CI010, CI011, CI012]
FI002: 单位经济性桥

不同于单位经济性表,这张桥图展示目标成本主张在形成毛利前,必须先扛住物理、运营和服务交付成本。

与表格不同的视角:映射从主张到利润率的依赖链和不利成本泄漏,而不是列指标行。

[CI007, CI008, CI009, CI010, CI011, CI014]

4.4 资本充足性:大额融资,更大的证明负担

资本获取是 Panthalassa 最清晰的财务强项。公司于 2026 年 5 月 4 日宣布完成由 Peter Thiel 领投的 $140M Series B 轮融资,GeekWire 报道累计融资约 $210M。该轮资金用于完成波特兰附近的试点制造设施,并加速 Ocean-3 部署;但公司没有披露账上现金、月度 burn、runway、债务、项目融资或走向商业规模所需的预期 capex。两份 SEC Form D 文件提供了有用颜色,但不是完整轮次账目:一个 Series B 配售载体报告向 14 名投资者售出 $319,000,之后一个 B Plus 载体报告向 31 名投资者售出 $1,935,313。这些看起来是部分 SPV 配售文件,而不是头部轮次。融资依赖仍然很高,直到 Ocean-3 产出可审计运营数据或首批商业合同。[CI017, CI019, CI020, CI021, CI022, CI023]

资本充足性表
资本项目金额 / 状态日期解读尽调备注
Series B 轮$140M2026-05-04公开披露的主要新增融资Peter Thiel 领投;资金用于工厂和 Ocean-3 部署
累计融资~$210M2026-05-04Series B 轮后累计融资意味着此前融资约 $70M
估计此前资本~$70MSeries B 轮前由累计融资减去 Series B 轮推算不是单独审计披露
Form D Series B 轮 SPV$319,000 / 14 名投资者2026-03-25 申报部分配售工具不是完整 Series B 轮
Form D B Plus 轮 SPV$1,935,313 / 31 名投资者2026-07-21 申报后续部分配售工具不是完整 Series B 轮
账面现金未披露需要管理账和银行余额
烧钱速度 / 现金跑道未披露需要月度烧钱和本轮后现金跑道
债务 / 项目融资未披露需要授信额度、担保权益或项目融资计划
下一轮融资触发点Ocean-3 数据 / 首批合同2026-2027下一轮大概率取决于验证节点用董事会计划和里程碑预算核验

融资金额公开;实际流动性、烧钱速度和项目融资能力不公开。

[CI017, CI019, CI020, CI021, CI023, CI024]
FI003: 财务估算区间

公开数值锚点主要是融资和目标成本区间;运营类财务指标仍然缺失。

混合单位的区间标签明确列出单位;融资数字是公开数据,成本数字是目标。

[CI007, CI008, CI019, CI020, CI021, CI022]
FI004: 资本强度 / 现金流图

资本先从投资者流向工厂和试点部署,之后才可能出现公开收入流;因此 Ocean-3 验证是闸门事件。

现金流顺序根据资金用途披露和缺失的流动性指标推断,不来自公司预算。

[CI017, CI019, CI020, CI026, CI027, CI028]

4.5 公开财务缺口主导尽调

公开记录在核心运营指标上很稀疏。收入、ARR、毛利率、贡献毛利、营运资本、CAC、销售回本、现金余额、burn、runway、客户合同、利用率、活跃用户和项目融资条款都缺失。对一家私有、未产生收入的前沿基础设施公司而言,这种缺失并不意外,但它阻止了常规软件、能源或数据中心投资判断。正确处理方式是保留 null,而不是用融资公告或投资者引述倒推出隐含值来填补。财务尽调请求清单应从管理账、银行余额和 runway、客户管线和合同草案、节点 BOM、维护预算、保险报价、GPU 采购和折旧假设,以及 Ocean-3 遥测协议开始;这些协议需要区分总发电量、寄生负载、可用算力和正常运行时间。[CI001, CI005, CI006, CI013, CI015, CI018]

公开财务缺口表
指标公开值缺口类型重要性尽调路径
收入缺少公开来源无法分析收入质量和增长索取按产品和客户拆分的月度收入
年经常性收入(ARR)缺少公开来源没有经常性收入基数或留存证据若已有合同,索取 ARR/MRR 明细表
烧钱速度仅有私有证据决定现金跑道和下一轮时间索取现金流量表和经营计划
现金跑道仅有私有证据显示 Series B 轮后资本是否充足将账面现金与月度烧钱和资本开支计划核对
毛利率仅有私有证据海上算力经济性的核心证明索取单节点销货成本(COGS)、维护、电力、连接和折旧模型
获客成本(CAC)/ 回本周期缺少公开来源显示商业化效率和企业销售负担索取管线、销售周期、赢单率和获客支出
账面现金仅有私有证据用于判断偿付能力和融资依赖索取银行流水和董事会批准预算
客户合同 / 积压订单缺少公开来源验证商业需求和收入确认索取已签合同、LOI 和 SLA 条款

必填的 null 指标按设计保留为 null,因为保留的公开来源均未披露这些数据。

[CI001, CI005, CI006, CI015, CI018, CI026]

4.6 财务结论:有期权价值,但尚无收入质量

从财务上看,Panthalassa 是一项高期权价值的基础设施赌注,而不是一家已有可投资判断收入质量的公司。正向逻辑是资本获取可信、需求瓶颈清晰、在海上消耗电力的模式有差异化,以及大投资者对近期试点押注。反向逻辑同样直接:公开信息没有收入、没有客户、没有毛利率、没有 CAC、没有 burn 或 runway、没有现金余额,也没有经审计数据证明海洋维护和算力运营能守住模型中的单位经济。因此收入质量不适用;利润率路径未证明;资本密集度很高。可投资触发点不是另一段媒体引述,而是 Ocean-3 的实测输出、正常运行时间、维护节奏和首份商业合同经济性。在此之前,结论是观察并验证,且存在重大尽调阻碍。[CI034, CI035, CI036, CI037, CI038, CI039]

4.7 图表

Chapter 05

05产品与技术

5.1 产品定义和买方工作流

Panthalassa 不是在销售传统波浪能电站、购电协议或海上数据中心外壳。产品是一个自主漂浮节点,把海浪运动转成板上电力,并立即用这些电力在海上运行 AI 芯片。放到客户工作流里,买方会把合适的 AI 任务路由到 Panthalassa 容量,节点在密封海洋算力模块内执行推理或更长时间的算力工作负载,再通过卫星回传把结果送回陆地。公司明确声称电力永不输出上岸,因此客户收益不是更便宜的电网电,而是绕开土地、电网并网、冷却水和许可瓶颈的算力。这使工作负载选择成为核心:延迟出结果的推理、仿真和其他批处理类任务,比时延敏感的聊天机器人或搜索流量更适配。因此,产品必须作为一套耦合服务工作流来尽调,横跨海上发电、远程算力运营、卫星网络和客户任务编排,而不是作为单独的海洋发电机。[CE001, CE002, CE003, CE013, CE016, CE018]

工作流 / 用例表
用户任务当前工作流Panthalassa 方案可衡量收益限制
批量推理或延迟 AI 任务在受电网和冷却容量限制的陆地数据中心运行把任务路由到海上节点;通过卫星返回推理 token / 结果不新增陆地数据中心用电,也能增加算力延迟和带宽不适合许多交互式应用。
科学仿真 / 长时间计算任务使用陆地 HPC 或受电价影响的云区域在海上波浪发电处运行负载可能降低碳强度,且无需岸电缆需要已验证的调度器、数据传输和客户 SLA。
AI 算力扩张建设或租用陆地数据中心容量从分布式自主节点购买算力绕开部分土地、水、电网和许可瓶颈未披露客户或生产验收测试。
绿色算力采购为陆地算力签订可再生能源或抵消合同直接使用海浪能驱动的算力可再生发电与算力在物理上绑定更紧环境热影响和生命周期分析未公开。
远程基础设施演示分别试点孤立发电或海上算力在一个节点内演示发电、冷却、算力和回传集成验证点可降低规模化船队风险原型数据仍由公司转述。

用例强调适合卫星回传的工作负载;真实定价和 SLA 未披露。

[CE001, CE002, CE003, CE018, CE024, CE041]
FE002: 客户工作流 / 运行流程

服务工作流从可容忍卫星延迟的工作负载开始,以结果回传陆地结束;电力则留在板载侧。

流程假设存在合适的任务调度器和客户集成层,但公开信息未说明。

[CE001, CE002, CE003, CE018, CE040, CE044]

5.2 节点资产和运营架构

对一家早期硬科技公司而言,公开架构异常具体。Ocean-3 被描述为一座约 85 米的钢板结构,形状像高尔夫球放在球座上;投资者报道称,上部为约 50 米球体,水面以下有长颈结构。海浪抬升和降低该结构时,海水被迫沿中央管道进入内部水库,再通过单个涡轮排出发电。这些电力为密封 AI 服务器供电,服务器借助周围海水冷却。Ocean-2 是较小的原型类比,上部球体约 9 米;公开报道显示,它在 Puget Sound 条件下最高约 50 千瓦。因此,架构图应被理解为一个垂直整合资产:船体、波浪泵、水库、涡轮、电力电子、密封算力、海水换热、自主系统和卫星通信。设计可能优雅,因为它减少运动部件并避免岸上电缆;但商业证明仍取决于开阔海域条件下的生存能力、维护和实测输出。[CE004, CE005, CE006, CE007, CE008, CE010]

产品模块 / 资产矩阵
模块 / 资产使用方 / 负责人状态 / 成熟度差异化尽调缺口
Ocean-3 节点船体Panthalassa 运营与制造计划 2026 年试点;未在商业规模验证~85m 板钢自航结构,不接岸电缆远海耐久性、风暴生存能力、维修间隔和认证图纸。
Ocean-2 原型工程与海试团队在海峡 / Puget Sound 完成原型测试;公开报道 ~50 kW更小的 9m 球顶实物验证点带仪器记录的测试日志、独立核验和故障历史。
Wavehopper / Ocean-1 原型工程验证历史原型证据显示 Ocean-3 之前已有十年原型迭代规格和复盘教训未公开披露。
波浪泵 / 蓄水池 / 涡轮机发电子系统机制已有描述;船队输出未验证同一船体内用单蓄水池、单涡轮机完成转换效率曲线、污损耐受度和维护入口。
密封算力舱算力载荷团队 / AI 客户概念已有描述;无公开在线率数据发电资产内嵌海水冷却 AI 芯片热性能、进水控制、PUE 和消防安全。
卫星回传与自主运行远程运营 / 客户工作流已有公开描述;SLA 未知LEO 卫星链路允许节点部署在远海带宽、延迟、加密、故障切换和支持流程。

资产成熟度来自公开原型和路线图证据;各行不是可购买的商业 SKU。

[CE004, CE006, CE007, CE008, CE010, CE014]
技术 / 运营架构表
层级 / 组件作用依赖风险
板钢船体与球形顶部捕捉波浪运动,并承载蓄水池 / 算力结构沿海钢结构制造、涂层、压载设计腐蚀、疲劳、风暴载荷和拖航物流。
中央管与蓄水池波浪推动节点时,将海水泵入蓄水水头水动力调校和进水口耐久性生物污损或漂浮物可能削弱流量和输出。
单涡轮机与电力电子把蓄水流转换为船载负载用电涡轮机可靠性、变流器、控制系统冗余设计披露前,单点会卡住性能。
GPU / AI 算力载荷执行推理或其他适配的 AI 工作负载GPU 供应、密封机架、电源调理局部过热、硬件故障和维修入口。
海水冷却 / 换热通过密封舱壁和海水混合排出服务器热量舱体材料、监测、洋流PUE、废热影响和进水风险未知。
LEO / Starlink 回传在节点与陆地之间传送任务 / 结果卫星带宽、天线可用性、抗天气能力延迟高于光纤,限制实时工作负载。
自主运行 / 定位保持无需系泊或发动机即可保持或改变位置船体形状、压载、软件、天气数据远程运行没有公开可靠性统计。

架构揉合公司说法、独立描述和技术类比;确切工程规格仍属私有。

[CE007, CE008, CE010, CE011, CE012, CE014]
FE001: 产品架构图

单个节点把波浪捕获、蓄水池 / 涡轮转换、板载电力调节、密封计算、海水冷却、自主运行和 LEO 回传叠进一项资产。

层级标签综合公开架构描述;详细工程图纸未公开。

[CE004, CE007, CE008, CE010, CE011, CE014]

5.3 部署成熟度、集成和路线图

Panthalassa 的公开成熟度应评为从原型到试点,而不是商业化。序列从 2021 年 Ocean-1 开始,经过 2024 年 Ocean-2 和 Wavehopper 原型活动、2025 年报告中的 Puget Sound Ocean-2 近 50 千瓦测试,再到计划于 2026 年在北太平洋开展 Ocean-3 试点系列,用于演示 AI 推理并优化制造。商业部署目标约在 2027 年,长期愿景则是数千个节点。集成不是简单安装动作:它需要沿海工厂生产、拖出或下水物流、自主定点保持、远程监控、卫星连接、天气航线规划、维修流程、GPU 载荷采购和客户工作负载调度。由于已有原型和融资,路线图主张值得跟踪;但它还不能作为服务等级产品来放进银行级模型,因为公开来源没有披露正常运行时间、维修间隔、船队遥测、环境许可或客户验收测试。正确的投资判断姿态是低到中等 TRL,并把 Ocean-3 作为重大证据检查点。[CE019, CE020, CE021, CE022, CE023, CE024]

路线图 / 发布 / 开发阶段表
日期 / 阶段功能 / 里程碑状态含义来源依据
2021Ocean-1 原型在 Strait of Juan de Fuca历史原型显示早期海上工作,但不能证明商业性能公司材料与二手历史资料。
2024Ocean-2 与 Wavehopper 原型活动原型验证支撑大型试点前的能力说法PR Newswire 和公司视频这类替代证据。
2025Ocean-2 在 Puget Sound 测试,约 50 kW据报道的海试对发电是有用验证点,但仍远低于 Ocean-3 规模Wikipedia / TechEBlog / Puget Sound 背景。
2026Ocean-3 试点节点系列,北太平洋计划中 / 进行中海上 AI 推理的下一个主要 TRL 关口PR Newswire、TechRadar、Wikipedia。
2027商业部署目标愿景式路线图商业就绪取决于 Ocean-3 数据和支持模型公司转述的路线图。
长期数千个节点和 GW 级工厂产出愿景 / 期权价值若验证通过,可能形成制造护城河Gigascale 与公司陈述。

路线图揉合历史原型、2026 年试点计划和愿景式规模;商业日期不是有约束力的客户承诺。

[CE019, CE020, CE021, CE022, CE023, CE024]
FE004: 产品成熟度 / 能力图

发电和原型证据领先;信任、安全、商业 SLA 和舰队运营验证滞后。

能力成熟度是基于公开证据的分析师分类,不是公司披露的 TRL 评分。

[CE021, CE023, CE024, CE028, CE029, CE039]

