Joulent
为 AI 数据中心同址配套燃气电力,以 Project Kilby 和 National Grid Ventures $1.75B 投资为锚
Joulent 是押注 AI 数据中心共址供电的高估值期权:合同验证扎实、伙伴网络强,但仍未产生收入,客户单一,资本开支很重。
封面要素
公司概况
Joulent 是一家位于休斯敦、由 Engine No. 1 孵化的能源基础设施公司,为 AI 数据中心和工业用户建设大规模同址发电。 其旗舰 Project Kilby 是位于德克萨斯州 Reeves County 的 2.67-GW 表后燃气园区,由 Chevron 持有、与 Microsoft 签订的 20 年购电协议锚定。2026 年 7 月,公司从 National Grid Ventures 获得 $1.75B 战略投资,对应约 35% 股权,隐含估值约 $5B。
- 成立时间
- 2023-01-01
- 创始人
- Chris James
- 创立地点
- Houston, TX, USA
- 总部
- Houston, TX, USA
- 产品
- 通过同址、表后发电向超大规模数据中心交付稳定电力服务——GE Vernova 7HA 联合循环燃气轮机和 Caterpillar Solar Turbines 机组与 ERCOT 电网隔离运行——并以长期购电协议出售。
- 客户
- 超大规模数据中心运营商(锚定客户:Microsoft)和大型工业用电方。
- 商业模式
- 融资、建设、持有并运营同址发电资产,通过长期(10-20 年)购电协议售电;同时借助合资股权分享项目经济收益(Kilby 合资公司 的 50% 期权)。
- 阶段
- growth
- 融资情况
- 2026 年 7 月 1 日从 National Grid Ventures 获得 $1.75B 战略投资(约 35% 股权),隐含估值约 $5B。
执行摘要
主要优势
- 与投资级对手方 Microsoft 签下 20 年购电协议,需求和收入质量有锚。
- Chevron(燃料 / PPA)、GE Vernova(涡轮机)、National Grid Ventures(资本)拼出稀缺伙伴生态,复制难度高。
- 离网、表后共址能更快拿到电力,切中公共电网跟不上的稀缺且快速增长需求。
主要风险
- $7-9B 首个规模化燃气巨型项目执行风险高,首批供电要到 2028 年。
- Microsoft 是唯一锚定客户,客户集中度极高。
- Texas SB6、ERCOT 大负荷规则带来监管和并网复杂度,碳排与搁浅资产风险也在。
未决问题
- 公开财务几乎空白:收入、利润率、烧钱速度和完整股权结构表均未披露。
- 为 $7-9B 建设融资的 Kilby 完整资本结构(债务 / 股权 / 伙伴分摊)未披露。
- NGV 投资后的确切股权所有权和治理结构未公开。
- 没有第二个具名客户或项目来缓释集中度风险。
目录
01公司概览
1.1 身份与商业模式
Joulent 是一家总部位于德克萨斯州休斯敦、技术驱动的能源公司,为 AI 数据中心和工业用户建设大规模同址电力基础设施。 经过投资机构 Engine No. 1 与 GE Vernova 多年合作孵化后,公司于 2026 年 6 月公开亮相。其产品化的 “Across-the-Meter” 电力栈把燃气发电、电池储能和可再生能源同客户负荷同址部署,并管理并网接入,以工业级规模交付可靠电力。 公司主打以“美国创新的速度和规模”供能,瞄准超大规模算力面临的尖锐电力瓶颈。Joulent 是私营公司;截至 2026 年 7 月,最合适的阶段判断是成长型开发期,首个旗舰项目仍处于最终投资决策前。[CO001, CO002, CO003, CO015, CO024, CO030]
| 指标 | 数值 / 状态 | 截至 | 置信度 | 缺口 / 备注 |
|---|---|---|---|---|
| 隐含估值 | ~$5.0B(由 NGV 交易推算) | Jul 2026 | 中 | 由 $1.75B 换约 35% 推算 |
| 最新投资 | $1.75B 战略投资(National Grid Ventures) | Jul 1 2026 | 高 | 少数股权 |
| 出售股权 | 约 35% 出售给 National Grid Ventures | Jul 2026 | 高 | 其余股权结构未披露 |
| 旗舰项目 | Project Kilby,2.67 GW,美国西得州 | Jun 2026 | 高 | FID 目标为 2026 年底 |
| 锚定客户 | Microsoft(20 年协议) | Jun 2026 | 高 | 客户集中 |
| 总部 | 美国得克萨斯州休斯敦 | 2026 | 中 | — |
| 阶段 | 增长期 / FID 前开发 | Jul 2026 | 中 | 首批供电目标 2028 年 |
| 员工人数 | 未披露 | Jul 2026 | 低 | 私营公司 |
数值汇总自公司、投资人和媒体来源;估值由已披露的 $1.75B 换 35% 交易推算(估计)。「未披露」表示私营公司指标。
[CO006, CO007, CO008, CO009, CO025]Joulent 的身份、产品、旗舰项目、客户、资本与合作伙伴如何串在一起。
[CO001, CO015, CO028]1.2 领导层与治理
Joulent 的公开门面是创始人兼 CEO Chris James;他同时领导自己 2020 年创办的激进投资机构 Engine No. 1, 该机构因 2021 年赢得 ExxonMobil 三个董事席位而出名。这一双重身份给 Joulent 带来经验丰富、可信度高的能源转型领军人物, 也让关键人风险集中,因为公司尚未公开其他具名高管。James 将 AI 时代的领导力定义为一场竞赛:谁能以最快、 最可靠、最低成本交付能源和算力,谁就占优;这正是 Joulent 策略背后的命题。创始人之外的治理细节很薄:高管团队、 董事会构成,以及 Engine No. 1 与新投资方 National Grid Ventures 之间的控制权平衡尚未公开,管理层厚度由此成为重要尽调缺口。[CO004, CO005, CO026, CO036, CO032]
1.3 融资、估值与所有权
2026 年 7 月 1 日,英国公用事业公司 National Grid plc 的商业化非监管部门 National Grid Ventures 同意向 Joulent 投资 $1.75B,取得约 35% 股权;Kirkland & Ellis 为该交易提供法律顾问。该价格隐含约 $5B 的标题估值, 将这家年轻公司推至独角兽规模;但该数字来自交易推算,并非审计标记。National Grid 称这笔投资由资产负债表出资, 且是在其创纪录资本计划之外的增量投入;公司已为 Ventures 项目预留约 $1B 近期资本。这笔资金拟用于 Project Kilby 和其他大型项目,让 National Grid 直接押注美国快速增长的“大负荷”合同电力需求;但 Joulent 的剩余所有权和完整股权结构仍未披露。[CO006, CO007, CO008, CO021, CO022, CO023]
| 利益相关方 | 角色 | 控制权 / 经济重要性 | 尽调问题 |
|---|---|---|---|
| National Grid Ventures | 战略少数股权投资人 | 约 35% 股权;$1.75B 资本 | 治理权、董事席位、资本承诺 |
| Engine No. 1 | 创始方 / 发起方 | 孵化 Joulent;创始人-CEO 重叠 | NGV 入股后的剩余持股与控制权 |
| Chevron (Energy Forge One) | 电力 / PPA 合作伙伴 | 持有 20 年 Microsoft PPA;约 50% Kilby JV 权益 | 合同条款与股权分配 |
| GE Vernova | 燃机技术合作伙伴 | 供应核心 7HA 发电设备 | 设备交付档期与定价 |
| Microsoft | 锚定客户 | Kilby 电力 20 年包销 | 包销量和照付不议条款 |
| Caterpillar / Solar Turbines | 设备供应商 | Kilby 补充发电设备 | 供货范围与可靠性承诺 |
枚举覆盖已公开识别的利益相关方;完整私有股权结构和确切经济分配未披露,因此覆盖仍不完整。
[CO006, CO012, CO013, CO027]截至 2026 年 7 月,Joulent 的核心 KPI。
[CO007, CO008, CO039]1.4 旗舰项目与伙伴关系
Joulent 的旗舰是 Project Kilby:位于德克萨斯州 Permian Basin、Reeves County 近 Pecos 的 2,000 多英亩同址电力园区, 容量 2.67 吉瓦,约等于 200 万户家庭用电。Chevron 通过子公司 Energy Forge One LLC 持有与 Microsoft 签订的 20 年购电协议;Microsoft 邻近的 Pecos 数据中心园区计划在五至七年内新增约 2 GW 算力负荷;Joulent 持有该合资公司的 50% 股权期权。GE Vernova 提供核心涡轮技术,Caterpillar 的 Solar Turbines 部门补充发电能力。该项目预计耗资 $7-9B,目标是在 2026 年底前完成最终投资决策、2028 年首次送电,并声称可创造超过 $10B 州和地方税收以及约 2,000 个就业岗位。[CO009, CO010, CO011, CO012, CO013, CO014]
1.5 里程碑与轨迹
Joulent 的短历史可追溯到 Engine No. 1 2020 年创立和 2021 年 ExxonMobil 战役;随后约从 2023 年起, Engine No. 1 与 GE Vernova 用三年左右孵化电力平台。2026 年动能陡增:1 月宣布 Chevron、GE Vernova 和 Engine No. 1 将为美国数据中心同址建设最高约 4 GW 燃气电厂;6 月 22 日签署 Chevron-Microsoft 20 年 Kilby 协议,同时 Joulent 公开亮相;7 月 1 日获得 National Grid Ventures 投资。接下来的节点是 2026 年底前目标最终投资决策和 2028 年首次送电。不过独立观察者提醒,为数据中心新建大型燃气电厂存在执行、排放和搁浅资产风险,这给原本陡峭上行的轨迹降温。[CO029, CO031, CO006, CO009, CO032]
| 日期 | 事件 | 类型 | 金额 / 估值 / 状态 | 参与方 | 影响 |
|---|---|---|---|---|---|
| 2020 | Chris James 创办 Engine No. 1 | 创立 | — | Chris James | 发起平台的起点 |
| 2021 | Engine No. 1 赢得 ExxonMobil 三个董事会席位 | 治理 | — | Engine No. 1、ExxonMobil | 确立能源转型信誉 |
| 2023 | Engine No. 1 与 GE Vernova 开始开发 Joulent 电力平台 | 产品 | 多年建设 | Engine No. 1、GE Vernova | 技术和项目基础 |
| Jan 2026 | Chevron、GE Vernova 与 Engine No. 1 公布共址燃气方案(最高约 4 GW) | 合作 | 已公告 | Chevron、GE Vernova 与 Engine No. 1 | 释放 Kilby 式管线信号 |
| Jun 22 2026 | Chevron 与 Microsoft 签署 Kilby 20 年电力协议 | 规模化 | 已签署 | Chevron、Microsoft | 锁定首个旗舰项目包销 |
| Jun 22 2026 | Joulent 公开亮相 | 产品 | 已发布 | Joulent | 公司以独立品牌浮出水面 |
| Jul 1 2026 | National Grid Ventures 以 $1.75B 投资换取约 35% 股权 | 融资 | $1.75B / 隐含 ~$5B | NGV、Joulent | 独角兽级战略资本 |
| 2026 年底(目标) | Project Kilby 最终投资决策 | 规模化 | 计划中 | Chevron、Joulent、Microsoft | 进入建设的关口 |
| 2028(目标) | Kilby 首批供电 | 规模化 | 计划中 | Joulent、Chevron | 收入拐点 |
单一记录时间线;未来日期行是公司 / 合作伙伴目标,并非已完成事件。
[CO005, CO006, CO009, CO029]按日期梳理 Engine No. 1 成立至 Kilby 计划首次送电之间的关键节点。
[CO005, CO006, CO009, CO031]1.6 图表要点
02市场分析
2.1 市场边界与替代方案
Joulent 面向的市场,是服务“大负荷”数据中心——尤其是 AI 算力园区——的大规模合同电力供应和实体电力基础设施。 纳入范围的支出包括新增发电容量、并网接入、同址和表后电厂,以及长期购电协议;明确排除 GPU、服务器、制冷等 IT 硬件。 现状替代方案是通过受监管公用事业公司做常规并网接入,但多年排队正越来越卡住超大规模云厂商。相邻供给路径包括电网侧独立发电商、 核电重启和小型模块化反应堆、燃料电池或混合现场系统,它们在边际上竞争。表后和同址燃气发电已成为获得可靠电力最快的路径, Joulent 产品化的正是这一段。尽调必须划清边界,因为标题性的“数据中心”支出把 IT 硬件、地产和电力混在一起;只有把电力与并网层单独剥离, Joulent 的机会才看得清。公司竞争的不是更庞大的算力硬件预算,而是更窄的一片:按超大规模客户时间表交付的新可调度容量。[CM001, CM002, CM003, CM004, CM005, CM030]
| 层级 | 范围内 | 范围外 / 相邻 | 替代方案 / 备注 |
|---|---|---|---|
| 核心市场 | 面向 AI / 大负载数据中心的签约大规模电力 | GPU、服务器、冷却等 IT 支出 | 通过公用事业电网接入(慢) |
| 发电 | 共址燃气、储能、表后可再生能源 | 居民 / 零售电力 | 现场柴油备用发电机组 |
| 接入 | 专用变电站、输电连接 | 光纤 / 网络建设 | 标准 ERCOT 排队(多年) |
| 包销 | 10-20 年 PPA / 照付不议 | 现货商售电 | 受监管电价供电 |
| 相邻供应 | 核电重启 / SMR、电网级 IPP、燃料电池 | 屋顶光伏 | 公用事业自有发电 |
定义共址数据中心电力市场边界;明确排除 IT 硬件,电网接入是现状替代方案。
[CM001, CM002, CM003, CM004, CM029]从场址控制到送电的价值链漏斗,对应 Kilby 的开发路径。
[CM019, CM017, CM037]2.2 多视角测算市场规模
没有一个数字能概括这个市场,只能从多个视角交叉测算。按用电量看,IEA 预计全球数据中心用电到 2030 年大致翻倍, 达到约 945 太瓦时。按容量看,综合估算显示数据中心电力需求 2026 年接近 132 吉瓦,到 2030 年升至约 290 吉瓦; 仅 Gartner 就预计 2026 年用电增长约 26%。按资本看,McKinsey 和 JP Morgan 将本十年全球数据中心投资框定在 $5-7T,McKinsey 的标题数字接近 $6.7T;Goldman Sachs 则测算到 2030 年电力需求跃升 165%。2026 年超大规模云厂商资本开支约 $690B——仅 Microsoft 就接近 $190B——凸显这些曲线背后的支出。关键在于,任一同址开发商可触达的只是一小部分, 因此 Joulent 的可服务市场和可获得市场远小于标题总量,且没有公开量化。[CM006, CM007, CM008, CM009, CM010, CM011]
| 口径 | 指标 | 估算 | 来源依据 | 限制 |
|---|---|---|---|---|
| 用电量(TAM) | 2030 年全球数据中心用电量 | 约 945 TWh(大致翻倍) | IEA,经 S&P Global | 用电量,不是容量 |
| 容量(TAM) | 数据中心电力需求 2026 -> 2030 | ~132 GW -> ~290 GW | Gartner / Introl | 多来源混合 |
| 资本开支(TAM) | 2030 年全球数据中心投资 | ~$5-7T (McKinsey/JPM) | McKinsey、JP Morgan | 包含 IT + 设施 |
| 需求增长 | 2030 年数据中心电力需求相较 2023 | +~165% | Goldman Sachs | 相对值,不是绝对值 |
| SAM | 美国 AI 负载共址 / 表后燃气 | TAM 的一部分(未量化) | 分析师推断 | 无公开点估算 |
| SOM | Joulent 近期可触达管线 | 未披露 | 公司未披露 | 仅有 Kilby 2.67 GW 锚定项目 |
多种规模测算口径使用不同单位(TWh、GW、美元、%);Joulent 的 SAM 和 SOM 未公开量化。
[CM008, CM007, CM010, CM009, CM015, CM033]从数据中心总投资逐层收窄到 Joulent 已披露的近期落地规模。
[CM010, CM011, CM015, CM033]不同单位和方法口径下,已发布的数据中心需求与 capex 估算分布。
[CM006, CM007, CM008, CM009, CM010, CM011]2.3 买方细分与采用路径
需求集中在超大规模云和 AI 运营商——Microsoft 最突出——以及托管服务商、REIT 和大型工业用户。经济买方通常是超大规模云厂商, 通过 10-20 年购电协议锁定电力;预算归属基础设施和能源采购团队,它们推动数十亿美元级资本项目。采用路径漫长且重资本: 先选址和控制土地,再落实发电和并网,随后签长期购电协议、完成融资和最终投资决策,最后历经数年建设并送电。Project Kilby 完整展示了这一动作:Chevron 持有与 Microsoft 的 20 年 PPA,Joulent 持有股权期权。AI 运营商越来越多地把算力直接同专用发电同址部署, 以压缩通电时间;速度变成采购差异点,长期合同承购则成为投资案的支点。[CM016, CM017, CM018, CM019, CM020, CM035]
| 细分市场 | 买方 / 用户 | 付款方与预算负责人 | 合同 | 采用备注 |
|---|---|---|---|---|
| 超大规模 AI | Microsoft、其他云厂商 | 超大规模云厂商能源采购 | 10-20 年 PPA / 照付不议 | 锚定需求(Kilby) |
| 托管数据中心 / REIT | Vantage、Equinix 类运营商 | 托管运营商 + 租户 | 多年电力合同 | 将电力成本转嫁给租户 |
| 工业 / 大负载 | 制造业、电气化 | 企业能源团队 | 双边供电 | 相对 AI 需求居次 |
| 开发商 / IPP 包销 | 电力 JV 对手方 | 项目融资发起方 | 股权 + PPA 支持债务 | Chevron / Joulent JV 模式 |
| 电网平衡 | ERCOT / 监管方 | 电费用户 | 电价 / 并网 | 系统性规划约束 |
买方、付款方和预算负责人细分;超大规模云厂商是长期 PPA 下的锚定付款方。
[CM016, CM017, CM018, CM035, CM036]按属性拆分买方、合约结构与采购优先级。
[CM016, CM017, CM018, CM037]2.4 增长驱动与约束
主导驱动因素是 AI 算力激增。相比以往工作负载,AI 耗电高得多,又撞上动辄数年的并网排队;两者合在一起, 正是超大规模客户转向同址发电的核心原因。通电速度和长期承购为新电厂降风险。驱动因素之外,约束也很硬。 Texas Senate Bill 6 和 ERCOT 新的大负荷并网标准,对 75 兆瓦及以上负荷施加研究费、按兆瓦计的财务担保和披露义务; 监管方也在处理集中需求带来的系统压力。资本强度——电厂每兆瓦成本以数百万美元计——限制供给扩张速度,燃气轮机紧缺又增加实体瓶颈, 排放审查升温则压迫新增燃气容量。最后,需求预测本身误差带很宽:AI 采用能否持续、效率提升或融资回撤,都可能显著改变轨迹; 公开估算分歧也凸显了这种不确定性。对尽调而言,约束集合和需求曲线同样重要,因为它决定哪些开发商真能把需求转成已送电、已签约的兆瓦。[CM021, CM022, CM023, CM024, CM025, CM026]
2.5 图表要点
03竞争对手
3.1 竞争格局
Joulent 在一个拥挤且快速变化的赛道中竞争:为 AI 数据中心提供可靠电力。格局横跨四类玩家:Crusoe 等垂直整合的 AI 电力开发商;VoltaGrid、Bloom Energy 等表后和现场供电专家;Vistra、Constellation、Talen 等燃气和核电独立发电商; 以及 Homer City 与 Kiewit 等可比的同址燃气开发商。所有玩家背后都有一个现状替代项:通过受监管公用事业公司做标准并网接入。 这仍是默认选择,但多年排队越来越受限。传统 IPP 现在直接面对分布式和同址电力竞争;超大规模云厂商也可以“内部自建”, 自行签约发电,这对每个第三方开发商都是隐性竞争者。这种广度意味着 Joulent 不能只在单一维度取胜,而要同时拼速度、选址、 燃料和资本。赛道也很流动:核电重启、燃气同址和表后燃料电池,资金都来自同一小群超大规模客户,因此即便技术不同, 竞争者也常在客户上重叠。对尚未完成 FID 的新进入者而言,实际问题不是竞争者是谁,而是谁真能按超大规模客户的时间表交付已送电兆瓦; 在这项测试上,已有运营电厂的既有玩家起跑领先。[CP001, CP002, CP003, CP008, CP012, CP013]
| 竞争对手 | 类型 | 规模 / 融资 | 目标客户 | 产品范围 | 战略方向 |
|---|---|---|---|---|---|
| Crusoe | 垂直整合 AI 电力 | 估值约 $10B;Series E 轮 $1.375B | 超大规模云厂商 / 自有云 | 数据中心开发 + 电力 + AI 云 | IPO 前规模化 |
| VoltaGrid | 表后燃气发电 | 未上市 | 数据中心 | 分布式现场燃气发电 | 模块化快速部署 |
| Bloom Energy | 现场燃料电池 | 上市公司 | 数据中心、工业客户 | 固体氧化物燃料电池发电 | 稳定现场发电 |
| Vistra | 燃气 + 核电 IPP | 大型上市公司 | 电网 + 数据中心 | 市场化发电资产组合 | 核电 + 燃气数据中心合作 |
| Constellation | 核电 IPP | 大型上市公司 | 超大规模云厂商 | 核电 PPA(TMI 重启) | 24/7 零碳电力交易 |
| Talen | 核电 IPP | 上市公司 | 超大规模云厂商(Amazon) | 核电邻近园区 | 同址核电供给 |
