# a16z 解析 Texas 为何让数据中心排队等待并网
> 原标题:Why Texas Is Making Data Centers Wait
- 来源:a16z:News
- 发布时间:2026-10-07T14:01:44.000Z
- AIHOT:https://aihot.news/items/yk3grmalszldrwd1p8bxk0cki
- 原文:https://www.a16z.news/p/why-texas-is-making-data-centers
## 摘要
a16z 的 Ryan McEntush 撰文分析 Texas 电网对数据中心按下暂停键的能源原因。
## 推荐理由
原文从并网队列、自备电源到灵活性机制逐层拆解 Texas 暂停审批的能源原因,为理解 AI 数据中心扩产瓶颈提供了可迁移的分析框架。
## 正文 · 中文译文

2024年底,得克萨斯州电网运营商的排队名单上有[63 GW的大型新客户](https://www.latitudemedia.com/news/ercots-large-load-queue-has-nearly-quadrupled-in-a-single-year/)。到今年6月,这一[数字达到了474 GW](https://www.utilitydive.com/news/texas-hits-pause-data-center-interconnections/827046/),是创纪录峰值需求的五倍多,其中约90%是数据中心。
随后得克萨斯州[按下了暂停键](https://www.texastribune.org/2026/08/03/texas-data-center-project-audit-greg-abbott/)。最初只是要求数据中心[自行承担](https://gov.texas.gov/news/post/governor-abbott-directs-puc-and-ercot-to-shield-texans-from-data-center-infrastructure-costs)电网升级费用,后来升级为[冻结](https://gov.texas.gov/news/post/governor-abbott-directs-tceq-to-halt-data-center-permits)所有新许可。如果你关心AI和美国的再工业化,这一点值得了解。它为什么这么做?
即将到来的[选举](https://www.texastribune.org/2026/09/21/texas-data-center-moratorium-water-energy/)是部分原因,但背后的深层问题值得深挖。首先是并网排队问题。这次暂停既关乎项目本身,也关乎项目如何申请审批。开发商惯常会在多个站点重复提交项目,其中许多是尚无客户的[投机性](https://www.bloomberg.com/news/articles/2026-08-12/most-electricity-sought-for-ai-data-centers-in-us-will-never-materialize)建设。这些新开发商往往连一块GPU都没插过电,更不用说发电厂了,因此处理低质量申请让负责该州大部分电网的ERCOT陷入[困境](https://www.texastribune.org/2026/06/17/texas-ercot-data-center-energy-grid/)。规划者无法分辨哪些是真实的,他们也不希望电费用户为那些永远不会落地的项目所建的升级买单。
第二个问题是社区认同。噪音、用水、排放和电费是任何附近居民[最关心的事](https://www.pewresearch.org/short-reads/2026/09/22/americans-views-of-data-centers-have-turned-more-negative/)。无论这些问题在[现实中](https://www.microsoft.com/en-us/microsoft-cloud/blog/2024/12/09/sustainable-by-design-next-generation-datacenters-consume-zero-water-for-cooling/)如何,当数据中心落户当地时,人们理应得到[直截了当的答复](https://www.texastribune.org/2026/09/14/texas-data-centers-abbott-water-survey-enforcement/),而他们[并不总能得到](https://www.ksat.com/news/texas/2026/08/10/while-data-centers-boom-in-texas-their-plans-stay-shrouded-in-secrecy/)([截至6月,377家公司中只有28家](https://www.texastribune.org/2026/08/03/texas-data-center-project-audit-greg-abbott/)回答了该州关于其资源使用的调查)。在胡德县,委员们被要求支持“爱国者项目”的税收减免,却不知道[它究竟是谁](https://www.texastribune.org/2026/06/02/texas-data-centers-hood-county-local-control-rural-water-power/)。平心而论,公司在选址时使用代号很常见,因为知名买家会推高价格,但一旦官员就税收减免进行投票,这一逻辑就更难站得住脚。
这一更广泛的政治背景很重要,但我将只聚焦[能源](https://a16z.com/decentralizing-the-electric-grid/)方面:为什么为数据中心供电如此困难,开发商仍在尝试的各种办法,以及事态可能走向何方。
### 并网究竟意味着什么
亚马逊购电的方式有助于说明发生了什么变化。过去十年,他们会找到一家有网络容量的公用事业公司,注册为大负荷用户,然后签订长期合同,在纸面上[匹配](https://sustainability.aboutamazon.com/renewable-energy-methodology.pdf)其用电量。这些合同往往[纯粹是财务性的](https://www.epa.gov/green-power-markets/financial-ppa),意味着你不必总是证明电力能输送到你那里,而且并网很容易,因为公用事业公司通常有闲置的网络容量。
这已不再成立。电网已被拉伸到极限,如今几乎任何新的大负荷或发电设施都需要升级。[电力流经每一条相连的路径](https://www.ferc.gov/sites/default/files/2020-05/ch1-3_0.pdf),所以从某座电厂购电并不会为你预留一条通往你建筑的线路(甚至还有输电拥堵[权](https://www.ercot.com/mktinfo/crr)市场)。
因此,买家首先转向他们能更直接主张的稳定电力(全天候可用),通常是通过[重启退役电厂](https://www.constellationenergy.com/news/2024/Constellation-to-Launch-Crane-Clean-Energy-Center-Restoring-Jobs-and-Carbon-Free-Power-to-The-Grid.html)或在现有电厂旁边建设。[两者都不是真正的出路](https://www.utilitydive.com/news/constellation-three-mile-island-crane-nuclear-ferc-waiver/821836/),而一些最激烈的争论围绕“共址”展开,[ERCOT将其描述为](https://www.ercot.com/files/docs/2026/06/01/ERCOT-Trending-Topic-Grid-Reliability-Data-Centers-Net-Metering-Arrangements.pdf)从邻近电厂在电力进入电网之前取电:
- **Susquehanna:**Amazon [买下了一座园区](https://www.ans.org/news/article-5842/amazon-buys-nuclearpowered-data-center-from-talen/),紧邻 Talen 的核电站,但联邦电网监管机构 FERC [否决了 Talen 直接向其输送更多电力的申请](https://www.utilitydive.com/news/ferc-interconnection-isa-talen-amazon-data-center-susquehanna-exelon/731841/)。公用事业公司曾辩称,这会使得该园区逃避电网费用,并将固定的输电成本转嫁给其他所有人。根据[重组后的协议](https://www.sec.gov/Archives/edgar/data/1622536/000162253626000066/tln-20260630.htm),Amazon [像其他任何客户一样支付输电费用](https://www.powermag.com/talen-amazon-launch-18b-nuclear-ppa-a-grid-connected-ipp-model-for-the-data-center-era/)。
- **Freestone:**CyrusOne 位于 Constellation 燃气电厂旁的 760 MW 园区需要获得[得克萨斯州公用事业委员会(PUCT)的批准](https://tcss.legis.texas.gov/resources/UT/htm/UT.39.htm#39.169),因为它实际上会从其他人那里夺走该电厂的大部分发电量。PUCT 在 5 月[表示同意](https://www.constellationenergy.com/news/2026/05/constellation-reports-first-quarter-2026-results.html),前提是该园区能在 ERCOT 发出呼叫后 30 分钟内[削减用电](https://www.ercot.com/files/docs/2026/06/01/ERCOT-Trending-Topic-Grid-Reliability-Data-Centers-Net-Metering-Arrangements.pdf)或切换到备用电源。