5.4 差异化、护城河和关键依赖

最强差异化在于融合两个通常分开的硬系统:同一自主船体上的波浪能发电和 AI 算力消耗。传统波浪项目通常尝试输出电力;传统数据中心在岸上解决土地、电网、冷却和人员约束;Project Natick 等水下数据中心类比验证密封海洋算力,但不验证自主波浪发电。如果 Panthalassa 的护城河最终出现,它将落在海洋工程诀窍、船体和压载设计、制造复制、远程运营软件、热集成,以及在能量密集波浪资源区运行的能力上。依赖地图同样苛刻。它依赖钢材制造和涂层、涡轮与电力电子可靠性、GPU 供给、密封容器热设计、Starlink 或可比 LEO 连接、海洋天气运营,以及许可或环境接受度。若干指标仍是愿景型:最高 90% 电力可用性、低成本能源,以及一座能达到 GW 级年节点产出的工厂。这些主张创造上行空间,但在成为耐久产品优势之前,需要带仪表的船队证明。[CE026, CE027, CE028, CE029, CE031, CE032]

FE003: 关键依赖图

商业就绪度取决于一条链:海事制造、可靠性、计算、卫星和信任控制都要跑通。

依赖按公开证据做定性排序,不来自内部项目计划。

[CE026, CE027, CE032, CE035, CE036, CE041]

5.5 信任、质量、安全和合规缺口

信任是产品故事里公开信息最薄的一环。一个漂浮 AI 数据中心必须在没有现场人员的情况下扛住腐蚀、生物污损、风暴、海水侵入、热循环、远程电力故障、网络中断和异常事故。技术参考来源把腐蚀和生物污损定义为普通海洋风险;New Scientist 则强调,数据中心事故中物理干预仍很常见,而 Panthalassa 的海洋站点让这种干预更难。密封、海水冷却的容器可能减少用水和冷却设备,但公开来源没有提供独立的 PUE、芯片温度、冷却回路、抗腐蚀涂层、防污、消防、风暴生存能力或平均服务间隔数据。安全和隐私同样未解:来源描述了卫星回传,但没有描述加密、租户隔离、密钥管理、事件响应或合规认证。环境信任也不完整,因为废热预计会散入海水,但邻近生态系统影响仍不清楚。质量控制表不应被读作今天已经通过的清单,而更像是 Ocean-3 和首批商业合同的尽调议程。[CE035, CE036, CE039, CE040, CE041, CE042]

信任 / 质量 / 合规表
控制项 / 质量指标状态范围缺口
腐蚀与涂层资质验证未公开披露钢制船体、涡轮机、舱体、紧固件需要材料规格、涂层寿命、检查计划、牺牲阳极或阳极保护策略。
生物污损管理未公开披露进水口、船体、换热表面需要防污方案、清洁计划和性能折减假设。
热管理与 PUE 验证概念上由海水冷却支撑密封算力模块和芯片运行需要实测 PUE、芯片温度、进水和故障数据。
远程运营可靠性仅有原型阶段证据自主运行、电力、网络和维修工作流需要 MTBF、MTTR、备件策略和异常事件预案。
数据安全与隐私未找到公开控制措施卫星回传和多租户算力需要加密、密钥管理、租户隔离、合规和事件响应。
环境热影响 / 海洋审查公开来源中影响不清废热、噪声、航行、海洋生物需要环境评估和监测结果。

各行是尽调控制项,不代表已通过认证;公开材料未披露正式鉴证或运营 SLA。

[CE011, CE035, CE036, CE039, CE040, CE041]

5.6 图表

Chapter 06

06客户

6.1 目标客户是需求侧代理,不是已拿下账户

Panthalassa 没有可按常规方式细分的公开客户基础。因此,正确的客户镜头是前瞻性的:需要高功率密度、可容忍延迟算力的超大规模云平台、大型 AI 基础设施买方、新云或商用 GPU 云运营商,以及 AI 实验室。本章分配的客户证明来源有意充当需求侧代理。Deloitte、IEA 相关数据中心需求证据和超大规模云参考来源显示,AI 工作负载正在挤压电力、冷却、土地和电网接入;它们并不显示 Panthalassa 已经把 AWS、Microsoft Azure、Google、某家新云或某个 AI 实验室转化为客户。隐含买方是基础设施或云采购,经济赞助人是试图锁定电力和算力容量的云或 AI 平台高管,用户则是愿意把批处理或推理任务送往海上的 AI 研究或产品团队。付款方可能购买已签约容量,而不是软件席位;但公开来源没有披露收入区间、价目表、渠道动作或已签容量承诺。[CU001, CU002, CU003, CU004, CU005, CU009]

客户分群表
客群买方 / 用户 / 付款方地域 / 垂直行业 / 规模渠道 / 用例收入区间或缺口
超大规模云平台基础设施采购付款;云容量团队购买;AI 服务团队使用全球超大规模;100 MW+ 数据中心级别;AWS/Azure/Google 作为代理样本海上承载可容忍延迟的 AI 推理或批处理任务无 Panthalassa 收入区间;仅为需求代理指标。
Neocloud / 商业 GPU 云运营商高管层基础设施采购方;GPU 云运营团队使用;容量转售方付款受电力和数据中心容量约束的大型 AI 云提供商批发海上算力容量或溢出容量未披露合同或分销渠道。
AI 实验室 / 前沿模型团队研究基础设施负责人购买;研究人员提交工作负载;实验室预算付款需要批处理、模拟或推理容量的大型 AI 模型开发者可接受卫星回传、周转需数小时至数天的任务未披露具名实验室试点或工作负载结果。
托管与数据中心运营商托管运营商购买容量;租户间接使用;运营商付款或转售受电力 / 制冷约束的多租户基础设施买方潜在容量合作,尚未显示为渠道未披露交易市场、交叉连接或托管合作。
硬件 / 基础设施合作伙伴合作伙伴采购和工程团队;不是终端客户Super Micro 等 AI/GPU 服务器供应商和投资方为节点供应 GPU 服务器或集成能力仅证明合作伙伴关系;不是客户收入。

Panthalassa 尚未披露已签约客户,因此分群只是前瞻判断;客户证明来源只提供需求代理, 不是购买证据。

[CU001, CU002, CU003, CU004, CU009, CU027]
FU001: 客户旅程图

潜在买方从 AI 电力痛点走向离岸计算扩张;客户验证在转化步骤仍缺失。

旅程根据目标买方工作流推断,因为 Panthalassa 未披露客户。

[CU001, CU005, CU013, CU032, CU041, CU044]

6.2 采用轨迹止于产品试点

公开采用轨迹是技术轨迹,而不是客户轨迹。Panthalassa 披露了 Ocean-1、Ocean-2 和 Wavehopper 等原型、2026 年 Ocean-3 试点系列,以及 2027 年商业部署目标。这些里程碑重要,因为可信的 Ocean-3 部署是任何设计伙伴对话的前提;但它们不是活跃使用、付费账户、重复购买、生产客户部署或利用率的证据。CBS 和公司描述了未来多个系统协同构成数据中心的情景,但当前公开记录仍缺少已签客户、LOI、客户试点、已部署客户地点或已签容量。因此漏斗异常陡峭:广泛需求可见,目标工作负载可以描述,产品试点也已规划,但公开证据中的每个商业证明步骤仍为零。尽管市场背景有吸引力,客户章节因此偏反向。[CU006, CU008, CU018, CU019, CU020, CU021]

客户增长 / 采用轨迹表
指标 / 里程碑数值或状态日期 / 时效性置信度含义 / 缺失分母
已签约客户2026-08-10 本轮证据公开信息没有已签约客户数;无法推断采用情况。
LOI / 客户试点2026-08-10 本轮证据未披露 LOI、设计合作伙伴或客户试点。
生产部署2026-08-10 本轮证据Ocean-3 是产品试点计划,不是生产级客户部署。
活跃账户 / 地点 / 利用率2026-08-10 本轮证据没有账户数、节点利用率或客户地点数据。
Ocean-1 / Ocean-2 / Wavehopper 原型技术海试2021 和 2024,另有 2025 报道仅为技术证据;没有客户工作负载证明。
Ocean-3 试点系列计划中的产品试点2026可能是采用前置条件;公开信息未绑定客户。
商业部署目标计划中的商业系统2027 目标未来目标;未披露已签约买方。

各行有意把技术里程碑和客户采用拆开;空值表示没有公开证据,不代表内部活动为零。

[CU006, CU008, CU018, CU019, CU020, CU021]
FU002: 采用 / 部署漏斗

需求信号广泛,但公开证据在已签客户、试点和收入处归零。

计数汇总公开证据类别,不代表内部 pipeline。

[CU007, CU008, CU018, CU019, CU021, CU024]

6.3 具名客户证明缺席

具名客户证明账本为空。表格刻意列出潜在设计伙伴细分,方便读者看到应该先测试谁;但每一行都标为目标,而不是已赢客户。这一区分是核心:AWS 和 Microsoft Azure 相关,是因为它们代表计算和电力需求巨大的超大规模云买方;托管和 AI 云运营商相关,是因为它们购买电力、冷却和高密度基础设施;AI 实验室相关,是因为部分工作负载可以容忍远离终端用户。这些事实都不是拿下具名客户、客户引述、试点公告、案例研究、采购记录或可背书部署。现有最佳证明是需求压力加产品市场逻辑;缺失的证明是一位客户明确表示 Panthalassa 以可衡量性能、可靠性和经济性运行了有用工作负载。在那之前,即使目标买方紧迫性可信,客户证明质量仍低。[CU007, CU008, CU022, CU024, CU026, CU030]

具名客户证明表
潜在客户 / 细分细分部署 / 用例生产还是试点结果 / 限制
Amazon Web Services(目标代理样本;未签约)超大规模云平台可容忍延迟的 AI 算力容量,或电力受限时的溢出容量未披露 Panthalassa 试点AWS 证据证明目标细分规模,不证明赢得客户。
Microsoft Azure(目标代理样本;未签约)超大规模云平台AI 平台容量、批量推理,或与可持续性挂钩的容量未披露 Panthalassa 试点Azure 证据证明目标细分相关性,不证明赢得客户。
大型 AI 实验室 / 前沿模型运营方(目标细分;未签约)AI 研究与产品组织长时模型任务、科学模拟,或低带宽推理未披露具名试点没有工作负载结果、正常运行时间、价格或客户推荐语。
Neocloud / 商业 GPU 云运营商(目标细分;未签约)AI 云基础设施买方批发海上 GPU 容量或容量转售未披露已签约设计合作伙伴没有渠道条款或采购承诺。

由于公开证据中没有已签约客户,这里的枚举只是目标设计合作伙伴类别样本。

[CU002, CU003, CU007, CU008, CU009, CU022]
FU003: 客户验证矩阵

市场需求证据质量强,但客户赢单、结果、留存和生产成熟度证据弱。

矩阵对证据类别做定性评分,因为没有客户案例研究。

[CU009, CU010, CU011, CU012, CU014, CU015]

6.4 留存、重复使用和满意度尚无法衡量

留存指标不应由市场需求倒填。没有披露客户,Panthalassa 也就没有公开 NRR、GRR、logo 流失、客户 cohort、续约期限、支持满意度评分或重复使用历史。正确表述是 null 指标加明确尽调问题。未来买方需要证据证明,海上节点能够连续数月产出可用算力、满足服务等级预期、在没有常规人工接近的情况下从故障中恢复,并接入采购和安全工作流。Uptime Institute 和 New Scientist 强调电力、网络、远程运营和时延都是可靠性约束,使留存门槛更高。实际中,第一次留存测试不会是续约 cohort;而会是 Ocean-3 或早期商业节点能否运行与买方相关的工作负载,时间长到足以证明第二次部署合理。在那之前,cohort 数字应刻意留空,而不是乐观填数。[CU014, CU015, CU023, CU024, CU031, CU035]

留存 / 重复使用 / 满意度表
指标数值 / 状态细分置信度尽调请求
净收入留存(NRR)所有细分首批商业部署后,索取按客户队列拆分的收入。
总收入留存(GRR)所有细分合同落地后,索取续约和流失记录。
客户流失 / 续约率所有细分索取客户名册、合同条款和续约结果。
合同期限 / 承诺容量超大规模 / AI 云目标客户索取已签署容量协议或 LOI。
客户满意度 / 推荐质量所有细分索取客户访谈和有记录的工作负载结果。
重复使用 / 利用率所有细分索取工作负载日志、节点可用性和利用率遥测。

Panthalassa 没有披露客户,因此留存指标为空;本表把每个缺失指标转成尽调请求。

[CU014, CU015, CU021, CU023, CU024, CU031]
FU004: 留存 / 重复队列

公开层面还没有客户队列,因此留存单元格留空,等待签约客户和使用遥测数据。

零值单元格表示公开留存证据为零,不是观察到客户流失;NRR、GRR、续约或重复使用数据均未公开。

[CU021, CU023, CU024, CU031]

6.5 扩张上行伴随极端集中度和采购风险

如果 Panthalassa 跑通,扩张可能很强:同一买方可把节点或船队作为容量产品继续追加,需求侧证据也显示超大规模云和 AI 云正在寻找受电力约束的算力。商业风险在于,这种上行会把未来收入集中在少数极成熟买方手中;这些买方尽调周期长、可靠性门槛高、会做安全审查,并且对价格有议价力。相比普通云容量,采购摩擦更高,因为基础设施新颖、远程、依赖卫星链路,并暴露在海洋风险下。Super Micro Computer 不应被误读为客户;它披露身份是投资者,且拥有相关 AI / GPU 服务器产品,因此可能是供应商或合作伙伴依赖。章节基础结论不是需求缺席,而是需求尚未转化为采用。下一道尽调门是已签设计伙伴证据,并且要绑定工作负载类型、正常运行时间目标、数据路径、经济性和扩张权。[CU016, CU017, CU025, CU026, CU030, CU031]

扩张与集中度风险表
扩张驱动因素集中度风险影响尽调路径
试点成功后扩张节点或船队容量高度依赖少数超大规模买方单一买方可能主导早期收入和条款索取目标账户管线、买方集中度上限和排他条款。
可容忍延迟的推理或批处理工作负载适配可服务工作负载可能窄于整体 AI 需求若延迟敏感应用主导需求,转化会受限用具名用户验证卫星回传下的工作负载基准。
硬件供应和集成伙伴依赖供应商或投资方伙伴,尤其是 AI/GPU 服务器延误或短缺可能卡住客户交付确认 Super Micro 角色、供应协议和备选供应商。
目标买方面临电网和土地约束采购团队可能仍偏好已验证的陆上园区新型海上风险会拉长安全、保险和可靠性审查向设计合作伙伴索取采购标准和风险签核。
未来分销商或类托管渠道没有公开渠道或市场证明可能被迫面向专业买方做缓慢的企业直销索取渠道计划、伙伴合同和容量转售权。
海洋耐久性和远程运营可靠性故障会在扩张前伤害续约对服务级别的怀疑可能阻碍多节点订单要求独立的 Ocean-3 正常运行时间、维护和事故数据。

签约合同出现前,扩张上行仍是假设;集中度风险是结构性的,因为相关买方少且强势。

[CU016, CU017, CU025, CU026, CU030, CU031]

6.6 图表

Chapter 07

07风险

7.1 严重性排序和剩余暴露

Panthalassa 的风险栈更偏向技术和运营生存能力,而不是普通软件执行。最高级别暴露在于:一座 85 米钢制、自主、承载算力的节点,能否在开阔海域扛住腐蚀、生物污损、波浪载荷、风暴损伤和定点保持要求,同时维持电力、冷却、网络和 GPU 可靠性。公司有实质设计缓释:简单钢结构、少量外露运动部件、无海底输出电缆、无锚、无发动机,并用分阶段原型走向 Ocean-3。这些缓释真实存在,但还不成熟,因为投资逻辑依赖承载算力的节点穿越数月海况,而不只是此前的能源试验。实际含义是,公司应按试点闸门约束的前沿基础设施赌注来投资判断。若 Ocean-3 在生存能力或维护画像上失败,会直接传导为更低利用率、更高维护船成本、收入延迟和明显更弱的估值逻辑。[CR001, CR002, CR003, CR004, CR005, CR006]