| Homer City / Kiewit | 同址燃气开发商 | 约 4.5 GW 项目(PA) | AI 数据中心 | 燃气供电数据中心园区 | 燃气模式直接对标 |
| 电网并网 | 既有公用事业模式 | 受监管 | 所有负荷 | 标准电网供电 | 受排队限制 |
涵盖直接同行、既有玩家、相邻玩家与现有方案;规模数字采用最新披露值。
[CP004, CP008, CP009, CP012, CP013, CP014]3.2 同业画像与融资
最直接的规模化同业是 Crusoe。这家垂直整合的 AI 基础设施公司在 2025 年底完成 $1.375B Series E 后估值约 $10B,且据报道 2026 年还在募 IPO 前融资轮;其数据中心开发、电力采购和 AI 云组合给了它 Joulent 还没有的收入基础。 表后专家 VoltaGrid 和 Bloom Energy 以分布式燃气和燃料电池发电竞争,Vantage、Equinix 等托管开发商则安排专用现场电力。 IPP 阵营中,Vistra 运营大型燃气和核电机组,Constellation 为 Microsoft 重启 Three Mile Island,Talen 与 Amazon 达成核电交易——这些无碳、全天候供给同燃气形成对照。Homer City 在宾夕法尼亚州约 4.5-GW、由 GE Vernova 供能的园区显示,其他玩家也在以吉瓦级复制同址燃气模式。[CP004, CP005, CP006, CP007, CP009, CP010]
| 能力 | Joulent | Crusoe | IPP(Vistra/Constellation) | BTM(VoltaGrid/Bloom) |
|---|---|---|---|---|
| 专属同址发电 | 是(燃气) | 是 | 部分 | 是 |
| 稳定燃料供应渠道 | 是(Chevron) | 部分 | 是 | 不一 |
| 稀缺涡轮机渠道 | 是(GE Vernova) | 部分 | 是(自有资产) | N/A(燃料电池) |
| 已披露超大规模云厂商锚定客户 | 是(Microsoft) | 是 | 是 | 部分 |
| 雄厚资产负债表 / 战略资本 | 是(NGV/Chevron) | 是(VC) | 是 | 不一 |
| 零碳选项 | 有限 | 部分 | 是(核电) | 部分 |
相对能力对比;“是 / 部分 / 有限”反映披露定位,不是审计后的基准。
[CP020, CP016, CP024, CP036]横轴看专注通电速度,纵轴看规模与资本实力。
[CP021, CP020, CP031, CP007]3.3 差异化与定价
Joulent 的差异化不在某个单点功能,而在一揽子组合:Chevron 的燃料供应和资产负债表、GE Vernova 稀缺的涡轮交付槽、 National Grid 战略资本和 Microsoft 锚定客户;单一竞争者很难复制这一拼装。约 $5B 隐含估值低于 Crusoe 约 $10B,差距既反映阶段,也反映 Crusoe 更宽的运营平台。全行业竞争围绕通电速度、选址、稳燃料或涡轮获取, 而非公开价格,因为价格通过双边 10-20 年 PPA 确定;Kilby 案中,Chevron 持有 Microsoft 协议。 稀缺燃气轮机是共同闸门因素;值得注意的是,Joulent 自己的供应商 GE Vernova 也为竞争项目供货,因此设备优势真实存在, 但并不独占。[CP019, CP020, CP021, CP022, CP023, CP024]
| 竞争对手 | 定价模式 | 合同期 | 打包方式 | 备注 |
|---|---|---|---|---|
| Joulent(Kilby) | 通过 JV 签双边 PPA | 20 年 | 同址燃气 + 电网连接 | Chevron 持有 Microsoft PPA |
| Crusoe | 综合成本 / 云服务 | 多年期 | 数据中心 + 电力 + 计算打包 | 垂直一体化 |
| IPP(核电) | PPA / 包销 | 10-20 年 | 稳定零碳电力 | 既有电厂 |
| BTM 专业商 | 容量 + 电量费用 | 多年期 | 模块化现场机组 | 快速部署 |
| 电网并网 | 受监管电价 | 持续 | 标准供电 | 受排队限制 |
定价靠双边长期合同敲定,而不是公开费率表;条款仅作指示。
[CP023, CP025, CP022]用护城河和就绪度 KPI,把 Joulent 与领先同行对标。
[CP021, CP004, CP020, CP025]3.4 护城河耐久性与竞争风险
Joulent 最耐久的护城河是合同:长期 PPA 一旦签下,就形成高转换成本;站址控制、并网权和涡轮交付槽也都是难复制资产。 但耐久性有条件。配电能力和资本化程度有利于 Vistra、Constellation 等既有玩家,它们已经拥有机组和电网关系; 超大规模客户也会在核电、燃气和电网之间多点下注,避免依赖。许多开发商都在复制同址燃气模式,模型正被商品化; 核电和可再生能源加储能等无碳竞争者可能凭排放和长期成本取代燃气,随着可持续审查升温,这一威胁更强。 与竞争者共享的涡轮稀缺,可能让 Joulent 落后于已有运营电厂的核电同业;资本更雄厚的既有玩家也能出价压过单项目新进入者。 护城河确实存在,但很大程度压在 Kilby 完成最终投资决策、并留住伙伴组合上。2028 年首次送电前,护城河的每一层都只是前瞻而非已证明; 投资人必须把这个核心竞争脆弱性同伙伴阵容的强度放在一起权衡。[CP025, CP026, CP027, CP028, CP029, CP030]
| 护城河 / 风险 | 评估 | 耐久度 | 缓释 / 威胁 |
|---|---|---|---|
| 长期 PPA 锁定 | 一旦签约,切换成本抬高 | 高 | 取决于能否到达 FID |
| 涡轮机 / 燃料渠道 | 稀缺,难复制 | 中 | 共用供应商也能支持对手 |
| 场址与并网权 | 绑定具体地点的资产 | 中高 | 监管(SB6)摩擦 |
| 燃气模式商品化 | 大量开发商跟进复制 | 中低 | 靠打包与锚定客户拉开差异 |
| 零碳替代 | 核电 / 可再生能源威胁 | 中 | 排放审查升温 |
| 资本规模相对既有玩家 | 低于 Vistra/Constellation 体量 | 中低 | NGV/Chevron 支持可对冲 |
登记护城河来源和被替代风险,并给出定性耐久度判断。
[CP025, CP026, CP029, CP030, CP031, CP033]四类玩家在发电、资本 / 燃料、设备和客户触达上的能力广度。
[CP020, CP011, CP016, CP036]3.5 图表要点
04财务
4.1 收入来源与定价
Joulent 的收入模型建立在向数据中心和工业用户供电的长期购电协议上。Project Kilby 方面,Chevron 通过 Energy Forge One 子公司持有与 Microsoft 的 20 年 PPA,Joulent 则通过合资公司 50% 股权期权参与,这是其获取项目经济收益的主要路径。 这类协议通常期限 10-20 年,带有照付不议式承诺;价格由双方谈判而非公开,在该细分市场常被引述为每兆瓦时数十美元。 电厂投运后,电力交付、容量可用时才会确认收入;今天 Joulent 在 Kilby 上尚未产生收入,目标 2028 年首次送电。 因此,收入线条结构清晰但数字不透明:合同存在,但价格、调价机制和 Joulent 精确经济份额都未披露。[CI001, CI002, CI003, CI004, CI005, CI006]
| 收入流 | 基础 | 状态 | 交易对手 | 备注 |
|---|---|---|---|---|
| Kilby 电力(PPA) | 20 年 PPA 下的电量 + 容量 | 已签约,运营前 | Microsoft(经 Chevron JV) | Joulent 拥有 50% 股权期权 |
| 新增项目 | 新园区未来 PPA | 储备 / 计划中 | 未披露超大规模云厂商 | 由 NGV 资本支持 |
| 股权 / JV 收益 | 项目现金流分成 | 潜在 | Chevron JV | 取决于 FID |
| 服务 / 开发 | 开发与管理费 | 未披露 | 合作伙伴 | 公开披露不足 |
收入来源已签约但尚未运营;只有 Kilby PPA 有具体披露。
[CI001, CI002, CI003, CI027]| 项目 | 模式 | 典型期限 | 指示性水平 | 置信度 |
|---|---|---|---|---|
| PPA 电量价格 | 双边 $/MWh | 10-20 年 | 数十 $/MWh(区间) | 低 |
| 容量付款 | 固定可用性费用 | 合同期 | 未披露 | 低 |
| 合同期限(Kilby) | 照付不议式 | 20 年 | 已签署 | 高 |
| 调价机制 | 指数挂钩 / 固定阶梯 | 合同期 | 未披露 | 低 |
| 股权回报 | JV 现金流分成 | 项目寿命 | 未披露 | 低 |
变现机制结构清楚,但除合同期限外,量化信息未披露。
[CI004, CI005, CI002, CI006]已签约 PPA 收入如何经 Chevron JV 流入 Joulent 的股权经济性。
[CI002, CI003, CI001]4.2 获客路径、成本结构与单位经济
Joulent 的获客顺序与常规相反:先锁定超大规模锚定承购,再投入资本,Microsoft 协议先于 Kilby 最终投资决策就是例证; 公司依赖 Chevron、GE Vernova、National Grid 等伙伴渠道,而不是传统销售团队。客户获取因此高度集中,首个项目由单一锚定客户支撑; 吉瓦级交易的销售周期横跨数月到数年,涉及选址、并网和签约。成本端,模型极度重资本:Kilby 预计耗资约 $7-9B, 同址燃气发电每兆瓦成本达到数百万美元,GE Vernova 涡轮和 Caterpillar 机组是主要支出项。利润率取决于合同 PPA 收入与天然气燃料、运营成本之间的价差,再用高利用率和长合同摊销前置重资本开支。重要的是,离网、表后设计让 Kilby 避免把电网升级成本社会化,也绕开多年并网排队;这既支撑独立项目经济性,也让其比依赖电网的建设更快转化为收入。[CI008, CI009, CI010, CI011, CI012, CI013]
| 驱动因素 | 对利润率方向 | 依据 | 备注 |
|---|---|---|---|
| 已签约 PPA 收入 | 正向 | 20 年 Microsoft 包销 | 投资级交易对手 |
| 天然气燃料成本 | 负向(可变) | 最大运营成本项 | 天然气价格敞口 |
| 涡轮机 / 设备资本开支 | 负向(固定) | GE Vernova、Caterpillar | 前期集中 |
| 利用率 | 正向 | 需要高容量因子 | 摊薄资本开支 |
| 合同期限 | 正向 | 长期限降低风险 | 20 年 |
单位经济性取决于 PPA 收入减燃料成本后的价差,再靠高利用率和长期限摊薄成本。
[CI014, CI015, CI017, CI018]从已签约 PPA 收入出发,扣除燃料、O&M 和 capex 摊销后,走到项目利润率。
[CI014, CI015, CI016, CI018]4.3 公开牵引与资本充足性
作为私营公司,Joulent 不披露审计收入、ARR 或利润;其公开牵引信号是已签 Microsoft PPA 和 $1.75B 战略投资。 Kilby 2.67 GW 已签约容量在尚无收入时充当标题性前瞻指标。资本方面,National Grid Ventures 于 2026 年 7 月 1 日同意以 $1.75B 投资取得约 35% 股权——隐含估值约 $5B——并称该交易由资产负债表出资,是其创纪录资本计划的增量投入; 公司已为 Ventures 预留约 $1B 近期资本。然而,相比 $7-9B 的 Kilby 建设规模,这笔股权资金偏小, 意味着还需要大量项目融资或伙伴资本,可能是以 20 年 Microsoft PPA 作担保的债务。下一个重大融资触发点是 Kilby 目标在 2026 年底完成的最终投资决策;Chevron 的燃料供应和资产负债表深度显著减轻了 Joulent 独自融资的负担。[CI019, CI020, CI021, CI022, CI023, CI024]
| 项目 | 金额 / 状态 | 来源 | 含义 |
|---|---|---|---|
| 已募战略股权 | $1.75B (Jul 2026) | National Grid Ventures | 约 35% 股权 |
| 隐含估值 | ~$5.0B | 由交易推导 | 收入前估值 |
| Kilby 建设成本 | ~$7-9B | 估算 | 仍需更多资本 |
| 追加融资 | 项目融资 / 合作伙伴(隐含) | 推断 | 以 20 年 PPA 支撑 |
| 下一触发点 | End-2026 FID | 公司 / 合作伙伴 | 进入建设的关口 |
能否撑住 $7-9B 建设成本,关键在于股权注资之外还能叠加多少项目融资和合作伙伴资本。
[CI023, CI024, CI028, CI029, CI030]| 指标 | 公开状态 | 为何重要 | 尽调路径 |
|---|---|---|---|
| 收入 / ARR | 未披露 | 验证变现 | 管理账 |
| 毛利率 | 未披露 | 评估盈利能力 | PPA + 燃料模型 |
| 烧钱 / 运营开支 | 未披露 | 评估现金跑道 | 公司财务报表 |
| 股权结构 / 持股拆分 | 部分披露 | 控制权与回报 | 股东协议 |
| 项目融资条款 | 未披露 | 杠杆与风险 | 债务条款清单 |
几乎所有运营财务数据都未披露,这给一家尚未产生收入的私营开发商留下实质性尽调障碍。
[CI019, CI022, CI034, CI037]围绕已披露和估算的核心财务参数给出区间。
[CI023, CI024, CI012, CI036]示意资本结构:从股权注资桥接到 Kilby 约 $8B 的融资需求(单位 $B)。
[CI028, CI012, CI016, CI030]4.4 财务结论与尽调阻塞点
财务图景更像一个高信心期权,而不是已验证业务。积极面是,一旦投运,Kilby 现金流将由与投资级交易对手签订的 20 年协议锚定, 支撑高收入质量和项目融资杠杆。反面是,资本强度和长建设周期把回报压到后端,也让模型暴露于成本超支和延期; 怀疑者警告,如果 AI 需求或气候政策转向,为数据中心新建大型燃气电厂可能面临搁浅资产和回报风险。最大的尽调阻塞点, 是公开财务几乎完全缺位——收入、利润率、烧钱速度、完整股权结构和项目融资条款都未披露——因此约 $5B 隐含估值是一个尚未产生收入、 由交易推导的标记,押注的是合同管线而非当前现金流。结论:结构有吸引力,经济性未验证;披露缺口必须在承销前补上。 2026 年底最终投资决策是第一个硬检查点,投资逻辑要么明显降风险,要么直接停滞。[CI032, CI033, CI034, CI035, CI036]
4.5 图表要点
05产品与技术
5.1 用客户工作流定义产品
Joulent 的产品不是设备,也不是软件,而是向超大规模数据中心和工业用户交付的稳定电力服务。公司融资、建设、持有并运营发电资产, 再通过长期协议售电;站在客户角度,买到的是有保障的容量,而不是设备。旗舰 Project Kilby 设计供给约 2.67 GW, 用于支撑 Microsoft 的 AI 算力。它解决的问题很急:超大规模客户需要数百兆瓦级稳定电力,但拥堵的公共电网无法按其时间表交付, 并网排队长达数年。Joulent 把发电与算力同址部署在离网、表后园区,并将负荷从 ERCOT 孤岛隔离,等于完全移除输电瓶颈, 缩短从涡轮到服务器的电气路径。放在工作流里看,客户用一座按私人时间表交付的定制电厂,替换了等待多年公共电网容量的过程。[CE001, CE002, CE003, CE004, CE005, CE006]
| 用户任务 | 现有流程 | Joulent 方案 | 可衡量收益 | 限制 |
|---|---|---|---|---|
| 锁定 GW 级稳定电力 | 排入多年电网并网队列 | 表后共址发电 | 通电时间提前数年 | 燃料 + 排放敞口 |
| 选址 AI 训练园区 | 争抢稀缺电网容量 | 把电力带到富气源场址 | 绕开电网瓶颈 | 西得州偏远区位 |
| 保障可靠性 | 依赖电网 + 备用发电机组 | 孤岛自供电园区 | 掌控可用性 | 必须自供电网服务 |
| 管理电力成本 | 承受电网拥堵电价 | 自有发电资产签长期 PPA | 20 年价格确定性 | 气价传导风险 |
| 满足容量增长 | 增量电网升级 | 分阶段加机组 | 随需求扩容 | 单个机组资本开支重 |
这些用例正好对应超大规模云厂商的痛点,主要取舍是区位和燃料敞口。
[CE004, CE006, CE019, CE024, CE027]电力如何从共址燃料生产出来,并送达 AI 算力负载。
[CE006, CE017, CE019, CE024]5.2 资产与产品线地图
发电机组以 GE Vernova 7HA 级重型燃气轮机为核心,配置为联合循环模块,并由 Caterpillar Solar Turbines 机组补充保供和灵活性。 园区位于德克萨斯州 Permian Basin 核心、Reeves County 近 Pecos,占地超过 2,000 英亩;电厂坐落在充足低成本天然气之上, 无需长管线也能取得本地燃料。资产通过合资公司持有,Chevron 承载 20 年 Microsoft 购电协议,Joulent 持有 50% 股权期权——这一结构把实体电厂转化为 Joulent 的经济产品。Kilby 之外,产品线最好理解为更多同址园区的管线,$1.75B National Grid Ventures 资本正是用来播种这些项目;但第二个项目尚未命名。因此,资产地图在 Kilby 的涡轮机组和场址上很具体, 在精确设备数量和未来项目细节上仍然很薄。[CE007, CE008, CE009, CE010, CE011, CE012]
| 模块 / 资产 | 主要用户 | 状态 / 成熟度 | 差异化 | 尽调缺口 |
|---|---|---|---|---|
| GE Vernova 7HA 联合循环机组 | Microsoft 计算负载 | 已订购,尚未开工 | 高效,支持 H2 | 燃机交付档期 |
| Caterpillar Solar Turbines 机组 | 园区保供 / 灵活性 | 规划中 | 模块化,可快速启动 | 确切数量未披露 |
| 孤岛微电网控制(GridOS 级) | 园区运营方 | 设计阶段 | 离网调度 | 2.67 GW 规模集成尚未验证 |
| Permian 气源接口 | 电厂燃料入口 | 场址可控 | 低成本气源就地配套 | 长期天然气合同条款 |
| Kilby 合资 / 股权结构 | Joulent、Chevron | 已签约 | 50% 股权选择权 | 完整经济条款未披露 |
| 未来园区储备 | 未披露超大规模云厂商 | 概念阶段 / 已获资金 | 靠 NGV 资金复制 | 未具名第二个项目 |
Kilby 的燃机资产已经落到具体机型,但数量和后续项目细节仍很薄。
[CE007, CE008, CE011, CE012, CE016]5.3 运营架构与技术机制
Kilby 的架构把已验证组件叠成一个孤岛系统。燃气轮机燃烧 Permian 燃料带动发电机,余热回收蒸汽循环捕捉排气热量驱动蒸汽轮机, 把燃料转电效率显著推高到简单循环之上;GE Vernova 7HA 平台的联合循环效率达到 63-64% 区间,处于商用燃气发电最高水平之一, 并设计为可掺烧氢气,为部分脱碳留出路径。由于园区离网运行,GridOS 类电网编排软件必须持续平衡发电、储能和负荷; 电厂还必须自供频率调节、旋转备用和黑启动能力,而这些通常由公共电网提供。表后交付缩短从涡轮到服务器的电气路径, 降低输电损耗和拥塞电价暴露。运营模型先把资本投进涡轮、余热回收和控制系统,再靠数十年高利用率调度赚钱。 每一层单独看都已验证,但为单一 AI 客户整合到 2.67 GW 规模,是核心技术未知数。[CE013, CE014, CE015, CE016, CE017, CE018]
| 层级 / 组件 | 作用 | 关键依赖 | 主要风险 |
|---|---|---|---|
| 燃气轮机(7HA) | 主力发电 | GE Vernova 供应 + 服务 | 燃机积压订单 / 交付档期 |
| 余热回收蒸汽循环 | 效率提升至 ~63-64% | EPC 集成 | 施工执行 |
| 燃料供应 | 燃烧输入 | Permian 天然气 + Chevron | 气价 / 可得性 |
| 微电网控制 | 平衡发电与负载 | GridOS 级软件 | 孤岛集成成熟度 |
| 电网服务层 | 调频 / 备用 / 黑启动 | 现场设备 | 自供可靠性 |
| 计算互连 | 向服务器送电 | 园区电气设计 | AI 负载可用性 |
单层技术都已验证;集成风险集中在孤岛控制和 EPC 执行。
[CE013, CE014, CE016, CE017, CE040]从共址天然气燃料到 AI 算力负载的分层架构。
[CE005, CE013, CE014, CE016, CE010]5.4 部署、可靠性、支持与路线图
部署路径从目标 2026 年底最终投资决策走向 2028 年首次送电,意味着一个压缩的多年工程、采购和建设周期,峰值用工超过 6,000 人。关键闸门依赖是涡轮可得性:GE Vernova 重型燃气轮机积压订单事实上已经排到 2020 年代末,因此锁定交付槽对达成 2028 年目标的重要性不亚于资本或许可。AI 训练园区需要极高可用性,冗余涡轮模块和现场备用是设计核心;分阶段建设也让容量分块上线, 在完整 2.67 GW 建成前先产生部分收入。长期支持和维护预计将依赖 GE Vernova 和 Caterpillar 的 OEM 服务协议, 而不是内部现场团队。整体看,路线图里程碑清晰,但 FID 之后每个日期仍是目标,不是已承诺、已融资的排期,延期风险真实存在。[CE020, CE021, CE022, CE023, CE024, CE025]
| 日期 / 阶段 | 里程碑 | 状态 | 含义 | 来源依据 |
|---|---|---|---|---|
| 2026 年 6 月 | Joulent 公开发布 | 已完成 | 公司对外亮相 | 公司 / 新闻稿 |
| 2026 年 7 月 | $1.75B NGV 投资 | 已完成 | 获得建设资金 | Business Wire |
| 2026 年底 | Kilby 最终投资决定 | 目标中 | 开工关口 | 分析师 / 合作方 |
| 2027 | EPC / 燃机安装 | 规划中 | 现场部署爬坡 | 推断时间线 |
| 2028 | 首电上线 | 目标中 | 收入起点 | CNBC / 分析师 |
| 2028 年后 | 2.67 GW 全部建成 | 规划中 | 园区完工 | 分析师 |
路线图的里程碑清楚,但 FID 之后所有日期都是目标,并非已承诺排期。
[CE020, CE021, CE022, CE025]5.5 差异化与竞争护城河