- **Armstrong County:**两个 Crusoe 数据中心,分别为 265 MW 和 260 MW,共用一个最多只能发电约 265 MW 的风电场。两者都适用与 Freestone 相同的规则,因此当 ERCOT 呼叫时,每个都必须削减其全部负荷。两者合计削减的负荷几乎是该风电场实际发电量的两倍,Crusoe 称这[过于苛刻](https://www.utilitydive.com/news/texas-approves-ai-data-center-co-location-next-to-wind-farm-with-curtailme/826617/),但 PUCT 维持了这一规定。
互联研究是电网规划者确定电网新增项目需要升级哪些线路和变电站的方式。ERCOT 过去按照[一次最多处理 50 个大型负荷](https://www.utilitydive.com/news/ercots-large-load-queue-jumped-almost-300-last-year-official/808820/)的规则逐一审查研究,但到 2025 年 11 月中旬,新增了 225 个请求。因此今年 6 月,PUCT [批准了一种新的批量处理方式](https://www.ercot.com/news/release/06182026-puct-approves-ercots)。在“Batch Zero”下,ERCOT 将 75 MW 及以上的大型负荷放在一起研究,并在它们之间分配电网容量。
这些研究主要询问的是当某些东西发生故障时会发生什么。[NERC 的标准](https://www.nerc.com/pa/Stand/Reliability%20Standards/TPL-001-5.1.pdf)是北美可靠性的基准,涵盖了一系列广泛的停电场景,[ERCOT 还增加了自己的要求](https://www.ercot.com/files/docs/2026/05/29/June-1-2026-Plannng-Guide.pdf)。规划者可能会模拟一台变压器故障,然后在此基础上再让一条线路或一台发电机停运。要在不削减更多负荷的情况下维持这一可靠性标准,通常意味着要建设额外的基础设施。
一个较新的风险是负荷突然同时从电网脱落。许多数据中心在第一次电压骤降时就切换到备用电源以保护其硬件,因此一次故障就可能在几秒内[从电网拉走巨大的负荷](https://www.jdsupra.com/legalnews/nerc-computational-load-updates-6702832/)。今年早些时候在弗吉尼亚州就[发生过](https://insidelines.pjm.com/pjm-dominion-review-large-load-transfer-event/)这种情况,但幸运的是处理得很好。
得克萨斯的容错空间更小,因为 ERCOT 的电网在很大程度上是[孤立的](https://www.ercot.com/files/docs/2025/05/12/ERCOT-Grid-Insights-Electricity-Connection-and-Transfer.pdf)。在某些情况下,ERCOT 一次只能损失[约 3.2 GW](https://www.ercot.com/files/docs/2026/09/16/QSA-Results-and-Large-Load-IROL-Mitigation-Plans_final.pdf) 的负荷,否则就会引发严重问题。一项新的电压[穿越规则](https://www.ercot.com/mktrules/issues/NOGRR282)现在要求新建数据中心在常规故障期间保持连接,我甚至听说有实验室在训练运行失败后运行虚拟任务,只是为了不让负荷突然下降。
所有这些听起来可能过于保守,但该系统的设计初衷就是将可靠性置于成本之上。简而言之,电网是为炎热的夏日和严寒的冬季而建的,那时故障可能意味着生死。正是这一职责使得电网规模的确定如此困难;你为几个高峰时段而建设,却要全年为其付费。
那么,这一切韧性究竟由谁来买单?得克萨斯州通常根据每个大客户在电网夏季高峰期的用电需求来分摊输电成本,因此,一个在最热的下午削减负荷的大型用电户,可以省掉其应承担的很大一部分输电费用。7月,监管机构[提议](https://interchange.puc.texas.gov/Documents/58000_8_1664102.PDF)改为按全部12个月的峰值来计算,并[收费](https://www.sos.texas.gov/texreg/archive/July242026/Proposed%20Rules/16.ECONOMIC%20REGULATION.html)时视大型负荷为满负荷运行。换句话说,大型购电方将按其电力管道需要多大来付费,而不是按实际流经的电量来付费。
听起来很简单,但分摊成本往往是并网过程中最慢的环节。为一个园区建设的升级工程,也可能提高现有客户的可靠性,或为未来难以预先估值的增长腾出空间。而且,如果任何新开发项目超出预算,或其负荷始终没有出现,其他所有人就被迫承担开发商承诺未覆盖的部分。
而这些承诺的代价低得惊人。PUCT的[新大型负荷规则](https://interchange.puc.texas.gov/Documents/58481_218_1684654.PDF)于10月8日生效,收取10万美元的固定研究费,外加每兆瓦5万美元的押金。押金能筛掉一些投机项目,但这道过滤网很薄。如果项目落后两年,ERCOT可以重新分配其容量,但即使项目失去容量,也只损失20%的押金(一个1 GW园区约1000万美元),外加公用事业公司已经花掉的部分。最重要的是,押金通常只担保为该项目建设的那部分升级,而不包括更昂贵的区域线路——这些线路的成本由电网上的所有人共同分担。
让一个大型数据中心完全并网可能需要[5到10年](https://arxiv.org/abs/2509.07218)。因此,为加快部署,[分阶段并网方式](https://www.crusoe.ai/resources/newsroom/crusoe-expands-ai-data-center-campus-in-abilene-to-1-2-gigawatts)正变得[越来越普遍](https://www.sdxcentral.com/news/xai-colossus-memphis-power-tva/),这给公用事业公司提供了更渐进的目标来规划,也让开发商有办法证明自己能够承受所申请的全部负荷。这一切还都假设Batch Zero在推进,而目前并没有。ERCOT甚至[暂停了](https://www.ercot.com/files/docs/2026/08/20/Batch-Zero-Verification-and-Audit-PUCT-Presentation.pdf)75 MW及以上数据中心的通电审批,其中包括17个已完成ERCOT其他所有步骤的项目。在审计的12月报告确定哪些项目符合资格之前,ERCOT无法将它们放在一起研究,因此时间表是[“待定”](https://www.ercot.com/files/docs/2026/09/11/14-Batch-Zero-Update.pdf),[负荷预测也处于搁置状态](https://www.ercot.com/files/docs/2026/08/20/Batch-Zero-Verification-and-Audit-PUCT-Presentation.pdf)。
### 自带电源解决了什么
对于面临并网排队困境的开发商来说,完全绕开电网看起来很有吸引力。闲置的GPU[成本](https://epoch.ai/data-insights/ai-datacenter-cost-breakdown)远高于运行它们的电费,所以速度最重要——人们称之为“通电时间”。这就是为什么开发商计划在场地内自带电源(“表后”),但几乎总是同时接入电网,或作为接入电网的过渡:
- **阿比林:**Oracle园区[依靠电网供电,燃气作为备用](https://www.crusoe.ai/resources/blog/an-inside-look-at-the-abilene-ai-data-center)。隔壁,Crusoe宣布了一个900 MW的Microsoft园区,配有[自建电厂](https://www.crusoe.ai/resources/newsroom/crusoe-announces-new-900-mw-ai-factory-campus-in-abilene-texas-to-support-microsoft-ai-infrastructure),但Crusoe在该场地的合作伙伴Lancium的CEO[将燃气发电描述为备用](https://westtexastribune.org/2026/05/05/lancium-ceo-addresses-abilenes-concerns/)。
- **沙克尔福德县:**Vantage正在建设一个为Oracle和OpenAI提供1.4 GW算力的园区,设计为离网运行,依靠[现场燃气](https://epoch.ai/publications/openai-stargate-where-the-us-sites-stand)。但其[场地规划](https://records.tceq.texas.gov/cs/idcplg?IdcService=TCEQ_EXTERNAL_SEARCH_GET_FILE&dID=8984027&Rendition=Web)中包括一个紧邻高压线的开关站——真奇怪!