运营 / 质量 / 安全风险登记表
排名失效模式可能性影响缓释成熟度残余风险敞口未解决缺口
1腐蚀、生物污损、风暴或海浪损伤削弱节点输出或生存能力严重低至中在 Ocean-3 扛过季节性海况前,这是最高残余风险敞口。6、12、24 个月独立检查数据。
2无人值守海上数据中心断电或网络中断海上环境里,正常运行时间事故的根因追查异常困难。事件响应预案和恢复遥测。
3Starlink 延迟或带宽把工作负载限制在有限推理或批处理任务若低延迟 AI 主导需求,可服务市场会收窄。实测延迟、带宽和客户工作负载适配验证。
4GPU、密封集装箱、制冷或海上载荷更换失败海上服务介入可能吃掉经济性。载荷更换计划和服务船成本曲线。
5废热或排放影响引发生态或安全异议海洋生态影响仍不清楚。环境基线、热羽流建模和监测计划。
6海上 AI 工作负载的网络安全和隐私控制披露不足公开来源不能证明完整安全体系。安全架构、认证、DPA 和事件桌面演练。

运营风险聚焦几类失效模式:海洋条件、无人值守运营、算力载荷,以及安全披露缺口。

[CR001, CR002, CR003, CR004, CR008, CR009]
FR001: 风险热力图

最高残余风险集中在高影响的技术、运营、通信、监管和单位经济性单元。

定性评分使用章节风险登记表,而不是数值概率模型。

[CR001, CR009, CR012, CR015, CR019, CR031]

7.2 监管、法律、环境和政策暴露

监管风险不是一张单一许可,而是一套彼此叠加的海上能源、海事、数据中心、环境和政策体系。BOEM 监管美国外大陆架上的可再生能源,DOE 的海洋能源项目则把波浪能定义为仍需研发、示范和降低障碍的行业。数据中心法律来源又叠加一层:选址和设计不同,NEPA、ESA、Clean Water Act、Clean Air Act、水、废弃物、雨水、分区、噪音、交通和公共土地审批都可能相关。符合条件的数据中心若获得联邦加速,可能有利于部分项目,但法律评论强调,州和地方审查、社区反对与诉讼仍是进度和成本风险。若美国境内拖带、维护或备件运输发生在美国港口之间,Jones Act 还会形成实际海事约束。环境审查尤其敏感,因为 New Scientist 报道称,海上废热对生态系统的影响仍不明确。[CR014, CR015, CR016, CR017, CR018, CR019]

监管 / 法律风险登记表
排名规则 / 事项辖区或领域可能性影响缓释成熟度残余风险敞口尽调路径
1OCS 海上能源审批和政策波动BOEM / 美国海上能源BOEM 暂停和租赁审查显示海上能源政策敏感。将部署海域逐一映射到 BOEM、Coast Guard、州级和国际水域主管机构要求。
2数据中心和海洋影响的环境审查联邦、州和地方环境法许可可能涉及 NEPA、ESA、CWA、CAA、水、废弃物、分区、噪声和公众反对。为 Ocean-3、美国商业节点和国际部署制作许可矩阵。
3Jones Act 和海事沿海运输约束美国国内海上运输美国港口之间的拖带、维护和备件运输,可能需要符合 Jones Act 的船舶和船员。与海事律师确认船旗、建造地、船东、配员和沿岸运输假设。
4数据中心诉讼和社区挑战风险州和地方审批即便有联邦加速,大负荷数据中心仍会面对诉讼和地方阻力。投入项目资本开支前,跟踪社区沟通、环境记录和诉讼卡点。
5IP、隐私、网络安全和事件响应披露缺口商业与技术法律留存来源中看不到公开专利、FTO、隐私、安全或事件响应材料包。索取专利清单、FTO 备忘录、DPA / 安全架构和事件响应计划。

本表是基于已分配监管和法律来源的部分风险登记;各行按严重性排序,但不是穷尽式许可意见。

[CR014, CR015, CR016, CR017, CR018, CR019]
FR002: 风险传导图

技术、监管和依赖风险会传导到正常运行时间、客户转化、capex、融资和估值。

不同视角:展示风险如何因果传导至经济性和估值,而不是依赖清单。

[CR008, CR009, CR012, CR018, CR019, CR029]

7.3 合作伙伴、平台、供应商与客户依赖

合作伙伴依赖集中在几处可能变成二元卡点的环节。通信依赖 Starlink 或类似 LEO 回传,现有证据显示,延迟和带宽会决定工作负载组合,因此近期产品更适合可容忍延迟的推理或批处理任务,而不是低延迟消费者应用。硬件敞口同样重要:Panthalassa 需要合格 GPU、服务器、电力电子设备、密封计算容器,并且要在海洋环境里可靠更换载荷。Super Micro 被列为投资方,但本次留存来源并未把这一点转化为已承诺的供应协议。公司还依赖监管方批准或容忍部署海域,依赖资本方在收入前资助制造,依赖未来超大规模云厂商或 AI 实验室客户把试点转成收入。由于公司没有披露付费客户,如果试点成功,早期收入可能集中在一两个大买方。[CR011, CR012, CR013, CR025, CR026, CR027]

合作伙伴 / 依赖风险登记表
排名依赖项交易对手 / 角色集中度失效场景严重性缓释残余风险敞口
1卫星通信Starlink / LEO 回传公开单平台依赖延迟、带宽、中断、定价或政策限制会削弱工作负载交付。只认证可容忍延迟的工作负载,并开发备用连接方案。
2GPU 和服务器供应Nvidia 级 GPU、Super Micro 风格硬件战略投入高度集中硬件短缺、热适配验证失败或载荷更换延迟会拖慢部署。锁定有约束力的配额,并完成海事适配测试。中高
3资本提供方Thiel 领投财团和未来项目融资收入前高度集中试点延误会迫使公司接受稀释性融资,或暂停制造扩产。按 Ocean-3 里程碑分阶段安排资本开支,并维持财团储备。中高
4监管机构和海事服务商BOEM、Coast Guard、船东、港口物流许可或 Jones Act 船舶约束会延迟部署和服务。中高按部署地域预先确认船舶资质和许可要求矩阵。
5未来超大规模或 AI 实验室买方未披露目标客户前瞻性集中试点后,首笔收入依赖一两个锚定客户。中高把试点转成多元化包销或算力容量协议。

由于没有披露客户,依赖分析把交易对手视为风险传导方,而不是已验证商业合同。

[CR011, CR012, CR013, CR025, CR026, CR027]
FR003: 依赖图

扩张前,Panthalassa 依赖一组窄口径的通信、硬件、监管、船舶、资本和客户交易对手。

不同视角:交易对手和资源依赖;与风险传导 DAG 的任何重叠是有意的,但不是重复分析。

[CR011, CR025, CR026, CR027, CR028, CR037]

7.4 财务、模型与单位经济风险

财务风险与海上生存能力紧紧绑在一起。公开融资事实很强:Panthalassa 2026 年 5 月宣布 $140 million Series B,累计融资约 $210 million。但这笔资金要在公司披露收入或付费客户之前,完成试点制造、部署 Ocean-3、证明海上推理可行,并降低制造和维护的不确定性。管理层和投资方材料给出诱人目标,包括约 $1,500/kW 的资本开支、高容量因子、低成本电力,以及可扩到约 1 GW/年的工厂。反方观点是,腐蚀、生物污损、保险、海上维护、船舶天数和 GPU 服务物流都可能抹掉模型收益。即便波浪能在物理上很丰富,投资判断要看的也不是总波浪资源,而是扣除寄生负载和服务干预后的交付算力成本。[CR027, CR028, CR029, CR030, CR031, CR032]

缓释措施与叫停标准表
风险监测指标阈值 / 事件行动含义
Ocean-3 生存性独立的运行时间、检查和按天气归一化的性能日志试点无法在代表性海况中持续安全运行不要资助商业船队;退回工程验证阶段。
腐蚀 / 生物污损退化检查报告、冷却温差、阻力 / 定位功耗、涂层状况材料退化或污损需要频繁上船干预按更高 O&M 重估经济性,或停止扩大部署。
电力与联网可靠性电力可用性、Starlink 运行时间、延迟和故障恢复中断超过客户 SLA 容忍度,或需要人工离岸干预仅限非关键工作负载,或暂停客户承诺。
客户转化已签 LOI、付费试点、利用率和工作负载匹配度Ocean-3 验证窗口结束后仍无锚定客户或付费试点将估值视为缺乏支撑;下一轮前要求拿到战略客户。
监管 / 法律阻滞许可矩阵、机构反馈、Jones Act 船舶方案、诉讼状态关键审批路径受阻,或服务物流不可行迁移部署、重做运营设计,或停止美国扩张计划。
单位经济性计入资本开支、维护、保险、卫星和 GPU 更新后的交付算力成本试点数据出来后,成本仍远高于目标,或波浪能 LCOE 差距仍在没有新的经济性证据,不要按 $1bn 级估值投资。

叫停标准把主要剩余风险转成可监测阈值,便于投资委员会分阶段决策。

[CR007, CR028, CR029, CR030, CR031, CR033]

7.5 团队、执行缓释与击穿论点的触发项

执行风险有部分缓释:团队来自航空航天、船舶设计、软件、制造、军方和研究机构,投资财团也够深,能资助更多迭代。风险仍然实质存在,因为公开来源把公司重心放在两位联合创始人身上,治理或董事会披露有限,许多具体运营控制仍不透明。Panthalassa 在 New Scientist 发表前也没有回应其带怀疑色彩的技术问题,因此独立尽调比媒体层面的背书更有价值。监控计划应当明确:要求独立 Ocean-3 遥测、扣除寄生负载后的净输出、Starlink 可用性和延迟日志、腐蚀试片或检查报告、生物污损与冷却衰减数据、每个节点的服务船天数、安全事故、许可里程碑、签约客户转化,以及更新后的单位经济桥接。若试点无法达到生存能力、服务节奏、客户转化、监管许可或成本目标,投资论点就会失效。[CR034, CR035, CR036, CR037, CR038, CR039]

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尽调路径
1创始人 / 技术领导层两位具名联合创始人支撑战略和技术可信度。专业团队和深厚的投资者财团。审查接班计划、技术决策权和董事会监督。
2海事运营组织从原型试验扩大到船队服务,尚无证据支撑。Ocean-1/Ocean-2 经验和 Ocean-3 试点计划。核查招聘计划、安全体系、服务流程和船舶合同。
3治理与披露董事会、控制、安全和事件流程披露有限。投资者监督很可能存在,但公开证据不足。索取董事会名单、委员会、保险、控制和事件治理材料。
4外部技术回应公司在发布前未回应一项质疑性询问。管理层在其他场合给过正面表述。要求就腐蚀、生物污损、Starlink 和维护问题给出书面回应。

人员风险的重点是执行能力和治理透明度,而不只是创始人履历。

[CR034, CR035, CR036, CR037, CR039]

7.6 展项

Chapter 08

08估值

8.1 投资建议与估值立场

投资结论是继续研究 / 跟踪,而不是买入。Panthalassa 具备风投式期权的形态,押注的是一个真实瓶颈:AI 算力受电力约束,如果海上波浪供能基础设施能绕开电网和土地瓶颈,就可能有价值。当前公开记录不足以支撑以隐含近独角兽估值投入新资金。公司有一轮 Peter Thiel 领投的 $140 million Series B,累计融资约 $210 million,但准确投后估值、优先权和完全摊薄所有权没有披露。公司也仍处于收入前,没有公开披露客户,也尚未展示独立的商业规模 Ocean-3 表现。即便上行期权有意义,入场价格纪律也难以判断,甚至可能已经偏高;风险评级为高。[CV001, CV002, CV003, CV009, CV034, CV035]

建议摘要表
决策项当前结论证据基础投资含义
建议继续研究 / 跟踪融资阵容和规模很强,但估值条款和商业化证据缺位。试点和客户证据改善前,不要按买入定价。
置信度低至中融资事实已获交叉验证;经济性和客户尚未验证。用里程碑闸门替代明确目标价。
风险评级海上生存性、延迟、单位经济性、客户和资本开支仍未解决。要求下行保护,或等证据出现。
估值立场未知至偏高已有接近 $1B 的说法,但投后估值和条款不明。仅凭公开证据,无法判断入场价格纪律。
目标回报 / 持有 / 退出仅持有 / 跟踪IPO 或战略退出仍需数年,并取决于收入证据。Ocean-3 和管线尽调后再评估。

建议对证据和价格高度敏感;公开估值条款缺失,无法精确计算持股或回报。

[CV001, CV002, CV003, CV009, CV034, CV035]
FV001: 建议逻辑

决策链把强期权价值转化为继续研究立场,因为验证和定价仍然缺失。

流程展示投资逻辑,不是机械评分模型。

[CV001, CV002, CV003, CV009, CV016, CV034]

8.2 正向论点与反向论点

核心论点不是泛泛的波浪能故事,而是共址逻辑。Panthalassa 想在海上用电、出售算力,避开拖慢陆地数据中心的输电、土地、冷却和电网并网问题。如果公司真能交付低成本、高可用能源,并把它封装成有用的推理容量,就可能在电网之外创造基础设施价值。反向论点也很直接。波浪能市场很小,经济性难做;海洋会腐蚀设备,维护成本高;LEO 延迟会限制工作负载;陆上 AI 云巨头在规模、采购、软件和客户上都有优势。投资方支持的经济模型不是客户证明,论点要先拿到现场数据和管线证据,才会变成可定价的投资案。[CV010, CV013, CV014, CV015, CV016, CV018]

论点 / 反论点表
维度论点反论点改变判断的证据
市场AI 电力需求制造了大型算力选址瓶颈。波浪能市场很小、也很早期。独立客户对可容忍延迟的离岸推理有需求。
产品海上共址绕开土地、电网和用水约束。海上维护和 Starlink 延迟可能限制可用工作负载。Ocean-3 遥测显示算力和联网可靠。
财务$140M Series B 轮为制造和首批部署供血。公司尚无收入、资本开支重,且未披露客户合同。可信 AI 买家签下收入合同、包销协议或 LOI。
单位经济性$0.02/kWh、$1,500/kW 和高可用性可能形成成本优势。目标未经独立证明,且可能漏掉维护、保险和物流。经审计的船队级成本、运行时间和维护数据。
竞争尚无直接同行把自主波浪发电和算力结合起来。陆上 AI 云既有玩家和其他前沿 DC 概念生态更强。清晰的工作负载细分场景和性价比证据。
风险投资者财团能为迭代供血。腐蚀、风暴、生物污损和远程维修可能吞掉回报。穿越恶劣海况事件的生存记录。

论点综合市场、产品、财务、客户、竞争和风险证据;所有判断变化都应由尽调闸门触发。

[CV010, CV013, CV014, CV015, CV016, CV018]
FV004: 投资 KPI

KPI 指标组给市场拉动和投资人质量打高分,但验证、客户和估值证据得分低。

分数是基于所引证据汇总出的投委会定性评级,不是经审计的 KPI。

[CV009, CV011, CV015, CV016, CV020, CV025]