Joulent 的主要差异化是通电速度:把发电带到场址,而不是等待输电,把稳定吉瓦级容量交付时间比并网替代方案提前数年。 外围是一道伙伴准入护城河——GE Vernova 涡轮的优先供给、Chevron 的燃料和资产负债表、Microsoft 锚定承购——新进入者很难拼出。 超过 2,000 英亩、具备同址燃气条件的实体场址控制,本身就是稀缺且难复制的资产;Engine No. 1 孵化和 Chris James 的能源转型网络,也给 Joulent 带来年轻开发商少见的资本市场和交易触达能力。关键但书是,表后概念并非专有: 从 Crusoe 到传统 IPP,竞争者都在推进同址命题。因此,能长期保持的优势不是想法,而是先发执行和集成质量。 放在成熟度地图上,组件已经验证;差异化但仍未验证的一层,是全规模孤岛集成。[CE027, CE028, CE029, CE030, CE031]
Joulent 位于依赖网络中心,网络横跨涡轮 OEM、燃料、承购方和监管机构。
[CE011, CE023, CE028, CE032]组件本身成熟;真正差异化、但仍未验证的,是 2.67 GW 规模下的孤岛式集成。
[CE014, CE016, CE029, CE040]5.6 信任、安全、合规与质量控制
作为大型燃烧源,Kilby 的涡轮需要获得联邦和州层面的空气质量许可,覆盖 NOx、CO2 等污染物;脱离 ERCOT 孤岛运行降低了并网审批暴露, 但不让电厂免于环境和安全监管。该设计遭到明确批评:分析师和环保组织认为,如果 AI 需求或气候政策转向, 为数据中心建设大型燃气电厂会带来实质碳排和搁浅资产风险;公司尚未发布公开生命周期排放模型。氢气掺烧提供合规对冲, 但当前掺混比例只能有限降低生命周期排放。可靠性和安全质量主要由 OEM 工程标准和长期服务协议约束,而非独立审计; Kilby 设计、排放建模或可靠性工程尚无第三方评审公开。West Texas 干旱地区的联合循环蒸汽与冷却用水来源, 也是一个未解决的运营和合规问题。简言之,合规义务能识别,但公开材料记录不足。[CE032, CE033, CE034, CE035, CE036, CE037]
5.7 图表要点
06客户
6.1 客户基础与细分
Joulent 瞄准两个客户群:超大规模数据中心运营商和大型工业用电方。现实中,当前客户基础是一个决定性账户——Microsoft; 它同时是买方、用户和付款方,签约约 2 GW 的 Project Kilby 容量以支撑 AI 算力。地理布局集中在德克萨斯州 Reeves County 的 Kilby 园区,核心用例是为无法及时获得电网容量的 AI 训练和推理供电。工业用户被列为目标细分, 但没有披露具体工业客户,也没有公布第二家超大规模客户。因此,除一个账户外,细分故事更多是愿景:市场机会横跨多个垂直领域, 但具体客户现实是一家超大规模云厂商,并且商业上通过 Chevron 合资 PPA 连接。这决定了后续所有关于采用、留存和集中的问题。[CU001, CU002, CU003, CU004, CU005, CU006]
| 分群 | 买方 / 用户 / 付款方 | 用例 | 规模 | 战略价值 | 缺口 |
|---|---|---|---|---|---|
| 超大规模数据中心 | Microsoft(三种角色合一) | AI 计算电力 | ~2 GW 已签约 | 锚定 / 决定性 | 单一账户 |
| 其他超大规模云厂商 | 未披露 | 未来园区 | 储备 | 若落地则高 | 未具名 |
| 大型工业用户 | 未披露 | 稳定工业电力 | 目标分群 | 分散风险 | 无具名客户 |
| 经合作方承接(Chevron JV) | Chevron 承接 PPA | Kilby 购电 | 20 年合同 | 支撑锚定交易 | Joulent 隔了一层 |
除单一超大规模云账户外,分群更多是愿景;只有 Microsoft 是具体客户。
[CU001, CU002, CU006, CU028]客户从电力短缺出发,签下长期合同,再走向潜在多园区扩张。
[CU005, CU013, CU027, CU029]6.2 采用轨迹与需求
采用是合同层面的,还不是运营层面的。Kilby 已签约但尚未供电,因此没有活跃使用、复购或利用率数据;标题指标是已签容量——总计约 2.67 GW,其中 Microsoft 实际预留约 2 GW——而非已部署或已利用兆瓦。首次送电目标为 2028 年,意味着今天不存在任何部署、 账户增长或利用率数据,当前运营容量为零。让预留有分量的是背后需求:Microsoft 更广泛的 AI 建设,包括大型 Fairwater 级园区,创造出 Kilby 合同要满足的算力用电胃口。尽调上,采用轨迹应被理解为一本由单一超大规模客户资本开支计划承销的远期订单; 所有兑现风险都集中在 2026 年底投资决策到 2028 年投运之间的建设期。[CU007, CU008, CU009, CU010, CU011, CU012]
| 指标 | 值 | 日期 | 来源依据 | 置信度 | 含义 | 缺失分母 |
|---|---|---|---|---|---|---|
| 签约容量(总计) | ~2.67 GW | 2026 | 分析师 / 媒体 | 中 | 大额远期订单 | 已部署 = 0 |
| Microsoft 预留量 | ~2 GW | 2026 | 分析师 / 媒体 | 中 | 锚定需求 | 确切条款未披露 |
| 运营容量 | 0 GW | 2026 | 推断 | 高 | 收入前 | 2028 年首电 |
| 具名客户 | 1 | 2026 | 公开记录 | 高 | 集中度 | 无第二个账户 |
| 首电目标 | 2028 | 2026 | CNBC / 分析师 | 高 | 多年等待 | 排期风险 |
采用完全来自远期签约;到 2028 年前,部署量和利用量都为零。
[CU007, CU008, CU009, CU010]从目标容量一路收窄到今天已交付 0 MW;漏斗已全部签约,但尚未交付(MW)。
[CU008, CU009, CU010, CU007]6.3 具名客户证明与参考质量
Microsoft 关系是唯一具名客户证明,由 Chevron 的 Energy Forge One 实体持有的 20 年购电协议支撑,不是试点,也不是意向书, 而是生产级商业承诺。Microsoft 公开谈到为 AI 基础设施锁定专用电力,印证了战略意图;这一证明也很新, 绑定 Joulent 2026 年 6-7 月的发布和投资公告。参考质量在合同确定性上很高,但在运营结果上必然很低, 因为 2028 年首次送电前不可能有结果。Microsoft 之外,没有公开第二个具名客户证明点;协议的定价、爬坡量和服务水平条款仍未披露。 因此,这一证明确认了一个头部超大规模客户承诺二十年,是强信号;但实际表现、单位经济,以及该模式能否吸引更多具名账户,仍全部开放。[CU013, CU014, CU015, CU016, CU017, CU018]
| 客户 | 分群 | 部署 / 用例 | 生产级 / 试点 | 结果 | 限制 |
|---|---|---|---|---|---|
| Microsoft | 超大规模云厂商 | ~2 GW Kilby 电力供 AI 计算 | 生产级承诺(20 年 PPA) | 已签约,尚未交付 | 2028 年前无运营结果 |
| 工业用户(类别) | 工业 | 稳定电力(目标) | 未披露 | N/A | 无具名账户 |
| 第二家超大规模云厂商(储备) | 超大规模云厂商 | 未来共址园区 | 未披露 | N/A | 尚未落地 |
只有 Microsoft 是具名且生产级客户;其他行只是已披露的目标分群,没有具名账户。
[CU013, CU014, CU015, CU018]仅 Microsoft 的合同确定性证据强;其他客户证据缺失。
[CU013, CU017, CU018, CU023]6.4 留存、耐久性与转换成本
耐久性是客户故事最强的一部分。20 年 PPA 在电力送出后提供极长合同寿命和收入可见度;这类协议常见的照付不议结构进一步锁定承诺。 客户转换成本很高,因为电厂与园区同址且专门建设,交易对手很难把负荷转到别处;Microsoft 的投资级信用也降低违约风险。 常规留存指标——净收入留存、总留存、流失和续约队列——全都为空,因为客户尚未上线;没有队列数据时, 合同期限就是主要耐久性信号。必须坦率指出,即使 20 年合同很强,也无法消除建设和交付阶段对关系的风险: 延期或超预算建设可能在项目产生收入前就拉紧协议。耐久性因此在合同上优秀,但在运营上未经验证。[CU019, CU020, CU021, CU022, CU023, CU024]
| 指标 | 值 / null | 分群 | 置信度 | 尽调问题 |
|---|---|---|---|---|
| 合同期限 | 20 年 | 超大规模云厂商 | 高 | 确认续约 / 延长期限条款 |
| 净收入留存 | null(收入前) | 全部 | 高 | 2028 年前不适用 |
| 毛留存 / 流失 | null(收入前) | 全部 | 高 | 投运后跟踪 |
| 交易对手信用 | 投资级(Microsoft) | 超大规模云厂商 | 中 | 确认担保方结构 |
| 切换成本 | 很高(共址) | 超大规模云厂商 | 中 | 评估退出 / 终止条款 |
持续性靠长约和高切换成本支撑;投运前,传统留存指标都是 null。
[CU019, CU021, CU022, CU024]20 年照付不议 PPA 下,已签约容量的耐久性(合同承诺,不是实际使用量)。
[CU019, CU023, CU022]6.5 扩张、集中与怀疑视角
客户集中度极高:实质上一个客户支撑整个旗舰项目,单一客户依赖是最主要商业风险。土地扩张路径是用 National Grid Ventures 资本复制更多超大规模园区的同址模式,但扩张高度依赖伙伴——Chevron 负责燃料和 PPA,GE Vernova 负责涡轮——也受制于吉瓦级交易漫长、 高摩擦的采购周期。怀疑者还给出两个反向视角:他们质疑在气候承诺下,超大规模客户依赖新建燃气电厂能否持久; 也指出 Microsoft 有多种替代电力来源——核电重启、可再生能源和竞争开发商——这限制了 Joulent 的议价能力。 围绕大型“燃气供数据中心”项目的社区和环保反弹,可能进一步压迫客户形象。Microsoft AI 资本开支下行或策略转向, 会直接威胁 Kilby 经济性;因此,走出 Microsoft、实现客户多元化,是降低客户基础风险的关键里程碑。截至 2026 年中, 这一里程碑仍未达成。[CU025, CU026, CU027, CU028, CU029, CU030]
6.6 图表要点
07风险
7.1 按严重度排序的风险概览
Joulent 的风险画像由两股力量主导:执行和集中。最大单一风险,是在压缩的 2026-2028 年时间表上交付一座 $7-9B、 吉瓦级燃气电厂,且首次送电预计要到 2028 年——这意味着任何抵消性收入出现前,会有数年的建设期无收入暴露。 资本强度放大下行,因为数十亿美元建设一旦超支,会直接侵蚀股权回报。执行之外,客户集中也是顶级风险:实质上一个客户 Microsoft 支撑旗舰项目。多数类别的剩余风险仍高,因为项目处于施工前且披露有限;缓释成熟度也不均衡——强合同和强伙伴缓释, 同弱公开透明度和未验证运营并存。按剩余严重度排序,优先风险是执行 / 施工、客户集中、涡轮供给、融资缺口、德克萨斯大负荷监管演变, 以及碳排 / 搁浅资产暴露。下文逐一审视,并给出可监控、会打破投资逻辑的触发因素。[CR001, CR002, CR003, CR004, CR005, CR006]
影响大而缓释成熟度低的项,剩余严重性最高——碳排和执行风险最突出。
[CR001, CR004, CR016, CR036]7.2 监管与法律风险
德克萨斯州监管背景因 Senate Bill 6 (2025) 而改变,该法收紧了大型电力负荷接入电网的规则;ERCOT 的大负荷流程(Project 58481)则对约 75 MW 及以上负荷施加研究费、披露和担保要求。SB6 还赋予 ERCOT 和公用事业公司在电网承压时削减或断开大型负荷的扩大权限; 法律分析师也提示,数据中心负荷面临更高披露和需求透明义务。Joulent 的离网、表后设计因园区负荷孤岛运行, 降低但没有完全消除这些并网规则暴露。另一个维度是,大型燃气轮机的空气质量许可(NOx 和 CO2)是必须完成且耗时的步骤; 大型化石能源项目存在环境诉讼或许可挑战这一现实法律风险;德克萨斯州西部干旱地区的用水许可,也给联合循环冷却增加依赖。 联邦能源监管还可能影响同址和批发电力安排。风险登记按严重度排序,但 Kilby 具体申报文件尚未公开,因此剩余暴露只能部分量化。[CR007, CR008, CR009, CR010, CR011, CR012]
| 规则 / 许可 / 案件 | 管辖区 | 状态 | 发生概率 | 严重性 | 缓释措施 | 剩余风险 | 尽调路径 |
|---|---|---|---|---|---|---|---|
| ERCOT 大负荷并网(58481) | Texas / ERCOT | 已生效 | 高 | 高 | 离网设计;遵守研究 / 安全要求 | 中 | 确认并网策略 |
| Texas SB6 大负荷规则 | Texas | 已生效(2025) | 高 | 中 | 法律顾问;需求披露 | 中 | 审查 SB6 合规计划 |
| 空气质量许可(NOx/CO2) | 联邦 + Texas | 必需 | 高 | 中 | 标准许可流程 | 中 | 获取许可申请 |
| 用水许可 | Texas | 必需 | 中 | 中 | 水源方案 | 中-高 | 审查水权 |
| 环境诉讼 / 异议 | 联邦 / 州 | 可能 | 低-中 | 中 | 完备许可流程 | 中 | 跟踪申报文件 |
| 排放 / 碳政策收紧 | 联邦 / 州 | 正在出现 | 中 | 高 | 可用 H2 的涡轮机 | 高 | 评估政策敞口 |
监管敞口真实存在,但离网设计缓释了一部分;各行按严重性排序,许可披露完成前,覆盖仍不完整。
[CR007, CR008, CR012, CR014, CR039]7.3 运营与供应链风险
运营风险集中在燃机供应。GE Vernova 重型燃气轮机积压订单实际上已经排到 2020 年代末,全球需求强劲带来交付档期风险,可能拖慢 Kilby 时间表。建设期风险包括高峰超过 6,000 名工人时劳动力能否到位、EPC 协调以及西得州天气。建成之后,2.67 GW 微电网要孤网运行,必须自己提供电网通常承担的调频、备用和黑启动服务,可靠性风险随之上升;单站电厂也意味着停机或设备故障会直接威胁锚定客户供电。即使靠近 Permian,天然气供应中断或价格飙升仍会压住运营和利润率。更麻烦的是,Kilby 这套具体设计还没有公开的可靠性、安全或质量审计记录,投运前运营风险大多无法量化。运营风险登记表按严重程度排序,燃机交付、建设超支和孤网可靠性排在最前。[CR016, CR017, CR018, CR019, CR020, CR021]
| 失效模式 | 发生概率 | 严重性 | 缓释成熟度 | 剩余敞口 | 未解决缺口 |
|---|---|---|---|---|---|
| 涡轮机交付延误 | 中 | 高 | 合作伙伴订单已就位 | 中-高 | 确认交付档期 |
| 施工延误 / 超支 | 中 | 高 | EPC + 有经验的合作伙伴 | 高 | 未公开进度细节 |
| 孤网可靠性不足 | 中 | 高 | 冗余模块(设计) | 中-高 | 2.67 GW 规模尚未验证 |
| 天然气供应中断 | 低-中 | 中 | 靠近 Permian | 中 | 燃料合同条款 |
| 单站点停运 | 低 | 高 | 冗余(规划中) | 中 | 无运营记录 |
运营风险集中在涡轮机供应、施工,以及尚未验证的孤网可靠性;各行按严重性排序。
[CR016, CR017, CR018, CR019, CR021]运营风险和集中度风险如何经由利润率、融资传导至估值。
[CR003, CR030, CR035, CR041]7.4 伙伴与依赖风险
Joulent 处在一张高度集中的依赖网络中心。Chevron 依赖度高——它持有 Microsoft PPA 并供应燃料,承诺本身就是承重梁;GE Vernova 在燃机和电网编排技术两端同样是高依赖。National Grid Ventures 是关键资本提供方,任何承诺变化都会拉紧融资;Microsoft 又同时是关键客户和需求理由,进一步放大集中度。ERCOT、Texas PUC 和环保机构等监管方是外部依赖,能直接卡住项目。JV 结构意味着 Joulent 的经济性取决于其无法完全控制的伙伴履约;燃料、燃机、资本或包销任一关键伙伴失手,都可能让旗舰项目停摆。人员侧,执行要靠一个年轻组织交付首个同等规模超级项目;公司依赖创始人兼 CEO Chris James,也需要足够深的项目开发和 EPC 管理梯队。各行按伙伴或执行失败的严重程度排序。[CR023, CR024, CR025, CR026, CR027, CR028]
| 依赖项 | 交易对手 | 角色 | 集中度 | 失效情景 | 严重性 | 缓释措施 | 剩余风险 |
|---|---|---|---|---|---|---|---|
| 燃料 + PPA | Chevron | 燃料供应,PPA 持有人 | 高 | Chevron 退出 / 履约不及预期 | 高 | 长期合资条款 | 中-高 |
| 涡轮机 + 控制系统 | GE Vernova | 设备 + GridOS | 高 | 交付 / 技术失败 | 高 | OEM 服务协议 | 中 |
| 资本 | National Grid Ventures | 领投方 | 高 | 承诺减少 | 中 | 已承诺 $1.75B | 中 |
| 承购 / 需求 | Microsoft | 锚定客户 | 很高 | 合同流失 / 资本开支削减 | 高 | 20 年 PPA | 中-高 |
| 监管 | ERCOT / Texas PUC | 审批 | 中 | 不利裁决 | 中 | 合规 | 中 |
每项关键输入都押在高度集中的合作伙伴身上;各行按合作伙伴失效的严重性排序。
[CR023, CR024, CR025, CR026, CR027]| 角色 / 职能 | 依赖或缺口 | 发生概率 | 严重性 | 缓释措施 | 尽调路径 |
|---|---|---|---|---|---|
| CEO / 创始人(Chris James) | 关键人依赖 | 中 | 中 | 经验丰富的团队 | 评估人才梯队 |
| 项目开发领导层 | 巨型项目执行能力 | 中 | 高 | 合作伙伴专长 | 审查团队履历 |
| EPC / 施工管理 | 现场执行能力 | 中 | 高 | 承包商选择 | 核验 EPC 合同 |
| 运营 / 电厂人员配置 | 孤网电厂运营 | 中 | 中 | OEM 支持 | 确认 O&M 计划 |
| 监管 / 法律团队 | SB6 / ERCOT 路径推进 | 低-中 | 中 | 外部法律顾问 | 确认顾问团队厚度 |
执行成败取决于一支年轻团队能否交付首个同等规模的巨型项目;各行按严重性排序。
[CR001, CR002, CR018, CR022]Joulent 的每项关键投入,都依赖四个高度集中的合作方和监管方。
[CR023, CR024, CR025, CR027]7.5 财务、碳和缓释触发点
从财务看,已募得的 $1.75 billion 股权资金相对 $7-9 billion 建设规模偏小,意味着还要承担大量追加融资风险;长期收入前阶段也会在首笔收入前持续烧钱。天然气价格上涨,或 PPA 定价固定且缺少充分传导,都会挤压利润率;以 20 年 PPA 加杠杆虽可行,却让模型暴露于利率和信用环境。几乎没有公开财务数据,本身就是尽调风险,限制独立验证;AI 数据中心资本开支下行会同时削弱需求和估值依据。缓释最弱的尾部风险是碳:批评者警告,如果 AI 需求或气候政策转向,为数据中心新建大型燃气电厂会有搁浅资产风险;环保组织强调气候张力;超大规模云厂商的气候承诺也可能长期压迫客户限制燃气电力。20 年 PPA、投资级包销方、强伙伴和可用氢燃机等关键缓释因素能降低风险,但不能消除风险。最清晰的论点破裂触发点,是 FID 滑过 2026 年底、Microsoft 合同丢失,或重大成本爆表。[CR030, CR031, CR032, CR033, CR034, CR035]
7.6 图表
08估值
8.1 投资论点与反论点
投资论点是:AI 数据中心共址电力稀缺且快速增长,Joulent 已是早期领先者;已签署的 Microsoft 20 年购电协议和 $1.75 billion 战略投资验证了需求,也降低了锚定项目风险。特权伙伴通道——Chevron 提供燃料和包销,GE Vernova 提供燃机,National Grid 提供资本——拼成了一个新进入者很难复制的生态优势。反论点同样清楚:Joulent 仍无收入、单项目、单客户,约 $5 billion 估值押的是执行期权;一旦执行滑坡,资本强度、碳暴露和建设风险都可能侵蚀回报。两种判断押的是同一组变量:如果 Kilby 按期完成最终投资决策和首批供电,且客户不变,正论点成立;任一项失败,反论点成立。因此,估值争论本质上是贴了价格标签的执行争论。[CV001, CV002, CV003, CV004, CV005, CV006]
| 论点 | 什么会改变判断 |
|---|---|
| 稀缺 AI 电力市场的早期领导者 | 竞争对手扩张更快,或需求降温 |
| 20 年 Microsoft PPA 验证需求 | PPA 重新谈判或流失 |
| 优势合作伙伴生态 | 合作伙伴承诺减弱 |
| 尚无收入、单项目、单客户 | 第二个客户 / 项目落地 |
| 资本密集且有碳风险 | 执行顺利且政策稳定 |
乐观与悲观情景都取决于少数几个变量:执行、客户广度和政策。
[CV001, CV002, CV004, CV005]市场规模和签约验证支撑判断,但风险和溢价估值压低结论,最终落到“观望”。
[CV001, CV002, CV009, CV010]8.2 建议、置信度和立场
综合看,建议是持续跟踪:今天的 Joulent 更像一个高质量期权,而不是已经可以承销的业务。置信度为中,最大约束是几乎没有公开财务数据,无法独立承销收入质量、利润率或回报。风险评级为高,由执行、客户集中和资本强度驱动;估值立场为溢价,因为约 $5 billion 标记已经计入尚未证明的成功执行。可支持的姿态,是在投入资本前跟踪 2026 年底最终投资决策和 2028 年首批供电里程碑,并随着每一步降险重新定价。与其他选择相比,信息不透明让直接投资为时过早;但若直接放弃,又低估了一个真正强伙伴、押注稀缺资源的优质期权。跟踪是穿针引线的选择:保留上行参与权,不预先支付完整执行溢价,同时定义足以支持更坚定立场的具体里程碑。[CV007, CV008, CV009, CV010, CV011, CV039]
| 建议 | 置信度 | 风险评级 | 估值立场 | 决策含义 |
|---|---|---|---|---|
| 观察 | 中 | 高 | 溢价 | 承诺投入前跟踪里程碑 |
| (相对)投资 | - | - | - | 信息不透明,投资还早 |
| (相对)放弃 | - | - | - | 直接放弃会低估强期权 |
| 重估触发项 | - | - | - | FID + 首次发电进展 |
净建议是观察,置信度中,估值立场为溢价;仍需等待执行里程碑兑现。
[CV007, CV008, CV009, CV010]IC 风格评分卡:市场和客户验证强,但经济性未经验证、高风险、披露不足抵消了优势。