- **佩科斯县:**Pacifico 将其在 GW Ranch 规划的微电网(一种表后配置)描述为[从不从 ERCOT 取电](https://www.pacificoenergy.com/gw-ranch),但该站点的新所有者亚马逊表示,它[设计为在能够接入电网后并入电网](https://www.aboutamazon.com/news/company-news/amazon-pecos-county-texas-data-center-clean-energy-jobs)。
- **西得克萨斯:**根据一项[20 年协议](https://www.chevron.com/newsroom/2026/q2/chevron-signs-20-year-power-agreement-with-microsoft-for-west-texas-data-center),雪佛龙计划在一座规划中的微软园区旁建设一座专用燃气电厂,该园区初期不会接入 ERCOT 电网,不过雪佛龙已申请接入。雪佛龙现在表示,许可冻结[可能将其最终投资决定推迟到 2027 年](https://www.semafor.com/article/10/01/2026/texas-politics-could-delay-investment-decision-for-huge-chevron-data-center),但仍预计在 2028 年首次供电。
- **阿姆斯特朗县:**Crusoe 正在一座风电场旁建设的谷歌园区[从第一天起就同时与风电场和电网相连](https://www.crusoe.ai/resources/newsroom/crusoe-developing-data-center-campus-in-armstrong-county-texas),Crusoe 称其为“[跨表](https://www.crusoe.ai/resources/blog/co-located-power-for-ai-data-centers)”,我承认这说法挺上口。
SemiAnalysis 已在追踪表后资产的[75 GW 设备订单](https://newsletter.semianalysis.com/p/everyone-says-datacenter-moratoriums),但我所知的每个站点都打算尽快接入电网,主要是因为现场供电几乎总是成本更高。所以,一旦电网可用,比如在第五年,你就切换过去。
与此同时,将现场资产与哪怕部分电网连接配对是明智的,原因与我们最初建设电网的原因相同。当一座电厂跳闸时,ERCOT 其余[1460 多台发电机组](https://www.ercot.com/files/docs/2022/02/08/ERCOT_Fact_Sheet.pdf)会顶上,但孤岛负荷(与电网切断)没有这种余裕。电网还提供我们习以为常的东西,比如[惯性](https://www.osti.gov/biblio/1659820)、[故障电流](https://www.osti.gov/biblio/1870461)、稳定电压和[黑启动](https://www.ercot.com/files/docs/2025/06/20/ERCOT-Grid-Insights-Black-Start.pdf)。孤岛必须自己提供所有这些——电力系统的人知道这有多难。
AI 工作负载让这件事更难。在 xAI 的第一个孟菲斯站点,每秒数次 10 到 20 MW 的波动[磨损了涡轮轴](https://newsletter.semianalysis.com/p/what-is-so-hard-about-behind-the),直到 xAI 增加了 150 MW 的特斯拉 Megapack。ERCOT 这种规模的电网会稀释这样的波动,但在私有电厂,涡轮机真切地感受到它。随着机架级功率密度增长、波动变得更加剧烈,预计会有更多电池和其他储能。
然后是站点的燃料供应。[冬季风暴 Uri](https://www.ercot.com/files/docs/2021/02/24/2.2_REVISED_ERCOT_Presentation.pdf)的教训是,燃气电厂可能[一起失效](https://www.nerc.com/globalassets/standards/projects/2021-07/ferc-presentation-phase-2.pdf)。而且由于它们往往在现场只储存很少燃料(NERC 称天然气为[“准时制”燃料](https://www.nerc.com/globalassets/who-we-are/news/2025/10/2025-october-reliability-insights_final.pdf)),备用涡轮机并不总能帮上忙。[有时管道甚至还不存在](https://sourcenm.com/2026/08/17/proposed-gas-pipeline-for-project-jupiter-data-center-delayed-to-2027-filings-show/)。太阳能用太阳和天气来换取这种燃料风险,而在吉瓦级规模下,你需要大量电池。要让一个 1 GW 园区撑过一个 14 小时的冬夜,需要从电池获取 14 GWh,[大约相当于截至 6 月 ERCOT 电网上所有电池储能的一半](https://modoenergy.com/research/en/ercot-bess-buildout-16-gw-q2-2026-growth-slowing)。对于一个追求 100% 正常运行时间的孤岛来说,覆盖阴雨一周或又一次 Uri 这样的罕见事件会变得昂贵。
无论使用何种燃料,你都希望孤岛能够优雅且可预测地失效,没有单点故障。即便是可靠性几乎达到电厂极限的核能,也仅运行约[92% 的时间](https://www.eia.gov/todayinenergy/detail.php?id=65904),主要是因为每个反应堆都要停机数周来换料。在这种情况下,冗余实际上意味着再来一整座电厂。因此,表后配置倾向于模块化,比如计划在 Shackelford 建设的[500 多台约 4 MW 的燃气发动机](https://newsletter.semianalysis.com/p/what-is-so-hard-about-behind-the),尽管数百台发动机维护起来可能很麻烦。
对于一个孤岛来说,完全可靠性非常、非常困难,但一些实验室和超大规模企业已经表明,在可靠性和速度之间被迫选择时,他们会选择速度。Meta 已经转向使用[没有备用发电机](https://newsletter.semianalysis.com/p/meta-superintelligence-leadership-compute-talent-and-data)的帐篷,而 SemiAnalysis 发现买家[越来越愿意接受停电](https://newsletter.semianalysis.com/p/what-is-so-hard-about-behind-the),一些孤岛设计的目标可用性低至 99%,即每年约 88 小时的停机时间。鉴于[GPU 时间如此昂贵](https://newsletter.semianalysis.com/p/the-great-gpu-shortage-rental-capacity),这仍然胜过等待数年,因此对许多买家来说,即使混乱,临时孤岛也是合理的。
一个将多个工厂和园区连接在一起的“私有电网”可以收回孤立孤岛所放弃的许多好处,至少在理论上是这样。但如今,德克萨斯州和大多数州一样,只允许在严格限制内建设自己的电力:
- **自给自足:**如果你只为自己、你的员工或你的租户供电,并且没有人转售电力,你就[不是公用事业公司](https://tcss.legis.texas.gov/resources/UT/htm/UT.31.htm)。
- **运营私有使用网络:**你也可以依靠共址发电运行,[将多余电力出售给电网](https://www.jdsupra.com/legalnews/private-use-networks-self-generation-4515203/),并在电力不足时从电网取电。这就是 Armstrong County 园区以及西德克萨斯园区一旦并网后的模式。
- **卖给邻居:**如果你卖给马路对面的工厂,你需要[零售电力供应商证书](https://www.puc.texas.gov/agency/rulesnlaws/subrules/electric/25.107/25.107.pdf)。
- **架设电线:**自己建线路,你很可能就会闯入当地公用事业公司的[服务区域](https://www.puc.texas.gov/agency/rulesnlaws/subrules/electric/25.101/25.101.pdf)。
卡托研究所的[消费者监管电力](https://www.cato.org/briefing-paper/case-consumer-regulated-electricity-private-electricity-grids-offer-parallel-path)提案将放宽这些限制,允许私营公用事业公司通过自己的网络为多个客户提供服务。这并非全新想法;犹他州的[SB 132](https://le.utah.gov/Session/2025/bills/enrolled/SB0132.pdf) 允许 100 兆瓦或以上的负载签订完全离网系统的合同。德克萨斯州目前不允许这样的网络服务多个客户,但如果电力需求仍然无法满足,我预计更宽松的州将凭借这种“狂野西部”式的公用事业结构赢得更大份额的建设。(然而,你也可能面临公用事业公司“[死亡螺旋](https://www.utilitydive.com/news/as-data-centers-go-off-grid-utilities-face-new-cost-and-planning-risks/811944/)”的风险,电网的固定成本将落在越来越少的客户身上。)
另一面是公用事业公司自己建设孤岛。在 ERCOT 之外,El Paso Electric 计划在 Meta 的新园区旁安装[来自 ERock 的 813 台小型燃气发电机](https://interchange.puc.texas.gov/Documents/59076_1_1564187.PDF)(总计 366 兆瓦),并由 Meta 出资,作为孤岛运行长达五年。这是一个选项,因为与 ERCOT 内的输配电公用事业公司不同,它仍然拥有发电厂。孤岛期结束后,它会将电厂接入其电网,并可能寻求将成本分摊给所有客户,尽管在 9 月,行政法官[建议](https://interchange.puc.texas.gov/Documents/59076_176_1686197.PDF)只有保护这些客户的情况下才批准。
任何以燃料运行的东西都需要与其运行方式相匹配的空气许可证,因此仅获准用于紧急情况的柴油发电机不能一直运行。乐观情况下,如果一切顺利,许可证可以[很快拿到](https://newsletter.semianalysis.com/p/what-is-so-hard-about-behind-the)。但有时[并非如此](https://news.constructconnect.com/oracle-drops-gas-turbine-plan-for-its-massive-project-jupiter-turns-to-bloom-fuel-cells)。在德克萨斯州,冻结令现在阻止数据中心获得许可证,直到审计完成。
我们还应该问,表后“桥接”天然气究竟在桥接什么。xAI 在孟菲斯的前两个数据中心以不同方式回答了这个问题。第一个在电网到来之前,利用[现有燃气总管](https://www.mlgw.com/images/content/files/pdf/new/xAI%202025%20Update.pdf)运行临时涡轮机,随后[开始拆除它们](https://memphischamber.com/blog/press-release/xai-phase-one-substation-63-providing-150mw-of-power-to-facility/)。在此过程中,它[在无空气许可的情况下运行了数十台](https://www.selc.org/news/inside-memphis-fight-against-xai/)。对于第二个,xAI 将电厂建在[州界对面](https://www.wsmv.com/2026/04/14/naacp-sues-xai-alleging-unlawful-operation-gas-turbines-southaven/)的密西西比州南黑文,但一项7月的命令[要求全部69台涡轮机在2027年中期退役](https://tippahnews.com/mississippi-news/mississippi-sets-retirement-schedule-for-xai-turbines-allows-some-to-run-past-deadline/),取而代之的是一座永久性1.2 GW电厂。一座桥通向电网,另一座通向xAI自有的发电厂。
然而在8月,联邦田纳西河谷管理局(TVA)[同意](https://www.aol.com/articles/tva-board-approves-agreement-spacexai-234602000.html)也直接为该数据中心供电。事实证明,想远离电网很难!