8.3 融资背景与入场纪律

融资背景要求纪律,而不是虚假的精确。公开文章给出头条轮次和接近 $1 billion 的叙事,SEC Form D 文件显示规模较小的相关 SPV 配额;没有任何来源披露完整股权结构、投后股数、清算优先权、参与权、按比例认购权或期权池处理。方法上,只有已知投入资本和完全摊薄资本结构时,投后估值才有意义。就 Panthalassa 而言,公开证据无法验证准确价格。理性的入场立场应当由里程碑设闸:把当前估值视为押注艰难基础设施突破的期权价格,而不是经验证的 DCF 或收入倍数。下一轮融资应在 Ocean-3 数据和可信客户承诺压缩主要不确定区间之后再评估。[CV003, CV004, CV005, CV006, CV007, CV008]

FV002: 估值敏感性

估值逻辑对试点成效、单位经济、客户验证和未披露条款最敏感。

数值是投委会评分中方向性的增减,量表为 -3 到 +3,不代表估值金额。

[CV003, CV007, CV009, CV015, CV016, CV020]

8.4 牛市、基准与熊市情景

情景分析给出的区间很宽,因为一个技术证明点就可能改变公司的风险类别。牛市情景下,Panthalassa 展示有用的海上推理,证明生存能力,并表明其 $0.02/kWh 和高可用目标经得住真实维护、保险和物流成本。那可能支撑数十亿美元级基础设施期权价值。基准情景下,试点跑通,但规模化缓慢且吃资本,因此隐含约 $1 billion 的估值只能维持,投资人等待订单和可重复制造。熊市情景下,腐蚀、生物污损、风暴、延迟或波浪能 LCOE 让系统不具经济性,陆基替代方案胜出。下行触发因素不是想象力不够,而是现场经济性无法足够快地收敛。[CV013, CV014, CV015, CV016, CV018, CV020]

牛 / 基准 / 熊情景表
情景假设估值 / 回报逻辑概率信号下行触发因素
Ocean-3 验证廉价可靠算力;客户签约;制造放量。如果 AI 电力需求愿意为离岸容量付费,公司就是数十亿美元级基础设施期权。重复部署、运行时间和已签约需求。若经济性和合同持续叠加,则无触发。
基准试点跑通,但制造和客户转化偏慢。在投资者等待规模化证据时,约 $1B 标记得以维持。遥测数据正面,但收入转化有限。下一轮持平,或带结构化条款。
生存性、延迟、LCOE 或维护经济性失效。叙事虽强,仍可能降价融资、困境出售或减记。试点错过、没有客户,或成本栈高于目标。避免追加新钱;可行时通过二级市场退出。

区间是情景带,不是目标价;取决于未披露条款和未经证明的技术里程碑。

[CV015, CV016, CV018, CV020, CV037, CV038]
FV003: 估值 / 回报区间

由于验证成败近乎二元、资本强度高,情景价值从可能归零到数十亿美元级期权都有。

USD 百万美元区间只是基于明确假设的示意性情景带;具体持股比例和投后估值未披露。

[CV003, CV015, CV037, CV038, CV039]

8.5 可比估值背景

可比集合只有在清楚标注不完美时才有用。Starcloud 是最接近的叙事类比,因为它在推进非地面数据中心时,据报道估值达到 $1.1 billion,但太空和海洋的执行风险不同。CoreWeave 可作为需求侧 AI 云参照,但它已经上市,成熟度远高于 Panthalassa。Crusoe 是 AI 基础设施类比,不过本次材料中关于其准确估值标记的公开证据较薄。CorPower Ocean 和 Eco Wave Power 是相关波浪能同业,但它们出售的是并网或沿海波浪电力,而不是海上算力。这些参照能三角定位期权价值和风险偏好;它们不能证明 Panthalassa 的价格。最诚实的结论是,前沿基础设施市场可以很早给出独角兽标记,但商业证明决定这些标记会复利还是坍塌。[CV028, CV029, CV030, CV031, CV032, CV033]

可比估值表
可比对象指标 / 估值参考为什么可比局限
Panthalassa约 $1B 说法;$140M Series B 轮;累计融资约 $210M标的公司,也是直接定价语境。确切投后估值和优先权未披露。
Starcloud据报融资 $170M、估值 $1.1B前沿非地面数据中心类比。太空数据中心不是海洋波浪算力。
CoreWeave上市 AI 云基础设施公司体现 AI 算力容量的需求侧价值。成熟上市公司,不是尚无收入的硬件基础设施。
CrusoeAI 基础设施私营公司类比;FACTS 包引用 2023 年约 $4.7B 标记可比的能源加算力叙事。用于精确核验估值的指定公开来源较薄。
CorPower Ocean拥有 10-30MW CorPack 阵列的波浪能技术公司波浪能执行和融资同行。电网供电设备,不是板载 AI 算力。
Eco Wave PowerNasdaq 交易的 WAVE 上市微盘波浪能参照公开波浪能市场情绪参照。沿海 / 岸上防波堤模式,不是自主远海算力。

可比集合不完整且刻意异质,因为公开市场没有纯粹的自主波浪能数据中心同行。

[CV001, CV002, CV003, CV028, CV029, CV030]

8.6 退出准备度与最终尽调问题

Panthalassa 今天还没有达到退出准备度。可信 IPO 需要收入、客户集中度披露、可重复单位经济、治理准备度,以及能把海洋算力从科学项目转成基础设施平台的公开市场叙事。战略收购只有在买方能够尽调现场表现和工作负载匹配后才合理。眼下的尽调问题因此很具体:确认投后估值和条款,检查 Ocean-3 遥测,验证客户管线,把能源经济性与维护和保险对齐,并审阅风暴和盐水暴露下的生存能力数据。任何入场前都应明确击穿论点的触发项:Ocean-3 部署失败、没有客户承诺、经济性显著高于目标、融资条款偏重优先权,或证据显示延迟和维护限制可服务的工作负载池。[CV024, CV025, CV026, CV027, CV034, CV035]

论点破裂与叫停触发因素表
触发因素阈值 / 事件传导到论点行动含义
Ocean-3 失手没有经过验证的部署,或反复出现重大延误融资需求出现前,试点证据未到位。暂停或避免入场。
经济性失手交付成本显著高于 $0.02/kWh,或资本开支高于目标相对陆上电力的成本优势消失。大幅下调定价。
生存性失败风暴、腐蚀、生物污损或维修事件造成长时间中断船队运行时间和保险假设失效。视为论点破裂。
客户缺口AI 算力买家没有可信 LOI、合同或试点尚无收入公司的估值缺少需求证明。保持仅研究立场。
结构化融资下一轮条款偏重优先权,或为降价融资此前估值标记站不住。除非条款保护下行,否则避免参与。
延迟 / 工作负载错配只有低价值工作负载适合卫星回传收入池小于论点预期。下调牛情景区间。

叫停触发因素设计成可监测尽调闸门,而不是泛泛风险。

[CV009, CV015, CV020, CV021, CV024, CV035]
最终尽调要求表
主题缺失证据重要性尽调路径
投后估值与条款确切投后估值、优先权、期权池、按比例跟投权和持股比例。决定入场价格、稀释和目标回报。索取条款清单和股权结构表。
试点结果Ocean-3 电力、运行时间、算力、冷却、延迟和自主运行数据。把原型叙事转成可支撑投资判断的证据。审查遥测、日志和第三方测试报告。
客户管线LOI、付费试点、工作负载画像和定价。尚无收入的公司需要需求证明。访谈管线客户并做客户尽调。
单位经济性每 kW 资本开支、维护、保险、船舶成本、利用率和毛利率。决定 $0.02/kWh 能否支撑算力经济性。用供应商报价和海试数据搭建自下而上的船队模型。
生存性风暴、腐蚀、生物污损、盐水侵入和维修历史。海上可靠性是核心熊情景。检查测试记录和独立海洋工程评审。
退出路径战略买家兴趣、IPO 准备度、治理和资本计划。界定持有期和融资依赖。访谈潜在买家,并审查董事会 / 治理材料。

尽调要求针对尚未解决的证据缺口,应在任何买入建议前完成。

[CV004, CV007, CV009, CV015, CV020, CV024]