[CV041, CV009, CV010, CV038]8.3 融资与估值背景
2026 年 7 月 National Grid Ventures 投资——以 $1.75 billion 换约 35%——对应约 $5 billion 投后估值。这是战略少数股权投资,不是控股交易,因此会影响优先权和治理动态,也可能留下稀释或优先权悬挂,压住未来普通股回报。入场纪律面临挑战,因为估值立在已签约管线而非当前盈利上:公开证据支持交易存在和规模,却没有独立估值交叉验证;这个标记是私募一级发行价格,不是经过市场检验或流动性背书的价格。尽调用语里,这个价格应理解为执行期权的溢价,而不是现金流倍数。若投资人按这个水平或更高价格买入,就是在承销首个同等规模超级项目完美交付;因此,必须先核实 $7-9 billion 建设背后的合同经济性和融资计划,才能把 ~$5 billion 视为底价,而不是乐观标记。[CV012, CV013, CV014, CV015, CV016, CV017]
8.4 牛市、基准和熊市情形
结果异常二元,因为价值集中在单一超级项目。牛市情形下,Kilby 按时执行,Joulent 用更多园区复制共址模式,成为价值远高于 $5 billion 的多 GW 电力平台。基准情形概率最高,因为需求已经签约:Kilby 建成,但有一定延误或成本超支,Joulent 仍是有价值但单资产的开发商,估值接近当前标记。熊市情形下,最终投资决策滑坡或 Microsoft 需求转弱,成本超支咬住回报,股权价值跌破入场价。下行触发点包括 FID 滑坡、PPA 重谈、资本开支冲破 $9 billion,以及 AI 资本开支收缩。概率信号偏向基准情形,但执行尾部风险实质存在,估值区间因此很宽——从熊市情形下股权受损,到成功扩张后达到当前标记数倍。比起任何单一点估计,这个跨度才是 Joulent 估值最诚实的总结。[CV018, CV019, CV020, CV021, CV022, CV023]
| 情景 | 假设 | 估值 / 回报逻辑 | 关键风险 | 概率信号 |
|---|---|---|---|---|
| 乐观 | Kilby 准时交付 + 复制扩张 | 多 GW 平台,远高于 $5B | 执行、资本 | 较低 |
| 基准 | 有一定延误 / 超支,单一资产 | 接近当前 ~$5B 估值 | 成本、集中度 | 较高 |
| 悲观 | FID 延误 / 需求转弱 | 股权减值至当前估值以下 | 超支、PPA 流失 | 中等尾部 |
结果围绕单一巨型项目呈异常二元;基准情景接近当前估值。
[CV018, CV019, CV020, CV021, CV022]| 触发因素 | 阈值 | 如何传导到投资论点 | 行动含义 |
|---|---|---|---|
| FID 延误 | 晚于 2026 年底 | 收入推迟、成本上升 | 重新测算 / 暂停 |
| Microsoft PPA 丢失 | 退出 / 重谈 | 失去需求锚 | 投资论点破裂 |
| 成本失控 | 超过 $9B | 侵蚀股权回报 | 重估估值 |
| AI capex 下行 | 超大规模云厂商削减支出 | 需求和可比估值转弱 | 下调评级 |
| 碳政策冲击 | 新增碳成本 | 搁浅资产风险 | 下修下行估值 |
每个触发因素都有可监测阈值,并会直接传导到估值判断。
[CV035, CV021, CV036, CV005]股权价值对关键价值驱动因素的百分比敏感性示意(方向性)。
[CV021, CV023, CV036, CV040]围绕约 $5B 隐含入场估值,牛市、基准、熊市情形下估值区间很宽(示意,$B)。
[CV012, CV018, CV019, CV020]8.5 可比公司、退出准备度和尽调
可比公司给当前标记框定区间。Crusoe Energy 是可比的 AI 数据中心电力和算力开发商,2025 年末轮次估值约 $10 billion;数据中心基础设施资产交易大约在 15-19x EBITDA;Vistra、Constellation、Talen 等受数据中心需求影响的上市独立发电商提供公开市场参照,而 2025-2026 年 AI 电力基础设施的私募战略轮集中在数十亿美元估值。合在一起看,Joulent 的 ~$5 billion 标记在主题区间内,但对一个无收入、单资产开发商而言偏贵;每个可比对象也都有局限:Crusoe 已有算力收入,IPP 是运营中的公用事业,基础设施倍数假设资产已经产生现金流。对 Kilby 做 DCF 会取决于 PPA 定价、天然气成本、利用率和折现率,这些大多未披露。退出准备度还早——无收入和单资产集中限制近期流动性——但 IPO、战略出售或 National Grid 买断都可行。最终尽调问题集中在财务、股权结构、PPA 经济性和项目融资计划;怀疑者也有道理,主题驱动估值可能跑在基本面前面。[CV024, CV025, CV026, CV027, CV028, CV029]
| 可比对象 | 指标 | 倍数 / 估值 / 状态 | 参考价值 | 局限 |
|---|---|---|---|---|
| Crusoe Energy | 私募轮估值 | ~$10B(2025 年底) | AI 电力 / 计算开发商 | 已有计算收入 |
| DC 基础设施 M&A | EV/EBITDA | ~15-19x | 资产类别倍数 | 假设资产已产生现金流 |
| 上市 IPP(Vistra/Constellation/Talen) | 公开市场 | 不一 | DC 需求敞口 | 运营型公用事业公司,不是开发商 |
| AI 电力私募融资 | 轮次估值 | 数十亿美元 | 主题定价 | 阶段不一 |
| Joulent(标的) | 隐含轮次 | 约 $5B(Jul 2026) | 直接可比 | 尚无收入、单一资产 |
可比对象显示,约 $5B 处在同一主题区间内,但对一个尚无收入的开发商而言偏贵;覆盖范围只是最相关参照中的一组样本。
[CV024, CV025, CV026, CV027, CV028]8.6 图表
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。关键财务、法律、技术和合同事实仍未公开;作出任何投资决定前, 应直接向管理层和原始文件核验。
证据索引
| 编号 | 陈述 | 可信度 | 来源 |
|---|---|---|---|
| CO001 | Joulent is a Houston, Texas-based, technology-driven energy company that develops large-scale power infrastructure for AI data centers and industrial users. | 高 | SO001, SO002, SO003 |
| CO002 | Joulent publicly launched in June 2026. | 中 | SO002, SO003 |
| CO003 | Joulent was developed over roughly three years by investment firm Engine No. 1 in collaboration with GE Vernova. | 中 | SO002, SO004 |
| CO004 | Chris James is founder and chief executive of both Engine No. 1 and Joulent. | 中 | SO002, SO005 |
| CO005 | Engine No. 1 was founded by Chris James in 2020 and won three ExxonMobil board seats in 2021. | 中 | SO006, SO005, SO007 |
| CO006 | On July 1, 2026, National Grid Ventures agreed to invest $1.75 billion in Joulent. | 高 | SO008, SO009, SO010 |
| CO007 | The National Grid Ventures investment buys an approximately 35% stake in Joulent. | 高 | SO011, SO008, SO009 |
| CO008 | The National Grid Ventures investment implies a Joulent valuation of approximately $5 billion. | 中 | SO011, SO012 |
| CO009 | Joulent's flagship initiative is Project Kilby, a 2.67-gigawatt co-located power campus in West Texas. | 高 | SO013, SO014, SO015 |
| CO010 | Project Kilby sits on a 2,000-plus acre site near Pecos in Reeves County, Texas, in the Permian Basin. | 中 | SO014, SO016 |
| CO011 | Project Kilby will supply electricity to a Microsoft-operated data center under a 20-year agreement. | 高 | SO013, SO017, SO018 |
| CO012 | Chevron, through its subsidiary Energy Forge One LLC, signed the 20-year power purchase agreement with Microsoft for Kilby. | 高 | SO017, SO013 |
| CO013 | GE Vernova is Joulent's turbine technology partner for its power projects. | 中 | SO004, SO019 |
| CO014 | Caterpillar's Solar Turbines division provides supplemental generation capacity for Kilby. | 中 | SO019 |
| CO015 | Joulent markets an 'Across-the-Meter' co-located power model that combines gas generation, battery storage and renewables. | 中 | SO002, SO010 |
| CO016 | Project Kilby is projected to cost roughly $7 billion to $9 billion. | 中 | SO015, SO016 |
| CO017 | Kilby targets a Final Investment Decision by the end of 2026. | 中 | SO016, SO017 |
| CO018 | Kilby targets first power delivery in 2028. | 中 | SO013, SO016 |
| CO019 | Joulent holds a 50% equity option in the Project Kilby joint venture. | 中 | SO016 |
| CO020 | Kilby is expected to generate more than $10 billion in state and local tax revenue and support roughly 2,000 jobs. | 中 | SO015, SO016 |
| CO021 | National Grid Ventures is the commercial, non-regulated arm of National Grid plc. | 高 | SO020, SO008 |
| CO022 | National Grid plc is undertaking the largest capital-investment program in its history, pledging at least £70 billion from FY26 to FY31. | 低 | SO021 |
| CO023 | Kirkland & Ellis advised Joulent on the National Grid investment and the Chevron partnership. | 高 | SO022, SO008 |
| CO024 | Joulent's custom 'power stack' manages energy sources and grid interconnection for reliable power at industrial scale. | 中 | SO001 |
| CO025 | Joulent is a privately held company. | 中 | SO002, SO003 |
| CO026 | Joulent's leadership beyond CEO Chris James has not been disclosed in detail publicly. | 低 | SO002 |
| CO027 | Microsoft is the anchor customer for Joulent's first project. | 高 | SO013, SO023 |
| CO028 | The National Grid capital is intended to fund Project Kilby and additional large-scale power projects. | 中 | SO008, SO010 |
| CO029 | In January 2026, Engine No. 1, Chevron and GE Vernova announced plans to co-locate up to roughly 4 GW of natural-gas plants for US data centers. | 中 | SO004, SO005 |
| CO030 | Joulent's headquarters is in Houston, Texas. | 中 | SO001, SO003 |
| CO031 | Joulent emerged from a multi-year incubation by Engine No. 1, surfacing publicly as a standalone company in 2026. | 中 | SO002, SO004 |
| CO032 | Skeptics warn that large new natural-gas plants built for data centers carry execution, emissions and stranded-asset risks. | 中 | SO024 |
| CO033 | National Grid earmarked roughly $1 billion of near-term capital specifically for National Grid Ventures projects. | 低 | SO021 |
| CO034 | The National Grid-Joulent investment is described as balance-sheet funded and incremental to National Grid's existing investment plan. | 低 | SO021, SO008 |
| CO035 | Joulent positions itself around delivering energy at the 'speed and scale of American innovation.' | 中 | SO001, SO002 |
| CO036 | CEO Chris James frames AI-era leadership as delivering energy and compute the fastest, most reliably and at the lowest cost. | 中 | SO002 |
| CO037 | The investment gives National Grid exposure to fast-growing US 'large load' contracted-power demand. | 中 | SO008, SO010 |
| CO038 | Joulent's exact employee headcount is not publicly disclosed. | 低 | SO002 |
| CO039 | Kilby's 2.67 GW of capacity is roughly equivalent to the electricity needed for about two million homes. | 中 | SO014, SO016 |
| CO040 | Microsoft's adjacent Pecos data-center campus is planned to add roughly 2 GW of compute load over five to seven years. | 中 | SO025, SO013 |
| CM001 | The relevant market is the supply of large-scale, contracted electricity and power infrastructure for 'large load' data centers, especially AI compute campuses. | 高 | SM001, SM002, SM003 |
| CM002 | Included spend covers new generation capacity, grid interconnection, co-located/behind-the-meter plants and long-term power purchase agreements; it excludes IT hardware such as GPUs, servers and cooling. | 中 | SM002, SM004 |
| CM003 | Adjacent markets include grid-scale independent power producers, regulated-utility PPAs, on-site backup generation and emerging nuclear small modular reactors. | 中 | SM003, SM005 |
| CM004 | The status-quo substitute is a standard grid interconnection through a regulated utility, which increasingly faces multi-year queues. | 高 | SM002, SM003, SM006 |
| CM005 | Data-center-driven behind-the-meter and co-located generation is emerging as the fastest route to power for hyperscalers avoiding interconnection delays. | 中 | SM002, SM004, SM007 |
| CM006 | Gartner projects data-center electricity consumption to grow about 26% in 2026. | 高 | SM008, SM009 |
| CM007 | Data-center power demand is estimated at roughly 132 GW in 2026, rising toward about 290 GW by 2030. | 中 | SM008, SM009 |
| CM008 | The IEA projects global data-center electricity use will roughly double to about 945 TWh by 2030. | 高 | SM010, SM009 |
| CM009 | Goldman Sachs projects data-center power demand will increase roughly 165% by 2030 versus 2023. | 高 | SM011, SM012, SM013 |
| CM010 | McKinsey estimates data centers could require roughly $6.7 trillion of global capital investment by 2030. | 高 | SM014, SM015, SM016 |
| CM011 | JP Morgan and McKinsey have forecast on the order of $5-7 trillion of global data-center investment this decade. | 中 | SM017 |
| CM012 | Hyperscaler capital expenditure is estimated at roughly $690 billion in 2026, with power availability a key bottleneck. | 中 | SM009 |
| CM013 | Microsoft alone guided to roughly $190 billion of capital expenditure in fiscal 2026, much of it AI infrastructure. | 中 | SM018 |
| CM014 | A large share of new data-center capex flows to power generation, interconnection and electrical systems rather than IT hardware. | 中 | SM014, SM019 |