### 灵活性可以换来什么
一个能够自我维持运行的园区,对电网来说也可能更容易接纳,甚至更受欢迎。它是一个大型付费客户,其需求在电力紧张时可以“灵活”调整,无论是削减用电量还是输出多余电力。xAI正是以此在其第二个孟菲斯站点获得了电网供电的批准。使其灵活承诺可信的,是它能够依靠自身电力承载全部负荷达四小时,以及孟菲斯电力、燃气和水务公司CEO所称的[“全球最大的并网电池系统”。](https://www.canarymedia.com/articles/batteries/xai-massive-battery-memphis-data-center)
电池只是实现灵活性的一种方式。园区还可以将[计算任务转移](https://cloud.google.com/blog/products/infrastructure/using-demand-response-to-reduce-data-center-power-consumption)到其他时段或其他数据中心,或切换到自有发电机。那么灵活性可以释放多少空间?[杜克大学的Tyler Norris及其同事](https://nicholasinstitute.duke.edu/publications/rethinking-load-growth)估计,在抛开输电限制的情况下,如果新增负荷愿意放弃其年度电网用电量的0.5%,ERCOT可以在不新增发电的情况下增加约10 GW的新负荷。由于平均削减持续约两小时,这也恰好是电池式的。在此基础上,[Camus、encoord和普林斯顿大学对PJM(美国最大电力市场)的研究](https://www.camus.energy/flexible-data-center-report)发现,每新增1 GW数据中心负荷,使其中20%的负荷具备灵活性,每年可为其他客户节省7800万美元,而其余80%自带容量则可使他们的账单再减少3.26亿美元。
要使灵活性有价值,重要的是在运营商需要时它始终可用。在德克萨斯州,发电机组已经按照[连接并管理](https://www.utilitydive.com/news/connect-and-manage-grid-interconnection-ferc-ercot-transmission-planning/698949/)模式这样运作。简而言之,只要ERCOT能在线路拥堵时将其削减,它们就可以提前接入。发电机组可以接受这一点,因为最坏情况也不过是一段时间内少卖些电,但一个已向客户承诺正常运行时间的数据中心则更具挑战性。一些负荷,如[比特币矿工](https://www.cnbc.com/2023/09/06/texas-paid-bitcoin-miner-riot-31point7-million-to-shut-down-in-august.html),已经做出了这种取舍。[Batch Zero](https://www.ercot.com/files/docs/2026/06/18/ERCOT-Trending-Topic-New-Batch-Connection-Process-for-Large-Electricity-Users.pdf)为此为大型负荷提供了两条可选路径:
- **自带电力:**园区可以将自有发电厂计入其规模,只要在电厂故障时能在[一分钟内](https://www.willkie.com/publications/2026/06/ercot-approves-implementing-new-batch-zero-process-for-large-load-interconnections)削减负荷。到目前为止,已有11个Batch Zero项目选择了这条路径。
- **同意削减:**园区可以按其申请的全部电量取电,但每当线路满载时,ERCOT可以自动削减超出其保证份额的部分。
园区也可以付费给邻居,让它们削减负荷。在 PJM,[Google 正在资助 Voltus](https://www.utilitydive.com/news/google-virtual-power-plant-vpp-pjm-voltus/821838/),在一个区域内聚合最多 100 MW 的电池、恒温器和其他灵活设备。但 PJM 只计算它相信该资源池能交付的部分,而且这是一项[容量交易](https://www.latitudemedia.com/news/google-is-voltus-first-bring-your-own-capacity-customer/),在峰值时帮助整个电网,而不是解决某一条拥堵线路的问题。这在得克萨斯更难推销,那是一个[仅能量](https://www.ercot.com/files/docs/2025/10/16/ERCOT-Grid-Insights-Energy-Only-Wholesale-Market.pdf)市场,输电是主要瓶颈。得克萨斯还[禁止](https://interchange.puc.texas.gov/Documents/58881_105_1628395.PDF)共址园区因[类似服务](https://interchange.puc.texas.gov/Documents/59220_69_1668205.PDF)获得报酬,因为根据 [SB 6](https://www.utilitydive.com/news/texas-law-gives-grid-operator-power-to-disconnect-data-centers-during-crisi/751587/),它们本来就必须在 ERCOT 要求时断电。
即便如此,得克萨斯在分布式资源方面一直走在前列。一个 [Base Power](https://a16z.com/announcement/investing-in-base-power/) 机队,[与合作社 GVEC 共同运营](https://www.gvec.org/gvec-and-base-power-partnership/),首次尝试就通过了 ERCOT 的试点测试,可直接向批发市场售电,如今正扩展到 50 MW。这些试点很重要;规划者需要这类证明,才会围绕这些机队进行设计。但一旦获得信任,分布式资源就能迅速增加容量,无需等待昂贵的新“[调峰](https://en.wikipedia.org/wiki/Peaking_power_plant)”电厂、线路、变电站或漫长的并网流程。
其他硬件也能帮上忙。与传统的钢制设备不同,固态变压器使用半导体开关,因此软件可以测量并引导流经它的电力。再加上诸如重新导线(用更高容量的导线更换线路)和动态线路额定值(根据实际天气而非最坏情况来评定线路)等网络升级,这种更高的可见性和控制力可以从纳税人已经出资建设的线路中榨取更多容量。这也是我们投资 [Heron Power](https://a16z.com/announcement/investing-in-heron-power/) 的一个重要原因,该公司将在[得克萨斯西部与 RWE 合作的电池站点](https://www.globenewswire.com/news-release/2026/09/21/3365509/0/en/heron-power-and-rwe-sign-pilot-agreement-outline-path-to-multi-year-capacity-reservation.html)安装其固态变压器。
问题在于,这些工具大多帮助运营商维持日常运行,但规划者在确定升级规模时并不总是把它们计算在内。不过,物理学上并没有什么强制要求如此。9 月[Piq Energy 的一项研究](https://www.piqenergy.com/research/piq-energy-base-power---flexibility-without-curtailment.pdf)利用 [Base Power](https://a16z.com/announcement/investing-in-base-power/) 关于潜在机队的数据发现,约 80 MW 的家用电池,如果战略性地选址以缓解输电约束,就能解决沃思堡附近一个假设的新建 100 MW 数据中心所引发的所有过载问题。
Batch Zero 尚未考虑这类事情。它[仍然按照灵活园区的全部计划负荷来规划升级](https://epeconsulting.com/epe-intelligence/news/new-pathway-and-deadlines-in-ercot-s-large-load-batch-zero-provisional-controllable-load-resource-pclr),因为这条路径是通往稳固服务的桥梁,而且没有选项可以永久保持灵活以换取更小的升级。规划者本可以通过计入灵活性和其他缓解相同瓶颈的资源来缩小升级规模,前提是这些资源能实时测量并可靠运行。
所有这些都通过更充分地利用已有的基础设施来节省时间和金钱,但无论多少灵活性,都无法让电网免于为即将到来的所有需求建设更多发电和线路。
### 迈向数百吉瓦
现场发电和灵活性将决定接下来几个园区如何通电,但我交谈过的实验室和超大规模企业最担心的是规模。它们的电力团队往往分成两组:一组为未来几年挑选站点和设备,另一组则研究如何在 2030 年后接入数百吉瓦的额外容量。
这数量可真不小!假设全年运行100 GW,那就是876 TWh,[约占](https://www.eia.gov/energyexplained/electricity/use-of-electricity.php)该国2025年用电量的五分之一。以下是电力开发商[告诉能源信息署](https://www.eia.gov/todayinenergy/detail.php?id=67205)(EIA)他们计划在2026年全电网新增的装机容量。
每一种电源都有帮助,而且其中很大一部分本来就要送往得克萨斯州。但按去年的平均容量系数(电厂实际发电量与其最大发电量的比值)计算,计划中的[太阳能、风能](https://www.eia.gov/electricity/monthly/epm_table_grapher.php?t=epmt_6_07_b)和[天然气电厂](https://www.eia.gov/electricity/monthly/epm_table_grapher.php?t=epmt_6_07_a)每年大约能发150 TWh,约为那876 TWh的六分之一。
公平地说,那根天然气柱很可能低估了[正在建设的规模](https://www.distilled.earth/p/eight-new-gas-plants-will-power-amazons),因为[EIA的调查只统计并网电厂](https://www.eia.gov/survey/form/eia_860/instructions.pdf)。这样一来,已下单的75 GW自备电源设备中,很大一部分可能是这类图表会遗漏的“幽灵容量”。尽管如此,天然气仍然受欢迎,因为它通常在你需要时就能运行。也正因如此,制约因素主要在于[能否拿到设备](https://www.sec.gov/Archives/edgar/data/0001996810/000199681026000147/gevpressrelease2q26.htm),所以买家正转向[更容易找到](https://www.latitudemedia.com/news/is-the-gas-turbine-bottleneck-solving-itself/)的替代方案,比如[往复式发动机](https://www.powermag.com/data-centers-are-turning-to-gas-generators-for-prime-power-to-eliminate-long-lead-times-for-grid-connections/)和[燃料电池](https://investor.bloomenergy.com/press-releases/press-release-details/2026/Bloom-Energy-and-Oracle-Expand-Strategic-Partnership-to-Deploy-up-to-2-8-GW-to-Accelerate-AI-Infrastructure-Build-Out/default.aspx)。