8.7 展项

免责声明

本报告是基于公开证据的尽调快照,不构成投资建议。重要的财务、法律、技术和合同事实仍未公开;任何投资决定前,都应直接向管理层和原始文件核验。

证据索引

结论
编号陈述可信度来源
CO001 Panthalassa is a Portland, Oregon company building autonomous floating platforms, called nodes, that convert ocean wave energy into electricity to run AI data centers at sea. SO001, SO002
CO002 Panthalassa is organized as a public benefit corporation focused on clean ocean-powered compute. SO002, SO023
CO003 Multiple 2026 sources place Panthalassa headquarters in Portland, Oregon, with prototype and testing activity in the Pacific Northwest. SO001, SO004, SO011
CO004 Panthalassa was founded in 2016, making it a roughly decade-old venture by its 2026 Series B. SO002, SO006
CO005 Panthalassa positions its business model as generating power at sea and selling AI compute rather than transmitting electricity to shore. SO001, SO005
CO006 Garth Sheldon-Coulson is identified as Panthalassa co-founder and chief executive officer. SO001, SO002
CO007 Garth Sheldon-Coulson previously worked as a senior investment associate and AI researcher at Bridgewater Associates before founding Panthalassa. SO001, SO006
CO008 Brian Moffat is identified as a Panthalassa co-founder and chief innovation officer with a background in ocean-energy research. SO006, SO007
CO009 Brian Moffat previously worked on wave energy at Spindrift Energy and holds three bachelor of science degrees from UC Irvine. SO006, SO008
CO010 Panthalassa reported roughly 120 employees around its May 2026 Series B. SO001, SO007
CO011 Panthalassa had grown from about 70 employees in early 2024 to roughly 120 by 2026, indicating steady scaling. SO001, SO011
CO012 Panthalassa announced a $140 million Series B financing on May 4, 2026. SO001, SO003
CO013 The Panthalassa Series B was led by Peter Thiel. SO001, SO005
CO014 Peter Thiel framed his investment by saying Panthalassa has opened the ocean frontier for compute beyond current imagination. SO003, SO001
CO015 Panthalassa disclosed approximately $210 million in total capital raised as of its 2026 Series B. SO001, SO007
CO016 Financial Times characterized Panthalassa as a "$1bn ocean data centre start-up," implying a near-unicorn valuation. SO014, SO015
CO017 Panthalassa does not publicly disclose an exact post-money Series B valuation in retained sources. SO001, SO014
CO018 Returning Panthalassa investors include Founders Fund, Gigascale Capital, Lowercarbon Capital, Unless, and WovenEarth. SO003, SO009
CO019 New Panthalassa Series B investors include John Doerr, Marc Benioff’s TIME Ventures, and Max Levchin’s SciFi Ventures. SO003, SO001
CO020 Additional new backers named include Susquehanna Sustainable Investments, Hanwha Group, Fortescue Ventures, Super Micro Computer, and Sozo Ventures. SO003, SO010
CO021 Local Oregon investors Portland Seed Fund and the Intrepid Oregon Fund are named among Panthalassa backers. SO003, SO018
CO022 A special-purpose vehicle related to the Panthalassa raise filed an SEC Form D on March 25, 2026 reporting $319,000 sold to 14 investors. SO025, SO021
CO023 A later related special-purpose vehicle filing reported roughly $1.94 million raised from 31 investors as of July 2026. SO025, SO021
CO024 Panthalassa deployed an early prototype, Ocean-1, in 2021 in the Strait of Juan de Fuca. SO011, SO004
CO025 Panthalassa tested a later prototype generating on the order of 50 kilowatts in Puget Sound in 2025. SO011, SO012
CO026 Panthalassa plans to deploy its Ocean-3 pilot node in the northern Pacific around August 2026. SO001, SO007
CO027 Panthalassa targets commercial deployment beginning around 2027. SO001, SO006
CO028 Panthalassa describes an eventual vision of thousands of autonomous nodes operating across open ocean. SO005, SO017
CO029 Panthalassa is pre-revenue with no publicly disclosed paying customers as of its 2026 Series B. SO001, SO024
CO030 Independent commentators caution that ocean environments are corrosive and mechanically harsh, raising execution risk for at-sea data centers. SO024, SO025
CO031 Panthalassa did not respond to at least one journalist request for comment on skeptical technical questions. SO024
CO032 Panthalassa frames its mission around abundant clean power and compute produced at sea as a strategic national asset. SO003, SO008
CO033 Panthalassa previously explored producing hydrogen or clean fuels before pivoting its emphasis toward AI compute. SO011, SO006
CO034 Panthalassa retains its own homepage presenting its ocean-compute mission and node concept. SO016
CO035 Coverage of the raise spans mainstream technology, climate, and business outlets, indicating broad 2026 media attention. SO001, SO007, SO008
CO036 John Doerr publicly praised the Panthalassa investment as strengthening American technological leadership. SO003, SO019
CO037 Panthalassa’s founding location and Pacific Northwest roots align with regional wave-energy and maritime testing infrastructure. SO011, SO017
CO038 Panthalassa’s public disclosure leaves valuation, revenue, and board composition largely unspecified in retained sources. SO001, SO014
CO039 Panthalassa’s leadership emphasizes engineering and ocean-energy expertise over a large disclosed executive bench. SO006, SO007
CO040 Panthalassa’s Series B is materially larger than its estimated prior lifetime funding, reflecting a step-change in capital. SO001, SO007
CM001 Panthalassa's relevant market is offshore-sited AI compute powered by ocean wave energy, not grid-scale wave electricity sold to shore. SM016, SM017, SM024
CM002 The market boundary includes onboard wave generation, offshore siting, seawater-cooled compute payloads, satellite or marine backhaul, and sale of AI compute capacity. SM016, SM019, SM021
CM003 The market boundary excludes conventional electricity transmission to shore because Panthalassa says electricity is consumed onboard by AI compute rather than exported. SM017, SM016
CM004 Status-quo substitutes for the same buyer budget include land data centers, colocation, hyperscale cloud regions, and grid-connected AI clusters. SM012, SM014, SM006
CM005 Adjacent infrastructure substitutes include offshore wind, floating solar, nuclear or small modular reactor power for data centers, and ordinary grid power procurement. SM008, SM013, SM021
CM006 The IEA projects global data-center electricity demand could reach about 945 TWh per year by 2030, exceeding Japan's current total electricity use. SM007, SM006
CM007 AI data-center demand creates a large top-down TAM lens for Panthalassa because the company sells compute capacity rather than wave-electricity equipment. SM007, SM006, SM011
CM008 SemiAnalysis frames data-center energy availability as a material constraint in the AI infrastructure race. SM006, SM007
CM009 Wikipedia's AI boom and cloud-computing context supports treating AI labs, cloud platforms, and neocloud buyers as demand-side customers rather than energy utilities. SM011, SM014
CM010 Mordor Intelligence estimates the wave-energy market at roughly 10 MW installed in 2026 growing to about 125 MW by 2031. SM001, SM002
CM011 Mordor's 2026-to-2031 wave-energy forecast implies a very high growth rate, reported around 65.7% CAGR, from a tiny installed base. SM001, SM005
CM012 Other analyst sources also describe wave energy as a fast-growing but early and forecast-heavy market through the early 2030s. SM002, SM004, SM003, SM005
CM013 Marine and wave energy remain small relative to mainstream renewable electricity markets and therefore cannot alone support a broad data-center TAM. SM001, SM009, SM013
CM014 Wave and marine energy LCOE estimates around $388 to $618 per MWh imply costs roughly three to six times solar in the current evidence base. SM001, SM008
CM015 EIA expects wind and solar to lead near-term U.S. power generation growth, reinforcing that mainstream renewable alternatives are more mature than wave power. SM008, SM013
CM016 Panthalassa sits at the intersection of two asymmetric markets: huge AI power demand and tiny commercial wave-energy deployment. SM007, SM001, SM016
CM017 The evidence-constrained SAM is wave-powered offshore compute, a narrower category than total AI compute or total wave-energy hardware. SM016, SM019, SM001
CM018 The evidence-constrained SOM for 2026-2027 is pilot and early commercial nodes rather than a fleet-scale share of the global AI data-center market. SM016, SM018, SM023
CM019 Panthalassa's reported Ocean-3 pilot timing around 2026 and commercial target around 2027 make early-node adoption the near-term sizing unit. SM016, SM019, SM023
CM020 Hyperscalers such as AWS, Azure, and Google are natural buyer archetypes because they control cloud-infrastructure budgets and need energy-backed compute capacity. SM014, SM012, SM006
CM021 Neocloud and AI-compute specialists such as CoreWeave-like and Crusoe-like buyers are plausible customers for batch or flexible compute capacity. SM014, SM006, SM015
CM022 AI labs are a plausible user segment when workloads can tolerate offshore networking limits and value incremental compute more than low-latency placement. SM011, SM021, SM024
CM023 Budget ownership for the target market sits primarily with cloud infrastructure, data-center capacity, AI infrastructure, and strategic energy procurement functions. SM014, SM012, SM006
CM024 Adoption is likely to start with pilots, non-mission-critical batch jobs, and capacity reservations before buyers trust offshore nodes for production workloads. SM024, SM021, SM019
CM025 Starlink-style satellite backhaul makes Panthalassa more credible for batch and inference workloads than for low-latency interactive AI services. SM024, SM021
CM026 A primary growth driver is the power scarcity and interconnection bottleneck facing land-based AI data centers. SM007, SM006, SM016
CM027 A second growth driver is land and permitting scarcity for conventional data centers near cheap power and network capacity. SM012, SM006, SM021
CM028 Decarbonization pressure and energy-sovereignty narratives strengthen interest in clean domestic compute supply chains. SM017, SM013, SM025
CM029 Panthalassa's partner materials cite aspirational economics such as very low power cost, high capacity factor, and factory-scale node production, but those figures are not independently validated. SM020, SM025
CM030 The largest adoption constraint is that commercial marine wave-energy systems remain unproven at data-center reliability and scale. SM001, SM009, SM010
CM031 High marine capital intensity is a constraint because Panthalassa must finance steel structures, offshore operations, and GPU payloads before revenue. SM020, SM016, SM018
CM032 Current wave LCOE materially weakens near-term ROI unless Panthalassa's integrated compute model delivers economics below generic marine-energy benchmarks. SM001, SM025, SM020
CM033 Regulatory and maritime permitting remain adoption constraints for offshore infrastructure even when power is not exported to shore. SM009, SM010, SM021
CM034 Trust is a gating factor because cloud buyers require uptime, security, maintainability, and predictable networking that offshore autonomous nodes have not yet demonstrated publicly. SM012, SM024, SM016
CM035 The market estimate range should preserve incompatible units: AI data-center power is measured in TWh, wave deployment in MW, and Panthalassa near-term SOM in nodes. SM007, SM001, SM016
CM036 Analyst wave-market estimates are speculative because they extrapolate from a very small installed base and differ by publisher and forecast horizon. SM001, SM002, SM003, SM004, SM005
CM037 Panthalassa's own total-addressable-market narrative is aspirational because retained sources do not disclose signed customers, pricing, node count, or contracted capacity. SM017, SM016, SM022
CM038 Ocean Energy Europe supports treating ocean energy as a policy-backed sector, but it does not convert Panthalassa's offshore-compute niche into a validated SAM. SM009, SM001
CM039 Gagadget and VKTR summarize the same 2026 Panthalassa funding and ocean-compute thesis, adding media confirmation but little independent market sizing. SM015, SM022
CM040 A diligence-grade market model must triangulate AI electricity demand, wave-energy deployment, buyer willingness to trust offshore compute, and Panthalassa's pilot conversion rate. SM007, SM001, SM006, SM016
CP001 Panthalassa is differentiated by fusing ocean wave-energy generation and at-sea AI compute on one autonomous floating node. SP019, SP023, SP025
CP002 Panthalassa avoids transmitting electricity to shore by using generated power onsite for AI chips and selling compute capacity. SP020, SP025
CP003 CorPower Ocean markets wave-energy converters and CorPack clusters that combine multiple devices into 10-30MW arrays. SP001
CP004 CorPower positions wave energy as steady renewable generation that can complement wind and solar across seasons. SP001
CP005 Oscilla Power develops the Triton wave-energy converter for energy, defense, homeland security, and oceanography use cases. SP002
CP006 Oscilla says the Triton wave-energy converter is backed by 16 granted patents plus additional pending patents. SP002
CP007 Eco Wave Power is a public Nasdaq-listed wave-energy company focused on onshore conversion of ocean and sea waves into electricity. SP011
CP008 Eco Wave Power reports a global project pipeline of 404.7 MW across planned projects in Portugal, Taiwan, and India. SP011
CP009 Aikido Technologies markets floating data centers integrated with offshore wind infrastructure and flat-pack platform assembly for AI-grade compute. SP003
CP010 Aikido coverage describes a 100 kW Norway proof-of-concept around a refurbished turbine due by the end of 2026 and links the company to an NVIDIA startup program. SP010, SP003
CP011 NetworkOcean says it builds and operates floating data-center barges and underwater data-center capsules that are cheaper than on-land alternatives. SP004
CP012 Microsoft Project Natick Phase 2 demonstrated a sealed subsea module with 12 racks, 864 servers, 27.6 PB of storage, and lower failure rates than a land control group. SP005, SP007
CP013 Public summaries report that Microsoft said Project Natick was no longer active in 2024. SP006, SP008
CP014 Data Center Frontier characterizes underwater data centers as technically plausible and environmentally intriguing but still commercially unproven. SP008, SP018
CP015 Shanghai underwater data-center reporting describes an initial 2.3 MW demonstration scaling toward a 24 MW second phase linked to offshore wind. SP008, SP009
CP016 Internet-Pros identifies Highlander Hailanyun, Subsea Cloud, NetworkOcean, and Panthalassa among 2026 ocean data-center builders. SP009
CP017 EcoPortal reports Aikido says GPU customers are circling its offshore wind data-center concept. SP010
CP018 Starcloud raised a $170 million Series A at a roughly $1.1 billion valuation to build space-based data centers. SP013, SP014, SP015
CP019 Starcloud sources report about $200 million of total capital raised and a roadmap toward larger 200 kW-class spacecraft. SP014, SP015
CP020 CoreWeave represents the land-based GPU-cloud status quo that AI-compute buyers can choose instead of offshore compute. SP016, SP018
CP021 Crusoe Energy Systems is a relevant energy-linked compute alternative in the broader neocloud and infrastructure substitute set. SP017, SP018
CP022 Ocean thermal energy conversion is an adjacent ocean-energy substitute that exploits temperature gradients rather than wave motion. SP012
CP023 New Scientist quotes Jonathan Koomey warning that wave power can work but the ocean is harsh because salt and waves attack equipment. SP018
CP024 Independent coverage says offshore computing still must prove it can compete economically with conventional data centers connected to grids and fiber networks. SP018, SP008
CP025 Conventional land data centers and neoclouds retain advantages in grid connection, fiber connectivity, scale, and buyer familiarity. SP018, SP016
CP026 China appears furthest along in commercializing submerged data centers at meaningful scale through Hainan and Shanghai projects. SP008, SP009
CP027 Panthalassa announced a $140 million Series B led by Peter Thiel in May 2026. SP019, SP023
CP028 Panthalassa’s Thiel-led 2026 Series B gives it a materially larger disclosed funding base than most private pure-play wave-energy developers. SP019, SP022
CP029 Panthalassa is pre-revenue with no publicly disclosed paying customers in retained 2026 sources. SP019, SP018
CP030 Panthalassa has more disclosed venture capital than many wave-energy peers, but that capital has not yet converted into commercial fleet proof. SP019, SP001, SP002
CP031 Direct wave-energy peers primarily sell or develop electricity-generation systems rather than integrated at-sea AI compute capacity. SP001, SP002, SP011
CP032 Offshore data-center peers primarily solve siting, cooling, or wind-powered compute rather than autonomous wave generation and onboard compute in one node. SP003, SP004, SP005, SP008
CP033 Status-quo substitutes include land hyperscale data centers, GPU-cloud neoclouds, internal build, grid power, and adjacent low-carbon power such as nuclear or offshore wind. SP016, SP017, SP018, SP010
CP034 Switching costs for Panthalassa may be moderate for batch or inference workloads, but latency and bandwidth can limit fit versus land-based clouds. SP018, SP025
CP035 AI-compute buyers can multi-home across land clouds, neoclouds, and experimental offshore capacity rather than committing exclusively to Panthalassa. SP016, SP018
CP036 Distribution power in AI compute remains concentrated among hyperscalers, GPU-cloud specialists, and hardware ecosystems rather than new offshore platforms. SP016, SP023, SP018