| CM015 | Only a fraction of total data-center demand is addressable by any single co-located power developer, so Joulent's serviceable market is far smaller than headline demand figures. | 低 | SM003, SM004 |
| CM016 | The primary buyers are hyperscale cloud and AI operators such as Microsoft, alongside colocation providers and large industrial users. | 中 | SM002, SM004 |
| CM017 | The economic payer is typically the hyperscaler, which contracts power under long-dated (10-20 year) purchase agreements. | 中 | SM020, SM021 |
| CM018 | Budget ownership sits with hyperscaler infrastructure and energy-procurement teams managing multi-billion-dollar capex programs. | 中 | SM018, SM009 |
| CM019 | The adoption path runs from site selection to generation and interconnection, then a power agreement, construction and energization over several years. | 中 | SM006, SM022, SM003 |
| CM020 | AI operators increasingly co-locate compute with dedicated generation to compress time-to-power from years to a shorter development cycle. | 中 | SM002, SM004 |
| CM021 | The dominant demand driver is the surge in AI compute, which is far more power-intensive than prior workloads. | 高 | SM011, SM009 |
| CM022 | Grid interconnection delays—often multiple years—are a central driver pushing customers toward co-located power. | 高 | SM001, SM003 |
| CM023 | Speed-to-power is a decisive competitive driver, favoring gas generation that can be built faster than new transmission. | 中 | SM002, SM004 |
| CM024 | Texas's SB6 and ERCOT's new large-load interconnection standards impose fees, financial security and disclosure requirements on 75-MW-plus loads. | 高 | SM006, SM022, SM003 |
| CM025 | High capital intensity—power plants costing millions of dollars per megawatt—constrains how fast new supply can scale. | 中 | SM019, SM023 |
| CM026 | Emissions scrutiny and sustainability commitments are a growing adoption constraint on new gas-fired data-center power. | 中 | SM024 |
| CM027 | Gas-turbine supply is tight, adding a physical constraint on how quickly co-located capacity can be delivered. | 中 | SM007, SM003 |
| CM028 | Utility and grid-operator planning strain from concentrated large loads is a systemic constraint regulators are actively addressing. | 中 | SM022, SM006 |
| CM029 | Nuclear—including restarts and small modular reactors—is a competing supply path for round-the-clock data-center power. | 中 | SM005, SM003 |
| CM030 | Traditional independent power producers face new competition from distributed and co-located power in serving data centers. | 高 | SM003, SM002 |
| CM031 | Fuel-cell and hybrid on-site solutions are an alternative behind-the-meter approach for some operators. | 中 | SM025, SM004 |
| CM032 | Market-sizing estimates diverge widely because they mix units—gigawatts of capacity, terawatt-hours of consumption and trillions of dollars of capex. | 中 | SM008, SM010, SM014 |
| CM033 | Public sources do not quantify Joulent's specific serviceable or obtainable market share. | 低 | SM004 |
| CM034 | Demand forecasts carry high uncertainty because AI adoption, efficiency gains and financing conditions could all shift the trajectory. | 中 | SM010, SM011 |
| CM035 | The US, and Texas in particular, is a focal geography for near-term large-load power demand. | 中 | SM001, SM006 |
| CM036 | Microsoft's Pecos/Kilby commitment illustrates hyperscaler willingness to underwrite dedicated, long-term co-located power. | 高 | SM026, SM021 |
| CM037 | Contracted, long-duration offtake de-risks new generation and is central to the market's investment case. | 中 | SM020, SM021 |
| CP001 | The competitive landscape spans vertically integrated AI-power developers, behind-the-meter specialists, gas and nuclear independent power producers, and the status-quo grid interconnection. | 高 | SP001, SP002, SP003 |
| CP002 | Traditional independent power producers increasingly face competition from distributed and co-located power in serving data centers. | 高 | SP002, SP001 |
| CP003 | Hyperscalers can also 'build internally' by contracting their own generation or utility deals, an implicit competitor to third-party developers. | 中 | SP003, SP004 |
| CP004 | Crusoe is a vertically integrated AI-infrastructure company that reached a roughly $10 billion valuation. | 高 | SP005, SP006 |
| CP005 | Crusoe closed a $1.375 billion Series E funding round in late 2025. | 高 | SP007, SP005, SP008 |
| CP006 | Crusoe has been reported to be raising a pre-IPO funding round in 2026. | 中 | SP009 |
| CP007 | Crusoe combines data-center development with its own power sourcing, competing directly on integrated speed-to-power. | 中 | SP007, SP009 |
| CP008 | VoltaGrid provides distributed behind-the-meter natural-gas power for data centers. | 中 | SP003 |
| CP009 | Bloom Energy supplies solid-oxide fuel cells as on-site generation for data centers. | 中 | SP010, SP003 |
| CP010 | Vantage Data Centers and Equinix are hyperscale/colocation developers arranging dedicated on-site power. | 中 | SP003 |
| CP011 | Behind-the-meter and fuel-cell approaches compete with co-located gas on speed and siting flexibility. | 中 | SP003, SP010 |
| CP012 | Vistra operates a large gas-and-nuclear generation fleet and is expanding data-center ties. | 中 | SP011, SP012 |
| CP013 | Constellation Energy is a leading nuclear operator pursuing data-center power deals, including a Three Mile Island restart for Microsoft. | 高 | SP004, SP013 |
| CP014 | Talen Energy has signed nuclear power arrangements with hyperscalers such as Amazon. | 中 | SP014, SP004 |
| CP015 | Constellation and Vistra shares have rallied as grid operators accelerate data-center deals. | 中 | SP015, SP016 |
| CP016 | Nuclear IPPs offer carbon-free, round-the-clock power, a differentiated posture versus gas developers. | 中 | SP004, SP013 |
| CP017 | Homer City, with Kiewit, is developing a roughly 4.5-GW natural-gas-powered AI data-center campus in Pennsylvania using GE Vernova turbines. | 高 | SP017, SP018 |
| CP018 | The Homer City project shows other developers pursuing the same co-located gas model at gigawatt scale. | 中 | SP017, SP019 |
| CP019 | GE Vernova's turbines power multiple competing data-center projects, so Joulent's key supplier also enables rivals. | 中 | SP020, SP018 |
| CP020 | Joulent differentiates on a bundled package of Chevron fuel and capital, GE Vernova turbines, National Grid capital and a Microsoft anchor. | 中 | SP021, SP022 |
| CP021 | Joulent's roughly $5 billion implied valuation sits below Crusoe's ~$10 billion, reflecting Crusoe's broader, revenue-generating platform. | 中 | SP005, SP022 |
| CP022 | Most peers compete on speed-to-power, siting and access to firm fuel or turbines rather than on price alone. | 中 | SP002, SP003 |
| CP023 | Pricing across the segment is set through bilateral long-term PPAs rather than published rate cards. | 中 | SP023, SP024 |
| CP024 | Access to scarce gas turbines is a key competitive gating factor across gas-based developers. | 中 | SP025, SP020 |
| CP025 | Long-dated PPAs (10-20 years) create high switching costs once a hyperscaler commits to a power partner. | 中 | SP023, SP024 |
| CP026 | Site control, interconnection rights and turbine delivery slots are durable, hard-to-replicate assets. | 中 | SP026, SP027 |
| CP027 | Distribution power favors incumbents with existing generation fleets and grid relationships. | 中 | SP002, SP011 |
| CP028 | Multi-homing is common: hyperscalers diversify power across nuclear, gas and grid to reduce dependency. | 中 | SP004, SP003 |
| CP029 | Co-located gas power risks commoditization as many developers replicate the model. | 中 | SP003, SP017 |
| CP030 | Nuclear and renewables-plus-storage rivals could displace gas on carbon and long-run cost grounds. | 中 | SP004, SP028 |
| CP031 | Better-capitalized incumbents (Vistra, Constellation) can outbid or out-scale a single-project entrant. | 中 | SP011, SP015 |
| CP032 | Crusoe's vertical integration into AI cloud gives it a revenue base Joulent lacks as a pure power developer. | 中 | SP009, SP005 |
| CP033 | Turbine scarcity shared with rivals could delay Joulent relative to nuclear peers with existing plants. | 中 | SP025, SP013 |
| CP034 | The status-quo grid interconnection remains the default competitor, constrained by multi-year queues. | 高 | SP002, SP027 |
| CP035 | GE Vernova's record turbine order backlog signals intense competition for gas generation capacity. | 中 | SP025 |
| CP036 | Joulent's Chevron partnership for fuel and balance sheet is a differentiator few pure-play developers can match. | 中 | SP021, SP029 |
| CP037 | No single competitor combines Joulent's exact mix of oil-major fuel, utility capital and hyperscaler anchor. | 低 | SP021, SP022 |
| CI001 | Joulent's revenue model is built on long-term power purchase agreements for electricity supplied to data centers and industrial users. | 中 | SI001, SI002 |
| CI002 | For Project Kilby, Chevron—through Energy Forge One LLC—holds a 20-year power purchase agreement with Microsoft. | 高 | SI003, SI002 |
| CI003 | Joulent holds a 50% equity option in the Project Kilby joint venture, its primary route to project economics. | 中 | SI004 |
| CI004 | Power purchase agreements in this segment typically run 10-20 years with take-or-pay style commitments. | 中 | SI002, SI004 |
| CI005 | Data-center PPA pricing is negotiated bilaterally rather than published, commonly cited in the tens of dollars per megawatt-hour range. | 低 | SI005, SI006 |
| CI006 | Revenue recognition would follow energy delivered and capacity made available under the PPA once the plant operates. | 低 | SI002 |
| CI007 | Joulent is currently pre-revenue on Kilby, with first power targeted for 2028. | 中 | SI007, SI004 |
| CI008 | Joulent's go-to-market centers on securing anchor hyperscaler offtake before committing capital, as evidenced by the Microsoft agreement preceding FID. | 中 | SI007, SI003 |
| CI009 | The sales cycle for gigawatt-scale power agreements spans months to years and involves siting, interconnection and multi-party contracting. | 中 | SI008, SI002 |
| CI010 | Customer acquisition is highly concentrated—one anchor customer (Microsoft) underpins the first project. | 中 | SI007 |