话虽如此,在我看来天然气仍像是一个不完整的答案。我不是绿色和平组织的斗士,但让100 GW的天然气电厂以[平均利用率](https://www.eia.gov/tools/faqs/faq.php?id=74&t=11)全天候运行,将释放该国[能源相关排放](https://www.eia.gov/environment/emissions/carbon/)的近8%。而且,最容易获得的设备往往[效率更低](https://www.eia.gov/electricity/annual/html/epa_08_02.html)。此外,运营这样一支机组群可能会严重考验我们的天然气供应。
太阳能之所以有吸引力,是因为它已经具备了实验室所追求的生产规模。问题在于,产能主要在中国。2025年全球[新增超过600 GW](https://www.solarpowereurope.org/press-releases/new-report-global-solar-market-hits-new-record-of-664-gw-installation-in-2025-as-global-solar-fleet-passes-3-tw),但仅中国一国就生产了[全球超过80%的太阳能组件](https://www.iea.org/reports/solar-pv-global-supply-chains/executive-summary)和[电池电芯](https://www.iea.org/reports/global-ev-outlook-2026/manufacturing-and-trade)。据我所知,中国供应商并不太介意卖给我们,至少部分原因是他们把我们扩大规模看作“小打小闹”。然而,据报道中国[曾权衡](https://finance.yahoo.com/sectors/energy/articles/exclusive-china-weighs-curbs-exports-074947904.html)限制特定太阳能制造设备的出口。华盛顿方面也施加了压力,包括[强迫劳动货物扣留](https://pv-magazine-usa.com/2022/11/11/due-to-forced-labor-concerns-u-s-ports-have-blocked-over-1000-shipments-of-solar-components/)、[新关税](https://pv-magazine-usa.com/2026/08/07/us-announces-tariffs-minimum-import-price-on-polysilicon-imports/),以及[逐步取消风能和太阳能补贴](https://www.irs.gov/pub/irs-drop/n-25-42.pdf)。尽管如此,埃隆仍打算靠自己实现美国太阳能制造[每年200 GW](https://pv-magazine-usa.com/2026/01/26/elon-musk-at-wef-spacex-and-tesla-to-produce-100-gw-each-of-pv-per-year-in-the-u-s-this-decade/)的目标,[显然](https://x.com/elonmusk/status/1925872357781983522?s=20)是吃了“太阳能药”。
扩大非天然气的稳定电源要复杂得多。[功率提升](https://www.prnewswire.com/news-releases/georgia-power-agreement-with-google-to-provide-approximately-900-million-in-projected-benefits-for-customers-and-advance-nuclear-energy-in-georgia-302885197.html)和重启可以从现有核电机组中[榨取](https://www.reuters.com/business/energy/google-enters-massive-36-gw-power-deal-with-constellation-energy-2026-10-06/)一点额外电力,但真正的上行空间在于新建反应堆,正如[我们之前所论证的](https://a16z.com/how-to-scale-nuclear-power/)。[Meta](https://about.fb.com/news/2026/01/meta-nuclear-energy-projects-power-american-ai-leadership/)和[亚马逊](https://www.aboutamazon.com/news/sustainability/amazon-smr-nuclear-energy)已签署大单,但其中大部分容量仍只是选项和目标。到目前为止,军方一直是更强劲的[买家](https://breakingdefense.com/2026/08/army-taps-5-vendors-to-build-nuclear-microreactors-with-2-2b-pot/),在其基地建设微型反应堆,这正是工厂预制反应堆学会更快、更便宜的方式(要怪就怪EPC,跟NRC一样)。地热能也可以利用钻探(得克萨斯州对此很在行)来实现可重复的电力——[谷歌](https://fervoenergy.com/fervo-energy-and-google-sign-396-mw-ppa/)和[Meta](https://www.businesswire.com/news/home/20240826755596/en/Sage-Geosystems-and-Meta-Announce-Agreement-for-Next-Generation-Geothermal-Power-Generation)似乎相当感兴趣。
无论电力来源是什么,最终都还是要等待变压器、开关设备等其他设备,以及所有安装这些设备的施工队伍。大型电力变压器现在需要[两年多](https://www.powermag.com/transformers-in-2026-shortage-scramble-or-self-inflicted-crisis/)才能到货,而 FCC 已经[限制了新的外国制造逆变器](https://docs.fcc.gov/public/attachments/DA-26-786A1.pdf),同时还发布了[8 月的紧急命令](https://www.whitehouse.gov/presidential-actions/2026/08/declaring-a-national-emergency-to-secure-the-united-states-bulk-power-system/),可能进一步禁止中国制造的设备进入电网。
还得有人来铺设电线。2008 年,得克萨斯州监管机构下令建设[CREZ 线路](https://www.bakerinstitute.org/sites/default/files/2020-11/import/ces-pub-texascrez-111720.pdf)来输送西得克萨斯的风电,然后把 70 亿美元的成本分摊到每个电费缴纳者身上。如今他们正在批准更大的[765 kV 线路](https://www.texastribune.org/2026/08/28/transmission-lines-texas-energy/),但进展[比许多人期望的要慢](https://thetexan.news/issues/energy/here-s-the-current-state-of-the-plan-to-build-765-kv-transmission-lines-across/article_b7e8c5bb-35c7-42d7-bed9-04f243de1927.html)。其中一部分(不幸的是)只是典型的施工问题,在任何地方都又慢又贵,但除此之外还有无数额外的监管障碍。如果国会能通过,[华盛顿正在成形的联邦许可协议](https://www.energy.senate.gov/press/rep/lee-capito-heinrich-and-whitehouse-introduce-bipartisan-american-affordability-and-jobs-act-of-2026/)可能有助于推进事情。
诚然,我更有信心的是我们会需要大量电力,而不是它具体会以什么形式出现。我打赌会有几种方案,取决于地理位置和灵活性,被反复建设。也许现场燃气和电池会先支撑一个园区,直到它的电网接入到位,然后留下来作为备用,在电网紧张时灵活调节。太阳能现在在适合的地方被叠加使用,而地热和反应堆则要等它们得到验证后才会加入。
2030 年之后,就更难说最终是谁来建设和拥有所有这些电力了。一种答案是同一家公司既建电厂又建园区,Google [收购 Intersect](https://abc.xyz/investor/news/news-details/2025/Alphabet-Announces-Agreement-to-Acquire-Intersect-to-Advance-U-S--Energy-Innovation-2025-DVIuVDM9wW/default.aspx) 可能就暗示了这一点。另一种答案是像 Chevron 或 Williams 这样的石油和天然气公司成为[更广泛的电网建设者](https://travisfisher.substack.com/p/socrates-south-shows-the-benefits),不难想象它们,或者像 NRG 和 NextEra 这样的“新公用事业公司”,会建设服务多个园区的私有电网,前提是法律允许。
我还不知道会朝哪个方向发展,但它们很可能都会从同样的设备制造商那里采购。这个市场巨大,而且出人意料地缺乏能力来满足正在拐点式上升的需求。鉴于此,我看到初创公司切入的两大主要方式:
- **整合:**有些公司采用熟悉的硬件,在整合或商业模式上取胜,就像 Base Power 把家用电池作为可信的聚合资源来运营。
- **技术:**另一些公司带来新的、更优越的技术,早期采用者会在供应受限的市场中冒险尝试,这正是 Heron Power 想用固态变压器实现的目标。
第三种卖点,粗略地说就是“更差但更快”,在短缺时期很畅销,而且可能利润极其丰厚,但我会问这些利润被重新投资到了哪里,因为如果更广泛的供应赶上来了,竞争可能会变得很难。
确实,这样的短缺是初创公司的机会,但理解为什么现有企业不急于填补这些缺口会有所帮助。上一次对涡轮机繁荣的押注以 GE Power 220 亿美元的减记告终,并[帮助让 GE 的 CEO 丢掉了工作](https://www.bostonglobe.com/business/2018/10/01/john-flannery-out-general-electric-ceo/7QRPwEHxACP1PON92Vq8WJ/story.html)。初创公司必须挺过现有企业正在为之做准备的萧条期,还要在全球范围内与 Siemens Energy、Mitsubishi Power 以及印度和中国等地的各类供应商竞争。客户今天可能会为速度买单,但要留住他们,需要可靠、价格有竞争力的设备,以及一支知道自己在做什么的服务团队。