CP037 Super Micro Computer participation in Panthalassa financing is a positive hardware-supply signal but is not evidence of customer demand. SP023, SP019
CP038 Panthalassa says Series B proceeds fund manufacturing and first deployments of autonomous ocean-powered computing systems. SP023, SP019
CP039 Project Natick is adverse evidence that underwater compute can succeed technically yet still fail to become an active commercial platform. SP005, SP006, SP008
CP040 Historical OTEC experience includes an ocean plant destroyed by weather and waves before generating net power, reinforcing marine durability risk. SP012, SP018
CP041 Public retained sources do not disclose comparable list pricing for Panthalassa, most offshore data-center peers, or wave-energy equipment vendors. SP023, SP003, SP004, SP001, SP002
CP042 Eco Wave Power explicitly links wave-energy growth to AI factories and digital infrastructure demand. SP011
CP043 TechRadar and Energy Digital both frame Panthalassa as a near-$1 billion ocean or wave-powered data-center company after the 2026 financing. SP024, SP021
CP044 CBS and Wikipedia describe Panthalassa nodes as self-propelled, cable-free systems that process AI tasks at sea and send answers by satellite. SP025, SP020
CP045 Aikido floating platforms and Mitsui ship-based computing illustrate likely entrant categories that attack offshore compute without Panthalassa wave integration. SP018, SP010, SP008
CI001 Panthalassa is pre-revenue in public sources: no retained source discloses recognized revenue, ARR, or paying customers as of the May 2026 Series B. SI014, SI023
CI002 Panthalassa's current revenue model is to sell AI inference compute capacity produced at sea, not to sell electricity transmitted back to shore. SI013, SI014, SI025
CI003 Official Panthalassa materials say node-generated electricity is used onboard to power AI chips and return inference tokens to land by low-earth-orbit satellite. SI013, SI025
CI004 Panthalassa and investor materials leave open a future clean-fuels or hydrogen use case, but 2026 commercialization emphasis is AI compute. SI021, SI022, SI016
CI005 Because no customer contracts, service terms, or billing units are public, revenue recognition for compute capacity cannot be assessed from retained evidence. SI014, SI023
CI006 No public source discloses Panthalassa pricing for AI tokens, GPU hours, capacity reservations, or clean-fuel offtake. SI013, SI014, SI023
CI007 Lowercarbon presents a company target of roughly $0.02 per kWh for electricity generation cost, which is investor-sourced and not audited field economics. SI022, SI023
CI008 Gigascale presents Panthalassa's manufacturing-cost target as around $1,500 per kW, but the figure is management/investor sourced rather than a disclosed realized cost. SI021, SI023
CI009 Gigascale reports a claimed node capacity factor up to roughly 90%, a key modeled driver of utilization and cost absorption. SI021, SI023
CI010 Gigascale says a roughly $1 billion coastal factory could produce about 1 GW of node capacity per year, framing scale economics as factory throughput rather than project-by-project construction. SI021, SI013
CI011 The cost structure is structurally capital intensive because Panthalassa must manufacture large steel marine nodes, outfit onboard AI compute, deploy offshore, and operate satellite-linked remote systems. SI013, SI021, SI018
CI012 Panthalassa nodes are described as large solid-steel structures around 85 meters long in 2026 secondary coverage, making steel fabrication a central capex driver. SI017, SI018
CI013 GPU or AI-chip payload cost, replacement cadence, depreciation, and supplier economics are not publicly disclosed. SI013, SI023, SI011
CI014 Offshore maintenance, corrosion, biofouling, insurance, and service-vessel costs are the principal adverse financial unknowns in the public record. SI023, SI018
CI015 Gross margin, contribution margin, working capital needs, and inventory financing are undisclosed because Panthalassa has no public operating financial statements. SI023, SI014
CI016 The company's go-to-market motion is best described as infrastructure-led: prove Ocean-3 AI inference at sea, then convert technical proof into hyperscaler or AI-compute buyer contracts. SI013, SI019, SI020
CI017 Panthalassa's 2026 Series B proceeds are officially earmarked to complete a pilot manufacturing facility near Portland and accelerate Ocean-3 node deployment. SI013, SI014
CI018 Public sources disclose no signed customers, LOIs, backlog, usage volume, GMV, or utilization metrics for the at-sea compute product. SI014, SI023
CI019 The strongest public traction metric is financing traction: a $140 million Series B led by Peter Thiel with a broad new and returning investor syndicate. SI013, SI014
CI020 GeekWire reported that Panthalassa had raised about $210 million in total capital after the Series B. SI014, SI023
CI021 The latest round implies prior disclosed or estimated lifetime capital of about $70 million before the $140 million Series B. SI014, SI013
CI022 Financial Times framed Panthalassa as a roughly $1 billion ocean data-centre startup, but retained sources do not disclose an exact post-money valuation. SI015, SI017, SI018
CI023 A related SEC Form D filing for RNN Ventures Panthalassa Series B a series of Allocations 2026 Master, LLC reported a first sale date of 2026-03-17, $319,000 sold, and 14 investors. SI002, SI023
CI024 A related SEC Form D filing for RNN Ventures Panthalassa B Plus a series of Allocations 2026 Master, LLC reported $1,935,313 sold to 31 investors and was signed on 2026-07-21. SI001, SI023
CI025 The 2026 Form D vehicles are partial special-purpose allocation filings and should not be read as the full $140 million Series B round. SI001, SI002, SI013
CI026 No retained source discloses Panthalassa cash on hand after the Series B, monthly burn, runway, debt, or project-finance commitments. SI014, SI023
CI027 Financing dependency is high because the business must fund factory completion, Ocean-3 deployment, large marine structures, and GPU payloads before revenue proof. SI013, SI021, SI023
CI028 The next financing or partnership trigger is likely Ocean-3 pilot performance and first commercial compute contracts rather than historical revenue growth. SI013, SI019, SI023
CI029 Public sources indicate Ocean-3 is intended to demonstrate AI inference capabilities at sea in 2026 ahead of commercial deployments targeted for 2027. SI013, SI014, SI018
CI030 Panthalassa's product can avoid subsea power export revenue economics only if customers accept satellite-linked, offshore compute rather than grid-delivered electricity. SI014, SI017, SI023
CI031 Public evidence supports demand-side pain points—grid capacity, cooling water, permitting, and land constraints—but not Panthalassa-specific sales efficiency metrics. SI013, SI019, SI020
CI032 CAC, sales cycle, payback period, pipeline conversion, and channel economics are not publicly disclosed. SI014, SI023
CI033 Strategic and hardware investors such as Super Micro Computer and Fortescue Ventures appear in the financing syndicate, but no retained source characterizes them as paying customers. SI013, SI014, SI007
CI034 DataDeep's adverse assessment argues that the $0.02/kWh thesis remains unproven once offshore operations, corrosion, biofouling, maintenance, insurance, and service vessels are priced in. SI023, SI018
CI035 Revenue quality is not yet rateable because the company has no disclosed revenue, customer concentration, contract duration, churn, or renewal data. SI014, SI023
CI036 The margin path is unproven because the public case depends on modeled power cost, capacity factor, manufacturing scale, and maintenance cadence rather than reported gross margins. SI021, SI022, SI023
CI037 Capital intensity is very high relative to current public traction because Panthalassa must build factories and fleets before commercial revenue is demonstrated. SI013, SI021, SI023
CI038 Financial diligence blockers include missing audited financials, revenue detail, ARR, burn, runway, gross margin, CAC, cash balance, customer contracts, and project-finance terms. SI023, SI014
CI039 Panthalassa's public benefit corporation status and Portland headquarters are consistent with its clean-infrastructure positioning but do not reduce financial-disclosure gaps. SI016, SI009, SI012, SI025
CI040 The Series B investor list includes John Doerr, Marc Benioff's TIME Ventures, Max Levchin's SciFi Ventures, Founders Fund, and Fortescue Ventures, providing reputational backing but not financial-statement evidence. SI013, SI003, SI004, SI005, SI006, SI007
CI041 Public context sources establish that levelized-cost claims are not the same as realized gross margin, especially for an offshore compute service with capex and operations costs. SI010, SI023
CI042 The financial verdict is watch-and-verify: compelling capital access and a clear demand narrative, offset by pre-revenue status, missing operating metrics, and unproven marine unit economics. SI013, SI014, SI023
CE001 Panthalassa's product is an autonomous floating node that converts ocean wave energy into onboard electricity and sells the resulting AI compute capacity rather than electricity. SE015, SE016, SE017
CE002 The customer workflow is framed as sending AI jobs to offshore compute and receiving inference tokens or completed results back over satellite links. SE016, SE019, SE022
CE003 Panthalassa says its nodes generate power at sea and do not transmit electricity back to terrestrial grids. SE016, SE019, SE022
CE004 The Ocean-3 class node is described as an approximately 85 meter tall plate-steel or solid-steel floating structure. SE018, SE021
CE005 Investor coverage describes the larger node shape as a spherical top about 50 meters in diameter with a long neck extending below the sea surface. SE021
CE006 Public descriptions place the smaller Ocean-2 spherical top at roughly 30 feet or 9 meters in diameter on a much longer tubular hull. SE022, SE024
CE007 Panthalassa's wave mechanism pumps seawater up a central tube or neck into an internal reservoir as waves move the structure up and down. SE017, SE021
CE008 The internal reservoir drains through a single turbine to generate continuous electrical power for the node. SE021, SE017
CE009 The architecture resembles an overtopping or oscillating-water-column-like hydro approach, but Panthalassa's published evidence emphasizes one reservoir and turbine rather than a multi-device wave farm. SE017, SE021, SE009
CE010 The compute payload is described as AI chips or servers housed in hermetically sealed containers below the sea surface. SE022, SE019, SE025
CE011 Surrounding seawater is positioned as the cooling medium for the sealed compute container and as a way to reduce the cooling burden of land data centers. SE016, SE018, SE014
CE012 Graphics processing units are a relevant compute primitive because GPUs are widely used for parallel workloads and AI acceleration. SE003, SE007
CE013 Inference engines apply rules or trained neural networks to generate predictions or decisions, matching Panthalassa's stated inference-compute emphasis. SE004, SE016
CE014 Panthalassa's nodes are described as self-propelled with no anchor, no cables, and no engine or grid connection at the operating site. SE017, SE016, SE018
CE015 The company says it has developed core power generation, propulsion, autonomy, and at-sea computing technologies over roughly a decade. SE016, SE015
CE016 Low-Earth-orbit satellite backhaul is the stated data path from the remote node to land. SE016, SE022, SE002
CE017 Starlink is a low-Earth-orbit satellite internet service, making it a plausible public analogue for Panthalassa's described backhaul. SE001, SE002, SE022
CE018 New Scientist reports that Starlink bandwidth and latency are likely to make the nodes more suitable for long-running batch, training, simulation, or delayed-result workloads than real-time chatbots or search. SE018, SE023
CE019 Ocean-1 was an early Panthalassa prototype deployed in 2021 in the Strait of Juan de Fuca. SE022, SE016
CE020 Panthalassa tested Ocean-2 in 2024 and reported Ocean-2 as part of the prototype sequence proving capabilities at sea. SE016, SE022, SE013
CE021 Public reporting states that a 2025 Puget Sound Ocean-2 test generated up to about 50 kilowatts in decent wave conditions. SE022, SE024, SE008
CE022 Panthalassa also names Wavehopper among prototypes used to prove at-sea capabilities. SE016
CE023 The company plans to deploy an Ocean-3 pilot node series in the northern Pacific in 2026 to demonstrate AI inference and refine manufacturing. SE016, SE019, SE022
CE024 Commercial deployments are targeted around 2027, but public evidence still places the technology in low-to-mid maturity rather than commercial-scale operation. SE016, SE019, SE018
CE025 Panthalassa's long-term vision is thousands of nodes operating in the open ocean. SE022, SE019, SE016
CE026 Gigascale reports an aspirational manufacturing vision in which roughly $1 billion of factory investment could produce about 1 gigawatt of node capacity per year. SE021
CE027 Panthalassa says nodes can be mass-produced from plate steel in coastal factories, a manufacturing distinction versus conventional land data centers. SE016, SE021
CE028 Gigascale reports management's claim that node power can be available up to about 90% of the time. SE021, SE011
CE029 Capacity factor measures actual output relative to theoretical maximum output, so Panthalassa's 90% figure should be treated as an aspirational availability metric until fleet data exists. SE011, SE021
CE030 The earlier technology storyline included clean fuels or hydrogen as potential uses of open-ocean power before the current compute emphasis. SE021, SE010, SE022
CE031 The main product differentiation is vertical integration of wave generation, energy storage-by-reservoir, cooling, compute, autonomy, and satellite backhaul on one floating asset. SE016, SE017, SE021, SE025
CE032 Panthalassa's public materials and investor coverage imply a manufacturing and marine-engineering know-how moat, but they do not yet establish commercial-scale reliability. SE016, SE021, SE018
CE033 Microsoft Project Natick provides an external technical analogue showing factory-built subsea data centers can use controlled offshore environments and local renewable power, but it is not evidence that Panthalassa's autonomous wave nodes are proven. SE023, SE018
CE034 Project Natick reported less than 90 days from factory to operation and no water consumed for cooling, making it a useful benchmark for sealed marine compute deployment claims. SE023, SE014
CE035 Biofouling can affect human-made objects in water through microorganisms, plants, algae, or animals accumulating on surfaces. SE005, SE018
CE036 Corrosion is the gradual deterioration of materials, usually metals, by chemical or electrochemical reaction with the environment. SE006, SE018
CE037 Panthalassa's YouTube team surface is a developer-signal proxy indicating the company publicly recruits and showcases engineering capability, although the page itself is JavaScript-only. SE012, SE015
CE038 Panthalassa's Ocean-2 YouTube sea-trial page is a practitioner-signal proxy for prototype activity, but it cannot substitute for instrumented test data. SE013, SE022
CE039 The company has not disclosed third-party certifications, uptime statistics, formal security controls, environmental impact measurements, or independent reliability data for commercial nodes. SE015, SE016, SE018
CE040 New Scientist reports that remote ocean operations create reliability challenges because physical intervention remains common in abnormal data-center incidents. SE018, SE022
CE041 Data security for Panthalassa remains an underwriting gap because public sources describe satellite backhaul but do not specify encryption, key management, tenant isolation, or incident response controls. SE016, SE022, SE001
CE042 Waste heat from sealed server modules would dissipate into surrounding seawater, but potential nearby marine ecosystem effects remain unclear in the public record. SE018, SE007
CE043 The product line is best viewed as a prototype-to-pilot asset roadmap rather than a catalog of commercial SKUs available for purchase. SE022, SE016, SE019
CE044 Panthalassa's value proposition directly targets grid capacity, cooling-water scarcity, permitting delays, and land constraints faced by terrestrial AI data centers. SE016, SE015, SE019
CU001 Panthalassa's most natural target customers are hyperscale cloud and AI infrastructure buyers that need large amounts of power-dense compute capacity. SU001, SU002, SU011
CU002 Amazon Web Services is a demand-side proxy for the target cloud-buyer segment, not evidence that AWS has signed with Panthalassa. SU003, SU002
CU003 Microsoft Azure is a demand-side proxy for hyperscale cloud customers, not evidence that Azure is a Panthalassa customer. SU004, SU002
CU004 Colocation and merchant AI-cloud buyers represent adjacent infrastructure customers that buy space, power, cooling, and connectivity, but Panthalassa has no disclosed colocation channel. SU005, SU001, SU011
CU005 Panthalassa's stated business model is to consume wave power onboard and sell compute capacity rather than transmit electricity to shore. SU014, SU016, SU017
CU006 The Ocean-3 series is described as performing AI inference at sea, with data returned to land by low-Earth-orbit satellites. SU014, SU013, SU018
CU007 Retained 2026 public sources do not name any signed Panthalassa customers. SU013, SU014, SU015, SU024
CU008 Retained 2026 public sources do not disclose Panthalassa revenue, customer pilots, letters of intent, or production customer deployments. SU013, SU014, SU024, SU009