| CI011 | Partner channels (Chevron, GE Vernova, National Grid) substitute for a conventional sales organization at this stage. | 中 | SI009, SI010 |
| CI012 | Project Kilby is projected to cost roughly $7 billion to $9 billion to build. | 中 | SI002, SI004 |
| CI013 | Co-located gas generation is highly capital-intensive, with build costs commonly in the millions of dollars per megawatt. | 中 | SI011, SI006 |
| CI014 | Gross margins on power projects hinge on the spread between contracted PPA revenue and fuel plus operating costs. | 中 | SI005, SI006 |
| CI015 | Natural-gas fuel cost is the largest variable operating input and a key margin driver. | 中 | SI006, SI005 |
| CI016 | Working capital and construction financing needs are front-loaded across a multi-year build before any revenue. | 中 | SI011, SI012 |
| CI017 | GE Vernova turbines and Caterpillar Solar Turbines represent major equipment cost line items for Kilby. | 中 | SI013, SI011 |
| CI018 | Long-run project returns depend on high utilization and long contract duration to amortize heavy upfront capex. | 中 | SI012, SI004 |
| CI019 | Joulent is privately held and discloses no audited revenue, ARR or profit figures. | 中 | SI014, SI015 |
| CI020 | The primary public traction signals are the signed Microsoft PPA and the $1.75 billion strategic investment, not operating metrics. | 中 | SI010, SI003 |
| CI021 | Kilby's 2.67 GW of contracted capacity is the headline forward-traction metric in lieu of revenue. | 中 | SI016, SI004 |
| CI022 | No public data exists on Joulent's headcount, burn rate or operating expenses. | 低 | SI015 |
| CI023 | On July 1, 2026, National Grid Ventures agreed to invest $1.75 billion in Joulent. | 高 | SI015, SI010 |
| CI024 | The National Grid Ventures investment buys an approximately 35% stake, implying a valuation of roughly $5 billion. | 高 | SI017, SI010 |
| CI025 | National Grid characterizes the investment as balance-sheet funded and incremental to its existing capital plan. | 中 | SI015, SI010 |
| CI026 | National Grid is running the largest capital program in its history and earmarked roughly $1 billion of near-term capital for Ventures projects. | 低 | SI018 |
| CI027 | The NGV capital is intended to fund Project Kilby and additional large-scale power projects. | 中 | SI010, SI001 |
| CI028 | A $7-9 billion Kilby build implies substantial additional project finance or partner capital beyond the $1.75 billion equity injection. | 中 | SI004, SI011 |
| CI029 | The next major financing trigger is Kilby's targeted end-2026 final investment decision. | 中 | SI004, SI003 |
| CI030 | Project-finance debt secured against the 20-year Microsoft PPA is a plausible funding path for construction. | 低 | SI002, SI012 |
| CI031 | Chevron's involvement provides both fuel supply and balance-sheet depth that reduces Joulent's standalone financing burden. | 中 | SI009, SI003 |
| CI032 | Skeptics warn that large new gas plants for data centers carry stranded-asset and return risks if AI demand or policy shifts. | 中 | SI019 |
| CI033 | The capital intensity and long build cycle mean returns are back-ended and sensitive to cost overruns and delays. | 中 | SI011, SI012 |
| CI034 | The absence of public financials is a material diligence blocker for assessing revenue quality and margins. | 中 | SI015, SI014 |
| CI035 | Revenue quality could be high once operational, given a 20-year investment-grade-counterparty PPA, but is unproven pre-FID. | 中 | SI003, SI002 |
| CI036 | The implied ~$5 billion valuation is pre-revenue and rests on contracted pipeline rather than current cash flows. | 中 | SI017, SI004 |
| CI037 | Historical funding chronology (covered in Company Overview) shows the July 2026 NGV round as Joulent's first disclosed external equity. | 中 | SI010, SI014 |
| CI038 | Business Wire's primary release confirms the $1.75 billion strategic investment positioning Joulent as a technology-driven energy company. | 高 | SI020, SI010 |
| CI039 | Independent coverage corroborates that the National Grid stake is a strategic minority position rather than a control acquisition. | 低 | SI021, SI017 |
| CI040 | National Grid Ventures is an active US infrastructure investor, lending balance-sheet credibility to Joulent's financing plan. | 中 | SI022, SI023 |
| CI041 | An off-grid, behind-the-meter design lets Kilby avoid socializing grid-upgrade costs, which supports the project's standalone economics. | 中 | SI024, SI025 |
| CI042 | Co-located generation sidesteps multi-year interconnection queues, improving time-to-revenue versus a grid-dependent build. | 中 | SI025, SI003 |
| CI043 | The 20-year Microsoft power agreement is the contractual backbone underpinning Joulent's forward revenue. | 高 | SI026, SI003 |
| CE001 | Joulent's core product is dedicated electrical power capacity supplied directly to data centers and industrial users rather than a device or software license. | 高 | SE001, SE002 |
| CE002 | The offering is structured as power-as-a-service: Joulent finances, builds, owns and operates the generation and sells the electricity under long-term agreements. | 中 | SE001, SE003 |
| CE003 | Project Kilby is designed to deliver roughly 2.67 GW of generating capacity to serve Microsoft's AI data-center compute. | 高 | SE004, SE005 |
| CE004 | The product solves a hyperscaler job-to-be-done: obtaining multi-hundred-megawatt firm power on a timeline the congested public grid cannot meet. | 中 | SE006, SE003 |
| CE005 | Kilby is engineered as an off-grid, behind-the-meter campus that islands its load from the ERCOT transmission system. | 高 | SE007, SE005, SE001 |
| CE006 | By co-locating generation with compute, Joulent removes the multi-year transmission-interconnection wait that constrains conventional data-center siting. | 中 | SE006, SE007 |
| CE007 | Kilby's generation fleet centers on GE Vernova 7HA-class heavy-duty gas turbines configured in combined-cycle blocks. | 高 | SE008, SE009 |
| CE008 | Caterpillar Solar Turbines units supplement the GE Vernova fleet, adding modular gas-turbine capacity. | 中 | SE008, SE005 |
| CE009 | The campus spans more than 2,000 acres near Pecos in Reeves County, Texas, in the Permian Basin. | 中 | SE005, SE010 |
| CE010 | Kilby sits atop Permian Basin natural-gas supply, giving the plant local, low-cost fuel access. | 中 | SE011, SE003 |
| CE011 | The asset base is organized as a joint venture in which Chevron holds the Microsoft PPA and Joulent holds a 50% equity option. | 高 | SE011, SE003 |
| CE012 | Beyond Kilby, Joulent's product line is a pipeline of additional co-located power campuses funded by the National Grid Ventures capital. | 低 | SE012, SE001 |
| CE013 | Combined-cycle configuration recovers exhaust heat via a steam turbine, materially raising fuel-to-power efficiency over simple-cycle operation. | 高 | SE009, SE013 |
| CE014 | GE Vernova's 7HA turbines reach combined-cycle efficiencies in the 63-64% range, among the highest in commercial gas generation. | 高 | SE009, SE014 |
| CE015 | The 7HA platform is engineered to co-fire hydrogen blends, giving a partial future decarbonization pathway. | 中 | SE014, SE015 |
| CE016 | Grid-orchestration software such as GE Vernova's GridOS is used to balance generation, storage and load across an islanded campus. | 中 | SE016, SE017 |
| CE017 | Operating as an islanded microgrid, Kilby must self-supply frequency regulation, spinning reserve and black-start capability normally provided by the grid. | 中 | SE007, SE016 |
| CE018 | The operating model front-loads capital in turbines, heat-recovery and controls, then earns over decades of high-utilization dispatch. | 中 | SE003, SE018 |
| CE019 | Behind-the-meter delivery shortens the electrical path from turbine to server, reducing transmission losses and exposure to grid congestion pricing. | 中 | SE007, SE006 |
| CE020 | A final investment decision on Kilby is targeted for the end of 2026. | 中 | SE003, SE005 |
| CE021 | First power from Kilby is targeted for 2028, implying a multi-year engineering-procurement-construction cycle. | 高 | SE004, SE003 |
| CE022 | Construction is projected to require more than 6,000 workers at peak, indicating a large field-deployment effort. | 中 | SE010, SE019 |
| CE023 | GE Vernova's heavy-duty gas-turbine backlog is effectively sold out into the late 2020s, making turbine slot allocation a gating deployment dependency. | 中 | SE014, SE018 |
| CE024 | Reliability for an AI training campus depends on high availability, so redundant turbine blocks and on-site backup are integral to the design. | 中 | SE007, SE005 |
| CE025 | The phased buildout lets capacity come online in blocks, allowing partial revenue before the full 2.67 GW is complete. | 低 | SE003, SE006 |
| CE026 | Support and operations are expected to lean on partner OEMs (GE Vernova, Caterpillar) for long-term service agreements on the turbine fleet. | 低 | SE018, SE015 |
| CE027 | Joulent's primary differentiation is speed-to-power: delivering firm gigawatt-scale capacity years faster than a grid-interconnected alternative. | 中 | SE006, SE007 |
| CE028 | Privileged partner access—GE Vernova turbines, Chevron fuel and Microsoft offtake—forms a defensible ecosystem moat around each project. | 中 | SE011, SE008 |
| CE029 | Site control over 2,000+ acres with co-located gas supply is a scarce, hard-to-replicate physical asset. | 中 | SE005, SE011 |
| CE030 | Engine No. 1 incubation and Chris James's energy-transition network give Joulent capital-markets and dealmaking reach uncommon for a young developer. | 中 | SE002, SE012 |
| CE031 | The behind-the-meter model is being pursued by multiple competitors, so first-mover execution—not the concept—is the durable edge. | 中 | SE006, SE020 |
| CE032 | As large combustion sources, Kilby's turbines require federal and state air-quality permitting for emissions such as NOx and CO2. | 中 | SE021, SE013 |
| CE033 | Islanding off ERCOT reduces some interconnection-approval exposure but does not exempt the plant from environmental and safety regulation. | 中 | SE007, SE021 |
| CE034 | Critics argue the gas-fired design carries meaningful carbon and stranded-asset risk if AI demand or climate policy shifts. | 中 | SE022, SE023 |