站点开发的每一步都很艰难。像 GE Vernova 这样的供应商现在仅为了预留一个制造档期就收取[不可退还的定金](https://www.gevernova.com/sites/default/files/gev_webcast_transcript_10222025.pdf),让开发商陷入[先有鸡还是先有蛋的困境](https://newsletter.semianalysis.com/p/what-is-so-hard-about-behind-the)。贷款方希望与可靠客户签订长期合同,而客户希望有可信的时间表,而该时间表又需要定金,开发商通常需要贷款方来提供这笔资金。引入一个未经证实的供应商,这个循环就更难闭合(而且一旦[断裂](https://amarillotribune.org/2025/12/12/fermi-stock-drops-after-termination-of-150-million-construction-agreement/),市场会迅速察觉)。即便循环闭合,其他一切仍然[必须顺利推进](https://www.bloomberg.com/news/articles/2026-09-24/oracle-cites-force-majeure-to-shield-itself-on-controversial-data-center)。
对于新供应商来说,即便只是进入园区的第一阶段也很困难。据我所知,几乎没有人在没有数千小时运行记录的情况下,为数据中心电力设备下过大额、确定的订单。例如,[FTAI 的大订单](https://www.globenewswire.com/news-release/2026/07/22/3331360/35538/en/ftai-announces-1-465-billion-gas-turbine-generator-set-order-through-j-f-power-systems.html)依托的是 CFM56,这款喷气发动机已累计[超过十亿飞行小时](https://www.safran-group.com/pressroom/cfm56-engine-fleet-surpasses-one-billion-engine-flight-hours-2019-06-04),而 Crusoe 则在 Boom 的首个发动机核心仍在[准备测试](https://aviationweek.com/business-aviation/aircraft-propulsion/boom-supersonic-gears-engine-prototype-core-testing)时[退出了 Boom](https://techcrunch.com/2026/09/25/crusoe-abandons-1-25b-plan-to-use-boom-turbines-at-ai-data-centers/)。因此,即便是打造出色技术的可信团队也面临冷启动问题。所以,让你的产品被设计进一个真实项目,远比早期定金或收入重要。
### 暂停之后
让我们回到得克萨斯州。我认为,核实 Batch Zero 中的项目是否[如其开发商所宣誓的那样](https://www.ercot.com/files/docs/2026/08/20/Batch-Zero-Verification-and-Audit-PUCT-Presentation.pdf)是公平的,可信的项目应该很快推进。广泛的许可冻结更难辩护,不过如果目标是在审计完成前不批准任何项目,我能理解该州的逻辑。话虽如此,我不喜欢它让一个已为其第一阶段提供资金的开发商——即便是自带电力的开发商——像投机者一样等待。得州现在就应该缩小范围,豁免仅为应急而获许可的发电机组以及已为其第一阶段提供资金的项目。该州环境监管机构于 10 月 19 日向州长提交的最新情况通报,是开启这场对话的明显起点。
一旦审计完成,正在进行的测试可以更加直接。一个缴纳了保证金、为其预留容量付费并达到分阶段里程碑的开发商,应该获得一个可以规划的并网日期。如果它同意在电网紧张时削减用电,也应该更早并网(Batch Zero [已经做了部分此类工作](https://epeconsulting.com/epe-intelligence/news/new-pathway-and-deadlines-in-ercot-s-large-load-batch-zero-provisional-controllable-load-resource-pclr))。更进一步,负责建设升级的公用事业公司或许应该像开发商一样为延误负责。也许得州[可以](https://capitol.texas.gov/tlodocs/86R/analysis/pdf/SB01938F.pdf)[开放竞争性投标渠道](https://www.texasenergyandpower.com/p/texas-competes-on-everything-except),并设置成本上限和罚则。
无论得州如何决定,电力供应很可能在数年内保持紧张,而瓶颈会不断转移。重要的是,数据中心所需的设备并非那么独特;它们只是最先触及这些限制。从这个意义上说,它们是一次完美的预演,因为我们在这个国家想要(重新)建设的大部分东西都将依赖同样的东西运行。
这也让它成为一个我不愿浪费的机会。AI 公司几乎愿意为电力支付任何价格,而且几乎可以在任何地方建设,因此它们可以帮助资助电网本就需要的升级。分析师 Hans Royal 估计,推理业务可以为电力支付荒谬的 [每兆瓦时 5,600 美元](https://computeheatrate.com/chr-index),仍能获得不错的回报,几乎是 [美国工业客户平均支付价格](https://www.eia.gov/electricity/monthly/update/end-use.php)的 60 倍。这可能会推高当地的天然气、电力、劳动力和材料价格,但它所资助的基础设施价值可能远超于此。诚然,这种论点目前很难推销出去。在 8 月的一项民意调查中,[56% 的得克萨斯州选民](https://texaspolitics.utexas.edu/blog/new-ut-texas-politics-project-poll-finds-talarico-leading-paxton-abbott-leading-hinojosa-continued-resistance-to-data-centers-2)表示,更多数据中心会推高当地能源账单。
这种担忧是合理的,因为诚实的答案是这取决于谁来买单。在理想情况下,开发商承担所有相关升级费用,并为其预留的容量付费,支持共享基础设施建设,从而改善我们所有人的电网。即使只是购买大量电力,一个园区也能将服务区域的固定成本分摊到更多销售上,从而缩小其他所有人的分摊份额。[劳伦斯伯克利国家实验室发现](https://eta-publications.lbl.gov/sites/default/files/2026-03/retail_price_trends_2026_edition.pdf),从 2019 年到 2025 年,负荷增长最快的州在扣除通胀后,平均电价普遍下降。想象一下,涌入 AI 的所有资金帮助下一家工厂接入电网、下一户家庭实现电气化,而无需承担难以负担的账单(也不会出现电网过于紧张、无法支持电动汽车或家用机器人的情况)。那户家庭可能永远不会打开聊天机器人,却仍能从中受益匪浅。我们已经看到了一些早期版本:
- **印第安纳州:**为亚马逊 New Carlisle 园区供电的公用事业公司[提议利用数据中心收入将家庭账单每年削减约 100 美元](https://www.wvpe.org/wvpe-news/2026-08-26/i-m-announces-electric-rate-cut-plan-enabled-by-data-centers)。
- **阿拉巴马州:**塔斯卡卢萨县的一个园区将在 20 年内支付 [2.7 亿美元的社区福利](https://yellowhammernews.com/tuscaloosa-county-lands-3-2-billion-data-center-first-in-alabama-to-pay-community-benefits-270-million-over-20-years/)。作为回报,它将获得约 3.145 亿美元的税收减免,不过学校税(约 1.315 亿美元)不予减免。
- **宾夕法尼亚州:**一位开发商[提出,如果其园区获批,将向 Hazle Township 的每户家庭提供 1 万美元](https://www.yahoo.com/news/us/articles/data-center-developer-offers-10-114000653.html),一些居民(或许正确地)称之为贿赂。
如果超大规模企业作为在当地设立园区的一部分,承担整个城镇的电费,我不会感到惊讶。大多数州,[包括得克萨斯州](https://texas.public.law/statutes/tex._util._code_section_36.003),不允许公用事业公司单独针对某个城镇,但在像[俄勒冈州的 Umatilla Electric](https://www.cobank.com/knowledge-exchange/earning-the-business-by-demonstrating-network-value) 这样数据中心消耗大部分电力的小型公用事业公司,为其大约 [17,000 个电表](https://www.umatillaelectric.com/about-us/who-we-are/)中的每个[典型家庭](https://www.eia.gov/electricity/sales_revenue_price/pdf/table_5A.pdf)承担账单,每年只需花费约 3,000 万美元。这样的协议可能划算,因为这不到一个 1 GW 园区[拥有和运营成本](https://epoch.ai/data-insights/ai-datacenter-cost-breakdown)的 0.5%。
在达成交易时,仔细阅读细则也很重要。在阿肯色州,谷歌同意为其数据中心向 Entergy 的太阳能电站预付 [4.43 亿美元](https://www.arkansasonline.com/news/2026/sep/03/googles-contribution-was-at-issue-in-arkansas/)。问题在于,Entergy 将其计为[电力预付款](https://www.arkansasonline.com/news/2026/sep/03/googles-contribution-was-at-issue-in-arkansas/),因此它仍然可以向所有客户收取该电站的全部成本外加回报。而 Entergy [起诉了两家报纸](https://www.cnbc.com/2026/09/02/google-data-center-arkansas-entergy-court.html),以阻止它们报道该合同,后在法官拒绝后撤诉。即使这笔交易合法,这也不是与社区建立信任的方式。
一个与周边社区坦诚相待并真正兑现承诺的园区,会赢得人们的支持,下一个园区也更容易被接纳。如果做错了,其他所有人都可能为此付出代价。在 PJM,[市场监测机构表示,数据中心占最新容量账单的 38%](https://www.utilitydive.com/news/pjm-data-centers-capacity-auction-imm-bowring/825626/),约 63 亿美元。我要说清楚:这一切都不构成全面暂停的理由。但这确实意味着,在我们扩大开发规模时,需要把规则和激励措施制定得当。AI 公司已经[承诺自行承担费用](https://www.powermag.com/hyperscalers-sign-white-house-pledge-to-fund-data-center-power-grid-upgrades/),而且大多数公司宁愿接入电网,也不愿回避它。让它们这样做,而不是把相关支出推向私人孤岛,或把我们无论如何都需要的基础设施升级费用转嫁到其他人的账单上,这似乎对所有人来说都是理想的选择。
得克萨斯州最先开辟了能源快车道;现在它有机会向我们展示一条更好的道路,我相信它会做到的。电力是美国检验自己是否仍能建设的地方。
如果你正在研究能源问题,我很想向你学习。请联系我。

[立即订阅](https://www.a16z.news/subscribe?)