CU009 The chapter's customer-proof sources are demand-side proxies showing target-buyer need; they are not purchase proof for Panthalassa. SU001, SU002, SU003, SU004
CU010 The IEA-linked data-center demand estimate retained in public sources projects data-center electricity demand could reach roughly 945 TWh per year by 2030. SU012, SU020, SU002
CU011 Deloitte cites strong demand from cloud providers and AI workloads as a driver of expanding data-center supply and capital spending. SU001
CU012 SemiAnalysis estimates global data-center critical IT power demand rising from about 49 GW in 2023 to 96 GW by 2026, with AI consuming about 40 GW. SU011, SU001
CU013 SemiAnalysis describes AI training workloads as relatively latency-insensitive and more dependent on abundant inexpensive electricity than proximity to population centers. SU011
CU014 New Scientist reports that Starlink backhaul makes Panthalassa most practical for jobs that run for hours or days rather than low-latency chatbots or search assistants. SU020, SU015
CU015 New Scientist quotes Uptime Institute warning that power and networking are top data-center outage causes and are uniquely difficult in remote environments with little or no staff. SU020, SU007, SU015
CU016 Super Micro Computer is disclosed as a Series B investor and is better classified here as a potential hardware supplier or partner, not as a customer. SU014, SU013, SU006
CU017 Supermicro's public product portfolio includes AI infrastructure and GPU server categories that make it relevant to the hardware supply chain. SU006
CU018 Panthalassa planned an Ocean-3 pilot series in 2026 and commercial deployments in 2027, but those are product milestones rather than customer adoption metrics. SU013, SU014, SU015
CU019 Ocean-1, Ocean-2, and Wavehopper sea trials are technical prototypes and do not establish paying-customer usage. SU014, SU013, SU024
CU020 CBS reports Panthalassa expects multiple systems to work together as a data center and targets offshore operation around August 2026, but this is a company expectation. SU018
CU021 No retained source discloses active accounts, utilization, repeat purchase, deployed customer locations, or contracted capacity for Panthalassa. SU014, SU013, SU024, SU020
CU022 Rows in the named customer proof table should be read as prospective target design-partner categories, not signed accounts. SU001, SU002, SU003, SU004, SU024
CU023 Because Panthalassa has no disclosed customers, NRR, GRR, logo churn, renewal rate, and contract-length metrics are not applicable from public evidence. SU013, SU014, SU024
CU024 No retained public source provides a named customer reference, satisfaction score, customer quote, renewal, or quantified outcome for Panthalassa. SU013, SU014, SU009, SU024
CU025 Future customer concentration risk is structurally high because a small number of hyperscale and large AI-cloud buyers control a large share of relevant capacity demand. SU002, SU011, SU003, SU004
CU026 Wikipedia's data-center article reports that AWS, Microsoft Azure, and Google Cloud collectively account for approximately 59% of global hyperscale data-center capacity. SU002
CU027 Hyperscale data centers are typically 100 MW or larger and are designed for cloud services, AI training, and large-scale data processing. SU002
CU028 AWS is a global cloud platform with more than $100 billion of 2025 revenue and compute services that include CPU and GPU capacity. SU003
CU029 Microsoft Azure is a global cloud platform with hundreds of services and AI offerings, making it a clear target-buyer proxy for large-scale compute capacity. SU004
CU030 Colocation centers provide power, cooling, space, physical security, and connectivity to tenants, making them a relevant channel analogy but not a disclosed Panthalassa route to market. SU005, SU014
CU031 Procurement friction should be expected because Panthalassa asks buyers to trust novel offshore, satellite-linked, remote infrastructure instead of audited land-based data centers. SU005, SU020, SU007
CU032 Panthalassa's near-term buyer fit is strongest for delay-tolerant, bandwidth-light inference or batch workloads rather than user-facing low-latency applications. SU014, SU020, SU011
CU033 Panthalassa says its nodes send inference tokens to land by satellite and use the surrounding ocean for cooling. SU014, SU017
CU034 Demand-side pressure is reinforced by public data-center energy and permitting constraints, including large delayed or blocked data-center projects cited in the data-center source. SU002, SU001
CU035 Wave power is not widely employed commercially, which weakens customer confidence in Panthalassa until Ocean-3 and later nodes prove durability. SU008, SU024, SU023
CU036 Mordor places the wave-energy market at a small installed base in 2026 relative to the scale implied by hyperscale AI infrastructure. SU023, SU008
CU037 Lowercarbon presents a target electricity generation cost of about $0.02 per kWh for Panthalassa nodes, but this is partner-claimed rather than customer-validated. SU025, SU024
CU038 Gigascale describes high capacity-factor and manufacturing-scale ambitions for Panthalassa, but those claims remain partner-side support rather than customer proof. SU019, SU024
CU039 Independent coverage frames the unresolved customer question as whether buyers will be willing to run AI workloads offshore under satellite-connectivity constraints. SU009, SU024, SU020
CU040 No retained public source identifies a Panthalassa marketplace listing, reseller, systems-integrator channel, or signed design-partner program. SU014, SU013, SU009, SU024
CU041 The practical buyer-user-payer split is likely infrastructure procurement as payer, cloud or AI platform operators as economic buyers, and AI research or product teams as users. SU003, SU004, SU011
CU042 Panthalassa's geographic customer surface is global in theory, but the initial deployment proof is tied to northern Pacific ocean operations rather than customer sites. SU014, SU013, SU018
CU043 There is no public evidence of revenue bands by customer size because there are no disclosed customers or contracts. SU013, SU014, SU024
CU044 If Ocean-3 produces credible field data, expansion could occur by adding node capacity or fleets for the same buyer rather than by deploying software seats. SU014, SU018, SU019
CU045 The main land-and-expand gate is conversion from product pilot data into auditable customer contracts or capacity commitments. SU014, SU024, SU020
CR001 Panthalassa's highest-severity risk is open-ocean survivability because its steel nodes, turbines, electronics, cooling surfaces, and station-keeping must work amid corrosion, biofouling, storms, and wave damage. SR011, SR009, SR010, SR018
CR002 New Scientist quoted Jonathan Koomey warning that the ocean is a harsh environment where salt and waves cause trouble for machinery. SR011, SR021
CR003 Corrosion is a natural deterioration process that degrades mechanical strength and permeability, and galvanic corrosion is especially relevant to the marine industry where salt water contacts metal structures. SR009
CR004 Biofouling accumulates organisms on water-exposed objects, can damage hulls and propulsion systems, and can increase hydrodynamic drag by up to 60 percent. SR010
CR005 Panthalassa's public design mitigates some mechanical exposure by eliminating hinges, flaps, gearboxes, anchors, engines, and shore-power cables. SR015, SR018, SR016
CR006 The company's mitigation maturity remains low-to-medium because Ocean-1, Ocean-2, and Wavehopper sea trials did not prove a compute-carrying commercial fleet operating through full seasonal marine conditions. SR014, SR015, SR021, SR024
CR007 Panthalassa plans to deploy Ocean-3 pilot nodes in the northern Pacific in 2026 and commercial systems in 2027, making Ocean-3 the decisive near-term survivability gate. SR015, SR014, SR017
CR008 Uptime Institute's public materials describe it as a standard bearer for digital infrastructure performance, and New Scientist cites Uptime's Jacqueline Davis that power and networking are the top two data-center outage causes. SR012, SR011
CR009 Power and networking failures are uniquely difficult for Panthalassa because its nodes are remote, minimally staffed offshore data centers rather than buildings with on-site technicians. SR011, SR012, SR016
CR010 New Scientist reported that physical intervention remains common in abnormal data-center incidents, including manual restarts, which raises residual exposure for no-staff offshore nodes. SR011, SR012
CR011 Panthalassa uses Starlink or low-Earth-orbit satellite backhaul as the link to shore, creating a single-platform communications dependency. SR015, SR014, SR016, SR018
CR012 Starlink latency and bandwidth make Panthalassa better suited to delay-tolerant batch or inference workloads than low-latency chatbots, search assistants, or tightly coupled training. SR011, SR016, SR021
CR013 Panthalassa's commercial viability narrows if inference power demand does not grow enough to rival training demand, because critics see its satellite-limited workload surface as constrained. SR011, SR021
CR014 BOEM oversees renewable energy resources on the U.S. Outer Continental Shelf, so U.S. offshore deployments can face federal offshore-energy jurisdiction and permitting processes. SR001, SR002
CR015 BOEM's 2025 renewable-energy page recorded a temporary halt and review of offshore wind leasing and multiple lease-area rescissions, showing policy volatility for offshore energy projects. SR001, SR008
CR016 DOE's Marine Energy Program frames marine energy as a developing resource needing R&D, demonstration support, and barrier reduction before cost-effective deployment. SR002
CR017 Environmental permitting for large data centers can involve NEPA, ESA, Clean Water Act, Clean Air Act, water, stormwater, hazardous-waste, zoning, noise, traffic, and local land-use approvals. SR003, SR004
CR018 Legal commentary in 2026 emphasized that federal data-center permitting acceleration does not eliminate state and local environmental review, public-land planning, or community resistance. SR003, SR004, SR005
CR019 Hunton and Data Center Knowledge reported rising data-center permitting hurdles, local resistance, and litigation risk tied to energy, land, water, air, noise, and environmental impacts. SR005, SR007
CR020 The Jones Act requires goods transported by water between U.S. ports to use vessels built in the United States, U.S.-flagged, U.S.-owned, and crewed by U.S. citizens or permanent residents. SR006
CR021 Jones Act cabotage rules may constrain Panthalassa's U.S. domestic towing, service-vessel, spares, and logistics options if nodes or materials move between U.S. ports. SR006, SR024
CR022 Waste-heat effects on nearby marine ecosystems remain unclear because New Scientist reported that currents may disperse heat but potential ecosystem effects are not yet known. SR011, SR010
CR023 No retained source establishes a tested privacy, security, or incident-response regime for AI workloads processed on autonomous offshore nodes. SR025, SR015, SR016
CR024 No retained source establishes a published patent portfolio, freedom-to-operate opinion, or IP moat for Panthalassa's combined wave-energy and offshore-compute system. SR025, SR016, SR014
CR025 Panthalassa depends on GPU and server hardware supply, and Super Micro Computer is named as an investor rather than a proven committed supplier. SR015, SR014, SR028
CR026 Panthalassa's use of AI chips offshore exposes it to GPU availability, hardware qualification, marine cooling reliability, and payload-swap execution risk. SR015, SR016, SR021
CR027 Panthalassa is pre-revenue with no publicly disclosed paying customers in the retained risk sources. SR014, SR011, SR016
CR028 The company raised $140 million in Series B funding in May 2026 and roughly $210 million total, but this capital must fund pilot manufacturing, Ocean-3 deployments, and proof of commercial economics. SR014, SR015, SR023
CR029 Panthalassa's model is capital intensive because it requires coastal manufacturing, steel marine structures, offshore operations, satellite connectivity, and GPU payloads before revenue is proven. SR015, SR020, SR021
CR030 Gigascale reports management targets near $1,500 per kilowatt capex, around 90 percent capacity factor, and a billion-dollar factory producing about 1 GW per year, but these are still investor or company-side projections. SR020, SR021
CR031 DataDeep characterized Panthalassa's two-cent-per-kilowatt-hour energy-cost claim as unproven at sea and vulnerable to offshore maintenance, insurance, corrosion, and biofouling costs. SR021
CR032 EIA reported that U.S. solar generation was forecast to grow 75 percent from 2023 to 2025, underscoring that land-based renewables continue to scale while wave economics remain unproven. SR013, SR002
CR033 The economic hurdle is severe because chapter facts place marine wave LCOE at roughly 3 to 6 times solar, while Panthalassa still must prove delivered compute costs after offshore maintenance. SR021, SR013, SR026
CR034 Founder concentration is material because public risk sources identify Garth Sheldon-Coulson and Brian Moffat as the named co-founders and do not disclose a broad executive bench or board. SR014, SR025, SR016
CR035 Panthalassa did not respond to New Scientist's skeptical inquiry before publication, which weakens third-party visibility on technical and economic rebuttals. SR011
CR036 Panthalassa's disclosed team draws from aerospace, naval architecture, software, manufacturing, military, and research backgrounds, which partially mitigates execution risk. SR025, SR020
CR037 A deep investor syndicate led by Peter Thiel, with returning Founders Fund, Gigascale, Lowercarbon, Unless, and WovenEarth, mitigates but does not remove financing dependency. SR014, SR015, SR029
CR038 The most important thesis-break triggers are Ocean-3 pilot failure, corrosion or biofouling degradation, storm survivability failure, regulatory blockage, no customer conversion, and delivered unit economics far above target. SR011, SR021, SR001, SR005, SR014
CR039 A prudent mitigation plan should require pilot-before-scale, independent telemetry, corrosion and biofouling inspections, service-vessel cost tracking, satellite performance logs, and explicit customer conversion milestones. SR021, SR011, SR012, SR015
CR040 Residual exposure remains high even after mitigations because many top risks are not fully testable until a compute-carrying Ocean-3 node operates at sea over multiple months and sea states. SR021, SR011, SR015
CR041 Offshore wind history shows that offshore infrastructure can scale, but costs, permitting, O&M, floating technology maturity, and U.S. policy volatility remain relevant analog risks for Panthalassa. SR008, SR001, SR021
CR042 Panthalassa's key customer concentration risk is prospective rather than observed because a first hyperscale or AI-lab buyer could dominate early revenue if commercial conversion occurs. SR011, SR014, SR015
CV001 Panthalassa announced a $140 million Series B on May 4, 2026, led by Peter Thiel. SV019, SV020
CV002 Panthalassa’s 2026 financing provides multi-year runway to fund the Ocean-3 pilot ahead of commercial revenue. SV019, SV021
CV003 Financial Times and later coverage framed Panthalassa as a near-$1 billion ocean data-centre startup, but no exact post-money valuation is publicly disclosed in retained sources. SV022, SV025, SV018
CV004 Post-money valuation analysis requires share price, fully diluted share count, and conversion or option-pool treatment, none of which is public for Panthalassa. SV009, SV011
CV005 A March 25, 2026 SEC Form D for RNN Ventures Panthalassa Series B reported a first sale on March 17, $319,000 sold, and 14 investors. SV016
CV006 A July 21, 2026 SEC Form D for RNN Ventures Panthalassa B Plus reported a first sale on July 15, $1,935,313 sold, and 31 investors. SV015
CV007 Series B financings can include investor-rights agreements, option-pool mechanics, convertible-security conversion, and other terms that affect dilution and preference overhang. SV011, SV009, SV012
CV008 A private near-unicorn valuation is a negotiated mark rather than a liquid public-market consensus. SV012, SV007, SV018
CV009 Panthalassa remains pre-revenue with no publicly disclosed paying customers or customer contracts. SV018, SV019
CV010 Panthalassa sells AI compute capacity produced at sea rather than transmitting generated electricity to shore. SV020, SV021, SV018
CV011 Panthalassa planned to deploy Ocean-3 pilot nodes in 2026 and target commercial deployments in 2027. SV020, SV019, SV021
CV012 Public prototype evidence covers Ocean-1, Ocean-2, and Wavehopper trials before the compute-carrying Ocean-3 pilot. SV020, SV018, SV021
CV013 Panthalassa investor materials and commentary cite a target delivered energy cost near $0.02 per kWh. SV024, SV018
CV014 Panthalassa investor materials cite roughly $1,500 per kW manufacturing economics and power availability up to about 90% of the time. SV023, SV018
CV015 The $0.02 per kWh, $1,500 per kW, and high-capacity-factor economics remain modeled or investor-supported targets rather than verified commercial fleet data. SV018, SV023, SV024
CV016 Independent AI-infrastructure analysis and reporting support the thesis that data-center power demand is becoming a major bottleneck, including an IEA-cited path toward about 945 TWh per year by 2030. SV029, SV017
CV017 Mordor Intelligence projects wave-energy installed capacity rising from about 10 MW in 2026 to 125 MW by 2031, implying high growth from a tiny base. SV002
CV018 Wave energy still faces high capex and LCOE gaps versus mature renewables, which is a direct economic headwind for Panthalassa. SV002, SV018