| CE035 | Environmental groups have questioned the emissions footprint of large behind-the-meter gas plants serving data centers. | 低 | SE024, SE023 |
| CE036 | Reliability and safety quality controls for the turbine fleet are governed by OEM engineering standards and long-term service agreements. | 中 | SE018, SE009 |
| CE037 | Hydrogen co-firing capability offers a compliance hedge but current blends only modestly reduce lifecycle emissions. | 低 | SE014, SE015 |
| CE038 | No public third-party audit of Kilby's design, emissions modeling or reliability engineering is yet available. | 低 | SE001, SE007 |
| CE039 | Water sourcing for combined-cycle steam and cooling in arid West Texas is an unresolved operational and compliance question. | 低 | SE021, SE019 |
| CE040 | The overall technology stack is proven at the component level (turbines, controls) but unproven at Joulent's specific 2.67 GW islanded integration. | 中 | SE005, SE006 |
| CE041 | Independent turbine-engineering commentary corroborates that modern H-class combined-cycle plants are a mature, well-understood technology base. | 低 | SE025, SE013 |
| CU001 | Joulent's customer base targets two segments: hyperscale data-center operators and large industrial power users. | 高 | SU001, SU002 |
| CU002 | The anchor and only publicly named customer to date is Microsoft, contracting power for its AI data centers. | 高 | SU003, SU002 |
| CU003 | Microsoft is the buyer, user and payer for Kilby's contracted power, making it the single decisive account. | 中 | SU002, SU004 |
| CU004 | Geographically, the initial customer footprint is concentrated in Texas at the Kilby campus in Reeves County. | 中 | SU005, SU006 |
| CU005 | The use case is powering AI training and inference compute that cannot obtain timely grid capacity. | 中 | SU007, SU006 |
| CU006 | Industrial users are named as a target segment but no specific industrial customer has been disclosed. | 低 | SU001, SU008 |
| CU007 | Adoption is contractual rather than operational: Kilby is contracted but not yet delivering power. | 中 | SU009, SU003 |
| CU008 | Microsoft has effectively reserved roughly 2 GW of Kilby's capacity for its compute needs. | 中 | SU010, SU006 |
| CU009 | First power to the customer is targeted for 2028, so active usage metrics do not yet exist. | 高 | SU003, SU009 |
| CU010 | The headline adoption metric is contracted capacity (2.67 GW total), not deployed or utilized capacity. | 中 | SU010, SU005 |
| CU011 | No repeat purchases, additional locations or utilization data are available at this pre-operational stage. | 低 | SU009, SU008 |
| CU012 | Microsoft's broader AI buildout (e.g. its Fairwater-class campuses) underpins the demand behind the Kilby reservation. | 中 | SU006, SU011 |
| CU013 | The Microsoft relationship is evidenced by a 20-year power purchase agreement held through Chevron's Energy Forge One entity. | 高 | SU004, SU002 |
| CU014 | This is a production-grade commercial commitment, not a pilot or letter of intent. | 中 | SU004, SU009 |
| CU015 | Microsoft has publicly discussed securing dedicated power for AI infrastructure, corroborating the strategic intent. | 中 | SU002, SU006 |
| CU016 | The proof is fresh, tied to Joulent's June-July 2026 launch and investment announcements. | 中 | SU012, SU011 |
| CU017 | Reference quality is high on contract certainty but low on operational outcomes, which cannot exist pre-2028. | 中 | SU009, SU003 |
| CU018 | Beyond Microsoft, no second named customer proof point has been made public. | 低 | SU008, SU001 |
| CU019 | The 20-year PPA gives exceptional contractual durability and revenue visibility once power flows. | 高 | SU004, SU009 |
| CU020 | Take-or-pay-style structures common to such PPAs further lock in customer commitment. | 低 | SU009, SU013 |
| CU021 | No net revenue retention, gross retention or churn metrics exist because the customer is not yet live. | 中 | SU003, SU009 |
| CU022 | Switching costs for the customer are very high given the co-located, purpose-built nature of the plant. | 中 | SU014, SU007 |
| CU023 | Contract length (20 years) is the primary durability signal in lieu of cohort or renewal data. | 中 | SU004, SU010 |
| CU024 | Counterparty credit quality is strong: Microsoft is an investment-grade offtaker, reducing default risk. | 中 | SU002, SU004 |
| CU025 | Customer concentration is extreme: essentially one customer underpins the entire flagship project. | 高 | SU003, SU015 |
| CU026 | This single-customer dependence is the dominant commercial risk in the customer base. | 中 | SU015, SU016 |
| CU027 | The land-and-expand path is to replicate the co-located model with additional hyperscaler campuses funded by NGV capital. | 中 | SU012, SU001 |
| CU028 | Expansion depends heavily on partners—Chevron for fuel and PPAs, GE Vernova for turbines—creating channel dependence. | 中 | SU004, SU006 |
| CU029 | Procurement friction is high: gigawatt power deals require long negotiation, siting and multi-party contracting. | 中 | SU007, SU006 |
| CU030 | A downturn in Microsoft's AI capex or a strategy shift would directly threaten Kilby's economics. | 中 | SU015, SU016 |
| CU031 | Diversifying beyond Microsoft is the key commercial milestone to de-risk the customer base. | 低 | SU008, SU001 |
| CU032 | Skeptics question whether hyperscalers' reliance on new gas plants is durable given climate commitments. | 中 | SU015, SU016 |
| CU033 | Community and environmental pushback around large gas-for-data-center projects could pressure customer optics. | 低 | SU016, SU008 |
| CU034 | Analysts note Microsoft has multiple power-sourcing options (nuclear restarts, renewables, other developers), limiting Joulent's leverage. | 中 | SU015, SU017 |
| CU035 | The absence of any disclosed second customer keeps concentration risk unmitigated as of mid-2026. | 低 | SU018, SU001 |
| CU036 | Even a strong 20-year contract does not eliminate construction- and delivery-phase risk to the customer relationship. | 低 | SU009, SU019 |
| CU037 | Public detail on pricing, volume ramp and service-level terms of the Microsoft agreement is not disclosed. | 中 | SU004, SU009 |
| CU038 | Mainstream and trade coverage identifies Microsoft as the AI-compute customer behind the multi-billion-dollar Kilby campus. | 中 | SU020, SU021 |
| CU039 | Long-dated data-center PPAs are structured to give developers a durable, contracted revenue book across the plant's life. | 中 | SU022, SU023 |
| CU040 | Partner-supplied orchestration and turbine platforms (e.g. GE Vernova) are integral to serving the customer's reliability requirements. | 低 | SU024, SU025 |
| CR001 | The single largest risk is execution: delivering a $7-9B gigawatt-scale plant on a compressed 2026-2028 schedule. | 中 | SR001, SR002 |
| CR002 | First power is not expected until 2028, so the business is exposed to years of pre-revenue construction risk. | 高 | SR003, SR001 |
| CR003 | Capital intensity magnifies downside: cost overruns on a multi-billion-dollar build directly erode equity returns. | 中 | SR004, SR001 |
| CR004 | Customer concentration is a top-tier risk, with essentially one customer (Microsoft) underpinning the flagship. | 高 | SR003, SR005 |
| CR005 | Residual exposure remains high across most risk categories because the project is pre-construction and largely undisclosed. | 中 | SR001, SR006 |
| CR006 | Mitigation maturity is uneven: strong contractual and partner mitigants, weak public transparency and unproven operations. | 中 | SR007, SR006 |
| CR007 | Texas Senate Bill 6 (2025) tightens rules for large electric loads interconnecting to the grid. | 高 | SR008, SR009 |
| CR008 | ERCOT's large-load interconnection process (Project 58481) imposes study fees, disclosure and security requirements on loads at and above roughly 75 MW. | 高 | SR010, SR011 |
| CR009 | SB6 gives ERCOT and utilities expanded authority to curtail or disconnect large loads under grid-stress conditions. | 中 | SR011, SR012 |
| CR010 | Legal analysts flag heightened disclosure and demand-transparency obligations for large data-center loads in Texas. | 中 | SR013, SR009 |
| CR011 | An off-grid, behind-the-meter design reduces—but does not fully remove—exposure to ERCOT interconnection rules. | 中 | SR014, SR008 |
| CR012 | Air-quality permitting for large gas turbines (NOx, CO2) is a required and time-consuming regulatory step. | 中 | SR015, SR012 |
| CR013 | Environmental litigation or permit challenges are a plausible legal risk for large fossil-fuel projects. | 低 | SR016, SR008 |
| CR014 | Water-use permitting in arid West Texas adds a further regulatory dependency for combined-cycle cooling. | 低 | SR012, SR013 |
| CR015 | Federal energy oversight (e.g. FERC-adjacent rules) can bear on co-location and wholesale-power arrangements. | 低 | SR009, SR011 |
| CR042 | Industry and legal commentary emphasize that Texas's large-load regulatory regime is still evolving, creating forward compliance uncertainty for data-center power projects. | 中 | SR017, SR018 |
| CR016 | Turbine supply is a critical operational risk: GE Vernova's heavy-duty backlog is effectively sold out into the late 2020s. | 中 | SR019, SR020 |
| CR017 | Gas-turbine lead times and global demand create delivery-slot risk that could slip Kilby's timeline. | 中 | SR015, SR019 |
| CR018 | Construction-phase risks include labor availability (6,000+ peak workers), EPC coordination and weather. | 中 | SR021, SR015 |
| CR019 | Operating an islanded 2.67 GW microgrid introduces reliability risk in self-supplying grid services. | 中 | SR022, SR023 |
| CR020 | Natural-gas supply disruptions or price spikes are an operational and margin risk despite Permian proximity. | 低 | SR023, SR004 |
| CR021 | Outage or equipment-failure events at a single-site plant would directly threaten the anchor customer's power. | 低 | SR023, SR022 |
| CR022 | No public reliability, safety or quality-audit record exists for Kilby's specific design yet. | 低 | SR024, SR022 |
| CR023 | Chevron dependence is high: Chevron holds the Microsoft PPA and supplies fuel, so its commitment is load-bearing. | 高 | SR007, SR001 |