_本通讯仅供参考,不应作为法律、商业、投资或税务建议加以依赖。此外,本内容不构成投资建议,也不适用于任何 a16z 基金的投资者或潜在投资者。本通讯可能链接到其他网站,或包含从第三方来源获得的其他信息——a16z 未对此类信息的当前或持续准确性进行独立核实,也未作出任何陈述。如果本内容包含第三方广告,a16z 未对此类广告进行审查,也不认可其中任何广告内容或相关公司。所提及、引用或描述的任何投资或投资组合公司并不代表 a16z 所管理工具的全部投资;请访问 [https://a16z.com/investment-list/](https://a16z.com/investment-list/) 查看完整投资列表。其他重要信息可在 [a16z.com/disclosures](http://a16z.com/disclosures) 找到。您收到本通讯是因为您此前已选择订阅;如果您希望退出未来的通讯,可以立即取消订阅。_
## 正文 · 原文

At the end of 2024, Texas’ grid operator had [63 GW of large new customers](https://www.latitudemedia.com/news/ercots-large-load-queue-has-nearly-quadrupled-in-a-single-year/) in its queue. By this June, that [figure reached 474 GW](https://www.utilitydive.com/news/texas-hits-pause-data-center-interconnections/827046/), more than five times record peak demand, about 90% of it data centers.
Then Texas [hit pause](https://www.texastribune.org/2026/08/03/texas-data-center-project-audit-greg-abbott/). What began as [directions](https://gov.texas.gov/news/post/governor-abbott-directs-puc-and-ercot-to-shield-texans-from-data-center-infrastructure-costs) for data centers to pay for their own grid upgrades has since escalated to a [freeze](https://gov.texas.gov/news/post/governor-abbott-directs-tceq-to-halt-data-center-permits) on all new permits. If you care about AI and American reindustrialization, this is important to understand. Why did it do this?
The upcoming [election](https://www.texastribune.org/2026/09/21/texas-data-center-moratorium-water-energy/) is part of it, but the underlying issues are worth digging into. The first is the interconnection queue. This pause is just as much about how projects apply for approval as it is about the projects themselves. Developers routinely spam projects across several sites, and many are [speculative](https://www.bloomberg.com/news/articles/2026-08-12/most-electricity-sought-for-ai-data-centers-in-us-will-never-materialize) builds with no customer yet. Often these new developers have never even plugged in a GPU, let alone a power plant, so dealing with low-quality submissions puts ERCOT, which runs most of the state’s grid, in a [tough spot](https://www.texastribune.org/2026/06/17/texas-ercot-data-center-energy-grid/). Planners can’t tell which are real, and they don’t want ratepayers paying for upgrades built for projects that never show up.
The second issue is community alignment. Noise, water, emissions, and power bills are [top of mind](https://www.pewresearch.org/short-reads/2026/09/22/americans-views-of-data-centers-have-turned-more-negative/) for anyone who lives nearby. And regardless of the [reality](https://www.microsoft.com/en-us/microsoft-cloud/blog/2024/12/09/sustainable-by-design-next-generation-datacenters-consume-zero-water-for-cooling/) of these issues, people deserve [straight answers](https://www.texastribune.org/2026/09/14/texas-data-centers-abbott-water-survey-enforcement/) when a data center comes to town, and they [haven’t always gotten them](https://www.ksat.com/news/texas/2026/08/10/while-data-centers-boom-in-texas-their-plans-stay-shrouded-in-secrecy/) ([as of June, only 28 of 377 companies](https://www.texastribune.org/2026/08/03/texas-data-center-project-audit-greg-abbott/) had answered a state survey on their resource use). In Hood County, commissioners were asked to support a tax waiver for “Project Patriot” without knowing [who it was](https://www.texastribune.org/2026/06/02/texas-data-centers-hood-county-local-control-rural-water-power/). To be fair, code names are common while companies shop for sites since a famous buyer can drive up prices, but that logic becomes tougher to defend once officials vote on tax breaks.
This broader political backdrop is important, but I’ll be focusing on just the [energy](https://a16z.com/decentralizing-the-electric-grid/) side: why it’s so hard to power a data center, all the ways developers are trying to do it anyway, and where things are likely headed from here.
### What connecting to the grid actually means
How Amazon purchases power helps illustrate what’s changed. For the last decade, they’d find a utility with network capacity, sign up as a large load, then enter long-term contracts to [match](https://sustainability.aboutamazon.com/renewable-energy-methodology.pdf) their use on paper. These were often [purely financial](https://www.epa.gov/green-power-markets/financial-ppa), meaning you didn’t always have to prove the power could reach you, and the grid connection was easy because utilities typically had spare network capacity.
That’s no longer true. The grid is stretched thin, and almost any new large load or generator now requires an upgrade. [Power flows across every connected path](https://www.ferc.gov/sites/default/files/2020-05/ch1-3_0.pdf), so buying from one plant doesn’t reserve a route to your building (there are even markets for transmission congestion [rights](https://www.ercot.com/mktinfo/crr)).
So buyers first went after firm power (available around the clock) they could claim more directly, typically by [restarting retired plants](https://www.constellationenergy.com/news/2024/Constellation-to-Launch-Crane-Clean-Energy-Center-Restoring-Jobs-and-Carbon-Free-Power-to-The-Grid.html) or by building next to existing ones. [Neither was a true escape](https://www.utilitydive.com/news/constellation-three-mile-island-crane-nuclear-ferc-waiver/821836/), and some of the loudest fights are over “colocation,” which [ERCOT describes](https://www.ercot.com/files/docs/2026/06/01/ERCOT-Trending-Topic-Grid-Reliability-Data-Centers-Net-Metering-Arrangements.pdf) as drawing power from a neighboring plant before it reaches the grid:
- **Susquehanna:** Amazon [bought a campus](https://www.ans.org/news/article-5842/amazon-buys-nuclearpowered-data-center-from-talen/) next to Talen’s nuclear plant, but FERC, the federal grid regulator, [rejected Talen’s bid](https://www.utilitydive.com/news/ferc-interconnection-isa-talen-amazon-data-center-susquehanna-exelon/731841/) to send it more power directly. Utilities had argued it would let the campus dodge grid fees and push fixed transmission costs onto everyone else. Under the [restructured deal](https://www.sec.gov/Archives/edgar/data/1622536/000162253626000066/tln-20260630.htm), Amazon [pays for delivery like any other customer](https://www.powermag.com/talen-amazon-launch-18b-nuclear-ppa-a-grid-connected-ipp-model-for-the-data-center-era/).
- **Freestone:** CyrusOne’s 760 MW campus beside Constellation’s gas plant needed [approval from the Public Utility Commission of Texas (PUCT)](https://tcss.legis.texas.gov/resources/UT/htm/UT.39.htm#39.169), since it effectively takes much of the plant’s output from everyone else. The PUCT [said yes](https://www.constellationenergy.com/news/2026/05/constellation-reports-first-quarter-2026-results.html) in May, as long as the campus can [cut its use](https://www.ercot.com/files/docs/2026/06/01/ERCOT-Trending-Topic-Grid-Reliability-Data-Centers-Net-Metering-Arrangements.pdf) or switch to backup power within 30 minutes of an ERCOT call.