CV019 Independent market estimates for wave energy differ materially, with 360iResearch, DataM Intelligence, and PW Consulting publishing divergent 2026-2033 market-size and CAGR figures. SV003, SV004, SV005
CV020 Saltwater corrosion, biofouling, and storm damage are material adverse risks for ocean-based data centers and wave-energy equipment. SV025, SV018, SV029
CV021 Starlink or other LEO satellite backhaul may constrain Panthalassa to longer-running inference or batch workloads rather than latency-sensitive chatbot and search use cases. SV021, SV029, SV018
CV022 Remote offshore sites make physical access, power reliability, and networking failures harder to manage than staffed land data centers. SV021, SV029
CV023 Land-based hyperscale data centers and AI-cloud incumbents retain economies of scale, mature supply chains, and established customer channels that could overwhelm Panthalassa if ocean economics disappoint. SV018, SV017, SV026
CV024 No independent retained source verifies an in-water Ocean-3 compute deployment as of the run date. SV018, SV020
CV025 The Series B syndicate includes Peter Thiel, John Doerr, TIME Ventures, SciFi Ventures, strategic investors, and returning climate backers. SV020, SV019
CV026 Gigascale Capital and Lowercarbon Capital publish supportive views of Panthalassa economics and ocean-power strategy as existing investors or partners. SV023, SV024, SV014
CV027 Investor and partner validation is not equivalent to customer proof or a bankable revenue contract. SV024, SV023, SV019
CV028 Starcloud raised $170 million and reached a reported $1.1 billion valuation for space-based data centers in 2026, making it a frontier-infrastructure valuation analog. SV006, SV030
CV029 CoreWeave is now a public AI-cloud company and is therefore a useful demand-side analog but a poor stage analog for pre-revenue Panthalassa. SV026, SV017
CV030 Crusoe is an imperfect AI-infrastructure comparable because public retained evidence for the specific valuation mark is thin in this chapter allocation. SV031, SV018
CV031 CorPower Ocean and Eco Wave Power demonstrate wave-energy peer activity, but their grid-connected wave-power models are not direct at-sea compute analogs. SV027, SV028, SV002
CV032 Eco Wave Power is publicly traded on Nasdaq under ticker WAVE, making it a public micro-cap reference rather than a venture-backed ocean-compute comp. SV028
CV033 Panthalassa and Starcloud both carry near-unicorn frontier data-center narratives before the relevant non-terrestrial infrastructure class is proven at scale. SV006, SV019, SV022
CV034 The investment recommendation should be research-more or track rather than buy because the valuation is imprecise, the company is pre-revenue, and the commercial pilot remains unproven. SV018, SV019, SV020, SV022
CV035 The valuation stance is unknown-to-stretched because a roughly $1 billion mark is cited but the exact post-money, liquidation preferences, and operating proof are missing. SV022, SV018, SV009
CV036 The appropriate risk rating is high because technical survivability, offshore operations, customer proof, and capital intensity all remain open at commercial scale. SV018, SV025, SV021, SV029
CV037 The bull case is multi-billion-dollar value creation if Panthalassa proves cheap, high-availability at-sea compute into sustained AI power demand. SV024, SV023, SV017, SV029
CV038 The base case is that a successful pilot but slow scale-up could support a roughly $1 billion hold rather than a compelling new-money markup. SV019, SV020, SV022, SV018
CV039 The bear case is a down-round or write-off if survivability, maintenance, latency, or wave-energy economics fail relative to land-based alternatives. SV018, SV025, SV002, SV029
CV040 A credible exit is years away and would most likely require either an IPO-ready revenue profile or strategic acquisition by an AI-cloud, hyperscale, energy, or infrastructure buyer. SV010, SV012, SV018, SV026
CV041 Final diligence must prioritize post-money and terms, Ocean-3 performance data, customer pipeline, unit economics, and survivability evidence. SV018, SV015, SV016, SV019, SV020
CV042 The investable path should remain milestone-gated until Ocean-3 results and credible customer demand convert the option value into underwriting evidence. SV020, SV018, SV019
来源
编号出版方标题引文
SO001 GeekWire Data centers at sea: Panthalassa nets $140M led by Peter Thiel for wave-powered AI
SO002 Wikipedia Panthalassa (company)
SO003 PR Newswire (Panthalassa) Panthalassa Raises $140 Million to Power AI at Sea
SO004 CBS News Using the ocean to power data centers
SO005 TechRadar Peter Thiel-backed company raises to send data centers out to sea
SO006 Energy Digital Panthalassa: The Floating, Wave-Powered Data Centre Unicorn
SO007 Technology Magazine Inside Panthalassa's Wave-Powered Data Centre in the Ocean
SO008 Sustainability Magazine Panthalassa: The Floating, Wave-Powered Data Centre Unicorn
SO009 Gigascale Capital Panthalassa: Scaling Ocean Power
SO010 Lowercarbon Capital Panthalassa
SO011 The Everett Herald That mysterious floating object near Everett? It's a prototype
SO012 TechEBlog Ocean-2, an Innovative Buoy That Aims to Create Clean Energy
SO013 OfficeChai Peter Thiel Leads $140 Million Investment In Panthalassa
SO014 Financial Times Peter Thiel backs $1bn ocean data centre start-up powered by waves
SO015 VKTR Panthalassa Raised $140 Million for Wave-Powered AI Data Centers
SO016 Panthalassa Panthalassa - Planetary-scale energy
SO017 Floating Solutions Peter Thiel backs $140M wave-powered AI data center startup Panthalassa
SO018 Idlen Panthalassa $140M Peter Thiel ocean wave AI data centers
SO019 American Entrepreneurship Today Panthalassa's $140M in Funding to Power AI with Ocean-based Computing Nodes
SO020 Information Today Europe Peter Thiel backs $1bn ocean data centre start-up powered by waves
SO021 IDCNova Panthalassa Secures $140 Million in Series B Funding to Deploy Ocean-Based AI Computing
SO022 Wikipedia Peter Thiel
SO023 Wikipedia Public-benefit corporation
SO024 New Scientist Can floating data centres meet AI's huge energy demand?
SO025 DataDeep Can Wave-Powered Ocean Data Centers Work? Inside Panthalassa's $1B Bet
SM001 Mordor Intelligence Wave Energy Market Size, Share & 2031 Growth Trends Report
SM002 DataM Intelligence Wave Energy Market Share, Size & Growth Report 2026-2033
SM003 PW Consulting Global Wave Energy Market 2026
SM004 360iResearch Wave Energy Market Size & Share 2026-2032
SM005 Research and Markets Wave Energy Market Size, Competitors & Forecast to 2032
SM006 SemiAnalysis AI Datacenter Energy Dilemma
SM007 International Energy Agency Energy and AI - Data centres
SM008 U.S. Energy Information Administration Wind and solar to lead U.S. power generation growth
SM009 Ocean Energy Europe Ocean Energy Europe
SM010 Wikipedia Wave power
SM011 Wikipedia AI boom
SM012 Wikipedia Data center
SM013 Wikipedia Renewable energy
SM014 Wikipedia Cloud computing
SM015 Gagadget Wave-powered AI data centers at sea: Panthalassa raises $140M
SM016 GeekWire Data centers at sea: Panthalassa nets $140M led by Peter Thiel for wave-powered AI
SM017 PR Newswire (Panthalassa) Panthalassa Raises $140 Million to Power AI at Sea
SM018 Energy Digital Panthalassa: The Floating, Wave-Powered Data Centre Unicorn
SM019 Technology Magazine Inside Panthalassa's Wave-Powered Data Centre in the Ocean
SM020 Gigascale Capital Panthalassa: Scaling Ocean Power
SM021 TechRadar Peter Thiel-backed company raises to send data centers out to sea
SM022 VKTR Panthalassa Raised $140 Million for Wave-Powered AI Data Centers
SM023 Wikipedia Panthalassa (company)
SM024 CBS News Using the ocean to power data centers
SM025 Lowercarbon Capital Panthalassa
SP001 CorPower Ocean CorPower Ocean - Wave Energy Converter Technology
SP002 Oscilla Power Oscilla Power - Triton wave energy
SP003 Aikido Technologies Aikido - Clean Marine Datacenters
SP004 NetworkOcean NetworkOcean - Floating and Underwater Ocean Data Centers
SP005 Microsoft Research Project Natick Phase 2
SP006 Wikipedia Project Natick
SP007 Blackridge Research Microsoft Data Center in Ocean - Project Natick
SP008 Data Center Frontier Innovations in Offshore Data Centers
SP009 Internet-Pros Underwater Data Centers 2026 - Highlander, Subsea Cloud
SP010 EcoPortal A floating wind turbine can now hide a 10 MW AI data center
SP011 Eco Wave Power Eco Wave Power
SP012 Wikipedia Ocean thermal energy conversion
SP013 GeekWire Starcloud raises $170M for space-based data centers, hits $1.1B valuation
SP014 SpaceNews Starcloud achieves unicorn status with $170 million raise
SP015 TechCrunch Starcloud raises $170 million Series A to build data centers in space
SP016 Wikipedia CoreWeave
SP017 Wikipedia Crusoe Energy Systems
SP018 New Scientist Can floating data centres meet AI's huge energy demand?
SP019 GeekWire Data centers at sea: Panthalassa nets $140M led by Peter Thiel for wave-powered AI
SP020 Wikipedia Panthalassa (company)
SP021 Energy Digital Panthalassa: The Floating, Wave-Powered Data Centre Unicorn
SP022 Technology Magazine Inside Panthalassa's Wave-Powered Data Centre in the Ocean
SP023 PR Newswire (Panthalassa) Panthalassa Raises $140 Million to Power AI at Sea
SP024 TechRadar Peter Thiel-backed company raises to send data centers out to sea
SP025 CBS News Using the ocean to power data centers
SI001 U.S. Securities and Exchange Commission Form D - RNN Ventures Panthalassa B Plus a series of Allocations 2026 Master, LLC
SI002 U.S. Securities and Exchange Commission Form D - RNN Ventures Panthalassa Series B a series of Allocations 2026 Master, LLC
SI003 Wikipedia John Doerr
SI004 Wikipedia Marc Benioff
SI005 Wikipedia Max Levchin
SI006 Wikipedia Founders Fund
SI007 Wikipedia Fortescue
SI008 Wikipedia Bridgewater Associates
SI009 Wikipedia Benefit corporation
SI010 Wikipedia Levelized cost of electricity
SI011 Wikipedia Nvidia
SI012 Wikipedia Portland, Oregon
SI013 PR Newswire (Panthalassa) Panthalassa Raises $140 Million to Power AI at Sea
SI014 GeekWire Data centers at sea: Panthalassa nets $140M led by Peter Thiel for wave-powered AI
SI015 Financial Times Peter Thiel backs $1bn ocean data centre start-up powered by waves
SI016 Wikipedia Panthalassa (company)
SI017 Energy Digital Panthalassa: The Floating, Wave-Powered Data Centre Unicorn
SI018 TechRadar Peter Thiel-backed company raises to send data centers out to sea
SI019 IDCNova Panthalassa Secures $140 Million in Series B Funding to Deploy Ocean-Based AI Computing
SI020 American Entrepreneurship Today Panthalassa's $140M in Funding to Power AI with Ocean-based Computing Nodes
SI021 Gigascale Capital Panthalassa: Scaling Ocean Power
SI022 Lowercarbon Capital Panthalassa
SI023 DataDeep Can Wave-Powered Ocean Data Centers Work? Inside Panthalassa's $1B Bet
SI024 Business Wire Panthalassa Raises $140 Million to Power AI at Sea
SI025 Panthalassa Panthalassa - Planetary-scale energy
SE001 Wikipedia Starlink
SE002 Wikipedia Low Earth orbit
SE003 Wikipedia Graphics processing unit
SE004 Wikipedia Inference engine
SE005 Wikipedia Biofouling
SE006 Wikipedia Corrosion
SE007 Wikipedia Green computing
SE008 Wikipedia Puget Sound
SE009 Wikipedia PacWave
SE010 Wikipedia Hydrogen economy
SE011 Wikipedia Capacity factor
SE012 Panthalassa (YouTube) Panthalassa — Building the Team
SE013 Panthalassa (YouTube) Panthalassa — Ocean-2 Sea Trial
SE014 Wikipedia Power usage effectiveness
SE015 Panthalassa Panthalassa - Planetary-scale energy
SE016 PR Newswire (Panthalassa) Panthalassa Raises $140 Million to Power AI at Sea
SE017 CBS News Using the ocean to power data centers
SE018 New Scientist Can floating data centres meet AI's huge energy demand?
SE019 TechRadar Peter Thiel-backed company raises to send data centers out to sea
SE020 Energy Digital Panthalassa: The Floating, Wave-Powered Data Centre Unicorn
SE021 Gigascale Capital Panthalassa: Scaling Ocean Power
SE022 Wikipedia Panthalassa (company)
SE023 Microsoft Research Project Natick Phase 2
SE024 TechEBlog Ocean-2, an Innovative Buoy That Aims to Create Clean Energy
SE025 Technology Magazine Inside Panthalassa's Wave-Powered Data Centre in the Ocean
SU001 Deloitte Insights Gen AI power consumption creates need for more sustainable data centers
SU002 Wikipedia Hyperscale data center
SU003 Wikipedia Amazon Web Services
SU004 Wikipedia Microsoft Azure
SU005 Wikipedia Colocation centre
SU006 Supermicro Supermicro Product Portfolio (AI/GPU servers)
SU007 Uptime Institute Uptime Institute
SU008 Wikipedia Wave farm
SU009 Cocoloop Panthalassa raises $140M for AI data centers at sea
SU010 Fizzty Ocean AI Data Centers: Panthalassa Gets $140M From Thiel
SU011 SemiAnalysis AI Datacenter Energy Dilemma
SU012 International Energy Agency Energy and AI - Data centres
SU013 GeekWire Data centers at sea: Panthalassa nets $140M led by Peter Thiel for wave-powered AI
SU014 PR Newswire (Panthalassa) Panthalassa Raises $140 Million to Power AI at Sea
SU015 Wikipedia Panthalassa (company)
SU016 Energy Digital Panthalassa: The Floating, Wave-Powered Data Centre Unicorn
SU017 TechRadar Peter Thiel-backed company raises to send data centers out to sea
SU018 CBS News Using the ocean to power data centers
SU019 Gigascale Capital Panthalassa: Scaling Ocean Power
SU020 New Scientist Can floating data centres meet AI's huge energy demand?
SU021 VKTR Panthalassa Raised $140 Million for Wave-Powered AI Data Centers
SU022 Technology Magazine Inside Panthalassa's Wave-Powered Data Centre in the Ocean
SU023 Mordor Intelligence Wave Energy Market Size, Share & 2031 Growth Trends Report
SU024 DataDeep Can Wave-Powered Ocean Data Centers Work? Inside Panthalassa's $1B Bet
SU025 Lowercarbon Capital Panthalassa
SR001 Bureau of Ocean Energy Management Renewable Energy on the Outer Continental Shelf
SR002 U.S. Department of Energy Marine Energy Program
SR003 Fresh Law Blog (Foley Hoag) Environmental Permitting for AI Data Centers
SR004 National Law Review White House Aims to Accelerate Environmental Permitting for Data Centers
SR005 Hunton Andrews Kurth The Rise of Data Centers Brings Environmental Permitting Challenges and Litigation Risk
SR006 Wikipedia Merchant Marine Act of 1920 (Jones Act)
SR007 Data Center Knowledge Data Centers Face Permitting Hurdles, Rising Litigation Risk
SR008 Wikipedia Offshore wind power
SR009 Wikipedia Corrosion
SR010 Wikipedia Biofouling
SR011 New Scientist Can floating data centres meet AI's huge energy demand?
SR012 Uptime Institute Uptime Institute
SR013 U.S. Energy Information Administration Wind and solar to lead U.S. power generation growth
SR014 GeekWire Data centers at sea: Panthalassa nets $140M led by Peter Thiel for wave-powered AI
SR015 PR Newswire (Panthalassa) Panthalassa Raises $140 Million to Power AI at Sea
SR016 Wikipedia Panthalassa (company)
SR017 CBS News Using the ocean to power data centers
SR018 TechRadar Peter Thiel-backed company raises to send data centers out to sea
SR019 Energy Digital Panthalassa: The Floating, Wave-Powered Data Centre Unicorn
SR020 Gigascale Capital Panthalassa: Scaling Ocean Power
SR021 DataDeep Can Wave-Powered Ocean Data Centers Work? Inside Panthalassa's $1B Bet
SR022 Financial Times Peter Thiel backs $1bn ocean data centre start-up powered by waves
SR023 Technology Magazine Inside Panthalassa's Wave-Powered Data Centre in the Ocean
SR024 The Everett Herald That mysterious floating object near Everett? It's a prototype
SR025 Panthalassa Panthalassa - Planetary-scale energy
SR026 Mordor Intelligence Wave Energy Market Size, Share & 2031 Growth Trends Report
SR027 SemiAnalysis AI Datacenter Energy Dilemma
SR028 Supermicro Supermicro Product Portfolio (AI/GPU servers)
SR029 Lowercarbon Capital Panthalassa
SR030 Sustainability Magazine Panthalassa: The Floating, Wave-Powered Data Centre Unicorn
SV001 Research and Markets Wave Energy Market Size, Competitors & Forecast to 2032
SV002 Mordor Intelligence Wave Energy Market Size, Share & 2031 Growth Trends Report
SV003 360iResearch Wave Energy Market Size & Share 2026-2032
SV004 DataM Intelligence Wave Energy Market Share, Size & Growth Report 2026-2033
SV005 PW Consulting Global Wave Energy Market 2026
SV006 GeekWire Starcloud raises $170M for space-based data centers, hits $1.1B valuation
SV007 Wikipedia Unicorn (finance)
SV008 Wikipedia Discounted cash flow
SV009 Wikipedia Post-money valuation
SV010 Wikipedia Initial public offering
SV011 Wikipedia Series B round
SV012 Wikipedia Venture capital
SV013 Data Center Dynamics Panthalassa unveils wave-powered floating data center platform
SV014 Gigascale Capital Gigascale Capital
SV015 U.S. Securities and Exchange Commission Form D - RNN Ventures Panthalassa B Plus a series of Allocations 2026 Master, LLC
SV016 U.S. Securities and Exchange Commission Form D - RNN Ventures Panthalassa Series B a series of Allocations 2026 Master, LLC
SV017 SemiAnalysis AI Datacenter Energy Dilemma
SV018 DataDeep Can Wave-Powered Ocean Data Centers Work? Inside Panthalassa's $1B Bet
SV019 GeekWire Data centers at sea: Panthalassa nets $140M led by Peter Thiel for wave-powered AI
SV020 PR Newswire (Panthalassa) Panthalassa Raises $140 Million to Power AI at Sea
SV021 Wikipedia Panthalassa (company)
SV022 Financial Times Peter Thiel backs $1bn ocean data centre start-up powered by waves
SV023 Gigascale Capital Panthalassa: Scaling Ocean Power
SV024 Lowercarbon Capital Panthalassa
SV025 TechRadar Peter Thiel-backed company raises to send data centers out to sea
SV026 Wikipedia CoreWeave
SV027 CorPower Ocean CorPower Ocean - Wave Energy Converter Technology
SV028 Eco Wave Power Eco Wave Power
SV029 New Scientist Can floating data centres meet AI's huge energy demand?
SV030 TechCrunch Starcloud raises $170 million Series A to build data centers in space
SV031 Wikipedia Crusoe Energy Systems