| CR024 | GE Vernova dependence is high for both turbines and grid-orchestration technology. | 中 | SR025, SR020 |
| CR025 | National Grid Ventures is the key capital provider; a change in its commitment would strain financing. | 中 | SR026, SR006 |
| CR026 | Microsoft is simultaneously the key customer and the demand rationale, compounding concentration risk. | 中 | SR003, SR005 |
| CR027 | Regulators (ERCOT, Texas PUC, environmental agencies) are external dependencies that can gate the project. | 中 | SR010, SR008 |
| CR028 | The joint-venture structure means Joulent's economics depend on partner performance it does not fully control. | 中 | SR001, SR007 |
| CR029 | Failure of any single critical partner (fuel, turbines, capital, offtake) could stall the flagship. | 低 | SR007, SR025 |
| CR030 | The equity raised ($1.75B) is small versus the $7-9B build, implying substantial additional financing risk. | 中 | SR006, SR004 |
| CR031 | Long pre-revenue period means sustained cash outflow before any offsetting income. | 中 | SR001, SR003 |
| CR032 | Margin compression risk arises if gas prices rise or PPA pricing is fixed without adequate pass-through. | 低 | SR004, SR023 |
| CR033 | Leverage against a 20-year PPA is plausible but exposes the model to interest-rate and credit conditions. | 低 | SR001, SR006 |
| CR034 | Near-total absence of public financials is itself a diligence risk, limiting independent verification. | 中 | SR006, SR027 |
| CR035 | A downturn in AI data-center capex would undermine both demand and the valuation basis. | 中 | SR005, SR028 |
| CR036 | Critics warn that large new gas plants for data centers carry stranded-asset risk if AI demand or climate policy shifts. | 中 | SR029, SR030 |
| CR037 | Environmental groups highlight the carbon footprint and climate tension of gas-fired data-center power. | 低 | SR016, SR030 |
| CR038 | Hyperscaler climate commitments could pressure the customer to limit or exit gas-based power over time. | 低 | SR030, SR028 |
| CR039 | Regulatory tightening on emissions or carbon pricing would raise operating costs and stranded-asset odds. | 低 | SR008, SR016 |
| CR040 | Key mitigants—20-year PPA, investment-grade offtaker, strong partners, hydrogen-capable turbines—reduce but do not eliminate the top risks. | 中 | SR007, SR020 |
| CR041 | The clearest thesis-break triggers are FID slippage past end-2026, loss of the Microsoft contract, or a major cost blowout. | 中 | SR001, SR003 |
| CV001 | The investment thesis is that Joulent is an early leader in the scarce, fast-growing market for co-located power for AI data centers. | 中 | SV001, SV002 |
| CV002 | A signed 20-year Microsoft PPA and a $1.75B strategic investment validate demand and de-risk the anchor project. | 高 | SV003, SV004 |
| CV003 | Privileged partner access (Chevron, GE Vernova, National Grid) forms a defensible ecosystem advantage. | 中 | SV004, SV005 |
| CV004 | The anti-thesis is that Joulent is a pre-revenue, single-project, single-customer developer valued at ~$5B on execution optionality. | 中 | SV006, SV007 |
| CV005 | Capital intensity, carbon exposure and construction risk could impair returns if execution slips. | 中 | SV007, SV008 |
| CV006 | The thesis hinges on Kilby reaching FID and first power on schedule; the anti-thesis hinges on the opposite. | 中 | SV009, SV010 |
| CV007 | On balance the recommendation is to monitor: a high-quality option that is not yet an underwritable business. | 中 | SV009, SV006 |
| CV008 | Confidence is medium, constrained by the near-total absence of public financials. | 中 | SV006, SV011 |
| CV009 | The risk rating is high, driven by execution, concentration and capital-intensity risk. | 中 | SV007, SV008 |
| CV010 | The valuation stance is premium: the ~$5B mark prices in successful execution not yet demonstrated. | 中 | SV012, SV006 |
| CV011 | A supportable posture is to track FID and first-power milestones before committing capital. | 中 | SV009, SV010 |
| CV012 | The $1.75B NGV investment for ~35% implies an approximate $5B post-money valuation. | 高 | SV003, SV006 |
| CV013 | This is a strategic minority investment, not a control transaction, shaping preference and governance dynamics. | 中 | SV006, SV003 |
| CV014 | Entry discipline is challenged: the valuation rests on contracted pipeline rather than current earnings. | 中 | SV006, SV012 |
| CV015 | Dilution and preference overhang from a large strategic stake could weigh on future common returns. | 低 | SV006, SV013 |
| CV016 | Public evidence supports the existence and size of the deal but not an independent valuation cross-check. | 中 | SV003, SV011 |
| CV017 | The ~$5B mark is a private, primary-round figure, not a market-tested or liquidity-backed valuation. | 中 | SV006, SV014 |
| CV018 | Bull case: Kilby executes on time, Joulent replicates the model, and it becomes a multi-gigawatt power platform worth well above $5B. | 低 | SV001, SV002 |
| CV019 | Base case: Kilby is built with some delay/cost overrun; Joulent is a valuable but single-asset developer near its current mark. | 低 | SV009, SV012 |
| CV020 | Bear case: FID slips or Microsoft demand softens, cost overruns bite, and the equity is impaired. | 低 | SV007, SV008 |
| CV021 | Downside triggers include FID slippage, PPA renegotiation, capex blowout above $9B, and AI-capex retrenchment. | 中 | SV009, SV007 |
| CV022 | Probability signals favor the base case given contracted demand but material tail risk on execution. | 低 | SV001, SV009 |
| CV023 | The valuation range spans a low near impaired equity to a high several times the current mark under successful scaling. | 低 | SV012, SV008 |
| CV024 | Crusoe Energy, a comparable AI-data-center power/compute developer, was valued around $10B in its late-2025 round. | 中 | SV014, SV015 |
| CV025 | Data-center infrastructure assets have traded at roughly 15-19x EBITDA in recent transactions. | 中 | SV012, SV016 |
| CV026 | Listed IPPs exposed to data-center demand (e.g. Vistra, Constellation, Talen) provide public-market reference points. | 中 | SV016, SV002 |
| CV027 | Private strategic rounds in AI-power infrastructure have clustered at multibillion-dollar valuations through 2025-2026. | 中 | SV013, SV014 |
| CV028 | Comparables suggest Joulent's ~$5B mark is within range for the theme but rich for a pre-revenue single-asset developer. | 中 | SV012, SV013 |
| CV029 | Every comparable has limits: Crusoe has compute revenue, IPPs are operating utilities, and infra multiples assume cash-flowing assets. | 中 | SV015, SV016 |
| CV030 | A DCF on Kilby would hinge on PPA pricing, gas cost, utilization and discount rate—all largely undisclosed. | 低 | SV008, SV009 |
| CV031 | Exit paths include an IPO (as peers like Crusoe are reportedly pursuing), strategic sale, or partner buyout. | 低 | SV014, SV015 |
| CV032 | Exit readiness is early: pre-revenue status and single-asset concentration limit near-term liquidity options. | 中 | SV006, SV009 |
| CV033 | National Grid's strategic stake could presage a larger takeout or provide a natural future buyer. | 低 | SV003, SV006 |
| CV034 | Final diligence asks center on financials, the full cap table, PPA economics, and the project-finance plan. | 中 | SV006, SV011 |
| CV035 | Thesis-break triggers are FID slippage past end-2026, loss of the Microsoft PPA, or a major cost blowout. | 中 | SV009, SV010 |
| CV036 | Skeptics caution that theme-driven valuations in AI-power infrastructure risk overshooting fundamentals. | 中 | SV007, SV008 |
| CV037 | The valuation is best treated as an option premium on execution rather than a multiple on current cash flow. | 中 | SV012, SV006 |
| CV038 | Independent verification of the valuation is limited by the private, undisclosed nature of Joulent's financials. | 中 | SV011, SV006 |
| CV039 | A monitor stance lets an investor re-rate on de-risking milestones without paying the full execution premium today. | 中 | SV009, SV012 |
| CV040 | The bull-bear valuation spread is unusually wide, reflecting binary execution outcomes on a single megaproject. | 低 | SV008, SV012 |
| CV041 | On IC-style scoring, market scale and proof are strong while economics, evidence quality and valuation discipline are weak. | 中 | SV002, SV006 |
| CV042 | Overall, Joulent is a compelling thematic option whose current price demands milestone-based, not immediate, commitment. | 中 | SV009, SV012 |
| CV043 | Independent market analyses project rapid growth in data-center and AI power demand, underpinning the market-scale leg of the thesis. | 中 | SV017, SV018 |
| CV044 | Third-party forecasts of surging AI compute and power capex support the durability of demand behind Joulent's model. | 中 | SV019, SV020 |
| CV045 | Joulent's Engine No. 1 lineage and named leadership lend execution credibility to the buildout ambition. | 中 | SV021, SV022 |
| CV046 | Multiple independent outlets reported the $1.75B strategic investment, corroborating the financing event behind the valuation. | 中 | SV023, SV024 |
| CV047 | National Grid's own investor communications frame the stake as a strategic infrastructure position, informing governance context. | 中 | SV025, SV026 |
| CV048 | Reported Kilby economics (multi-GW scale, ~$7-9B capex, 20-year PPA) are the operational basis for any valuation model. | 中 | SV027, SV028 |
| CV049 | Coverage of the Microsoft offtake and Kilby siting corroborates the contracted-demand assumptions in the base case. | 中 | SV029, SV030 |
| CV050 | Crusoe's own disclosures on its financing and scale sharpen the closest private comparable to Joulent. | 中 | SV031, SV032 |
| CV051 | Competitor moves—including nuclear restarts and rival gas developers—frame the public-market comparable landscape. | 中 | SV033, SV034 |
| CV052 | Partner and turbine-supply reporting confirms the dependency and scarcity factors weighing on the risk rating. | 中 | SV035, SV036 |
| CV053 | Supply-chain and regulatory analyses reinforce the execution and policy risks that justify a cautious stance. | 中 | SV037, SV038 |
| CV054 | Skeptical technology-press coverage underscores the risk that AI-power theme valuations overshoot fundamentals. | 中 | SV039, SV017 |