- **Armstrong County:** Two Crusoe data centers, of 265 MW and 260 MW, share a wind farm that can produce at most about 265 MW. Both got the same rule as Freestone, so each must shed its entire load when ERCOT calls. Together, they'd cut nearly twice what the wind farm can actually produce, which Crusoe called [excessive](https://www.utilitydive.com/news/texas-approves-ai-data-center-co-location-next-to-wind-farm-with-curtailme/826617/), but the PUCT kept it.
Interconnection studies are how grid planners identify the wires and substations that need upgrades for any addition to the network. ERCOT used to review studies individually under rules [built for up to 50 large loads at a time](https://www.utilitydive.com/news/ercots-large-load-queue-jumped-almost-300-last-year-official/808820/), but 2025 brought 225 new requests by mid-November. So this June, the PUCT [approved a new batch approach](https://www.ercot.com/news/release/06182026-puct-approves-ercots). Under “Batch Zero,” ERCOT studies large loads of 75 MW or more together and allocates grid capacity among them.
Primarily, the studies ask what happens when something breaks. [NERC’s standard](https://www.nerc.com/pa/Stand/Reliability%20Standards/TPL-001-5.1.pdf), the baseline for reliability across North America, covers a broad set of outage scenarios, with [ERCOT adding its own requirements](https://www.ercot.com/files/docs/2026/05/29/June-1-2026-Plannng-Guide.pdf). Planners might simulate a transformer outage, then knock out a line or generator on top of that. Maintaining that reliability standard without shedding more load often means building additional infrastructure.
A newer risk is load unexpectedly dropping off the grid all at once. Many data centers switch to backup power at the first voltage dip to protect their hardware, so one bad fault can [pull an enormous load](https://www.jdsupra.com/legalnews/nerc-computational-load-updates-6702832/) off the grid in seconds. This [happened](https://insidelines.pjm.com/pjm-dominion-review-large-load-transfer-event/) earlier this year in Virginia, but was fortunately handled well.
Texas has even less room for error since ERCOT’s grid is largely [isolated](https://www.ercot.com/files/docs/2025/05/12/ERCOT-Grid-Insights-Electricity-Connection-and-Transfer.pdf). Under some conditions, ERCOT can lose only [about 3.2 GW](https://www.ercot.com/files/docs/2026/09/16/QSA-Results-and-Large-Load-IROL-Mitigation-Plans_final.pdf) of load at once before causing serious issues. A new voltage [ride-through rule](https://www.ercot.com/mktrules/issues/NOGRR282) now requires new data centers to stay connected through routine faults, and I’ve even heard of labs running dummy jobs after a training run fails just to keep load from dropping abruptly.
All of this can sound overly conservative, but the system was designed to put reliability ahead of cost. Put simply, the grid is built for the hot summer days and frigid winters when failure can mean life or death. That duty is what makes sizing the grid so hard; you build for a few peak hours but pay for it all year.
So who does pay for all this resilience? Texas typically splits transmission costs by each large customer’s demand during the grid’s summer peaks, so a big load that ramps down on the hottest afternoons can skip much of its share of the transmission bill. In July, regulators [proposed](https://interchange.puc.texas.gov/Documents/58000_8_1664102.PDF) counting all 12 monthly peaks instead and [charging](https://www.sos.texas.gov/texreg/archive/July242026/Proposed%20Rules/16.ECONOMIC%20REGULATION.html) large loads as if they ran at full size. In other words, large buyers would pay in proportion to how big their electricity pipes need to be, not how much is flowing through them.
Sounds simple enough, but splitting up costs is often the slowest part of interconnection. An upgrade built for one campus may also improve reliability for existing customers or make room for future growth that’s hard to value up front. And if any new development runs over budget or its load never shows up, everyone else is forced to cover whatever the developer’s commitments didn’t.
And those commitments are surprisingly cheap to make. The PUCT’s [new large-load rules](https://interchange.puc.texas.gov/Documents/58481_218_1684654.PDF), effective October 8, charge a flat $100,000 study fee plus a $50,000-per-MW deposit. The deposit weeds out some speculative projects, but it’s a thin filter. ERCOT can reassign a project’s capacity if it falls two years behind, but even a project that loses its capacity forfeits just 20% of the deposit (~$10 million on a 1 GW campus), plus whatever the utility has already spent. Most importantly, the deposit typically only backs the upgrades built for that project, not the more expensive regional lines whose cost everyone on the grid shares.
Fully connecting a large data center can take [5 to 10 years](https://arxiv.org/abs/2509.07218). So to accelerate deployment, a [phased connection approach](https://www.crusoe.ai/resources/newsroom/crusoe-expands-ai-data-center-campus-in-abilene-to-1-2-gigawatts) is becoming [more common](https://www.sdxcentral.com/news/xai-colossus-memphis-power-tva/), which gives the utility more gradual targets to plan around, as well as a way for the developer to prove they can handle everything they’ve asked for. All of this also assumes Batch Zero is moving, which it isn’t right now. ERCOT has even [paused approvals](https://www.ercot.com/files/docs/2026/08/20/Batch-Zero-Verification-and-Audit-PUCT-Presentation.pdf) for data centers of 75 MW or more to switch on, including 17 that had finished every other ERCOT step. Until the audit’s December report settles which projects are eligible, ERCOT can’t study them together, so the timeline is [“TBD”](https://www.ercot.com/files/docs/2026/09/11/14-Batch-Zero-Update.pdf) and [the load forecast is on hold](https://www.ercot.com/files/docs/2026/08/20/Batch-Zero-Verification-and-Audit-PUCT-Presentation.pdf).
### What bringing your own power solves
For a developer facing the interconnection queue, skipping the grid entirely looks appealing. Idle GPUs [cost](https://epoch.ai/data-insights/ai-datacenter-cost-breakdown) far more than the electricity to run them, so speed matters most — what folks call “time to power.” That’s why developers are planning to bring their own power on site (“behind-the-meter”), but almost always alongside a connection to the grid, or as a bridge to one:
- **Abilene:** The Oracle campus [runs on grid power with gas backup](https://www.crusoe.ai/resources/blog/an-inside-look-at-the-abilene-ai-data-center). Next door, Crusoe announced a 900 MW Microsoft campus with its own [on-site plant](https://www.crusoe.ai/resources/newsroom/crusoe-announces-new-900-mw-ai-factory-campus-in-abilene-texas-to-support-microsoft-ai-infrastructure), but the CEO of Lancium, Crusoe’s partner on the site, has [described gas generation as backup](https://westtexastribune.org/2026/05/05/lancium-ceo-addresses-abilenes-concerns/).
- **Shackelford County:** Vantage is building a campus with 1.4 GW of compute for Oracle and OpenAI, designed to run off-grid with [on-site gas](https://epoch.ai/publications/openai-stargate-where-the-us-sites-stand). But [its site plan](https://records.tceq.texas.gov/cs/idcplg?IdcService=TCEQ_EXTERNAL_SEARCH_GET_FILE&dID=8984027&Rendition=Web) includes a switchyard beside a high-voltage line — curious!
- **Pecos County:** Pacifico describes its planned microgrid (a behind-the-meter setup) at GW Ranch as [never drawing from ERCOT](https://www.pacificoenergy.com/gw-ranch), yet the site’s new owner, Amazon, says it’s [designed to join the grid](https://www.aboutamazon.com/news/company-news/amazon-pecos-county-texas-data-center-clean-energy-jobs) once it can connect.
- **West Texas:** Under a [20-year agreement](https://www.chevron.com/newsroom/2026/q2/chevron-signs-20-year-power-agreement-with-microsoft-for-west-texas-data-center), Chevron plans to build a dedicated gas plant beside a planned Microsoft campus that won’t initially touch the ERCOT grid, though Chevron has applied for a connection. Chevron now says the permit freeze [could push its final investment decision into 2027](https://www.semafor.com/article/10/01/2026/texas-politics-could-delay-investment-decision-for-huge-chevron-data-center), but it still expects first power by 2028.
- **Armstrong County:** The Google campus Crusoe is building beside a wind farm is [tied to both the farm and the grid from day one](https://www.crusoe.ai/resources/newsroom/crusoe-developing-data-center-campus-in-armstrong-county-texas), with Crusoe calling it “[across the meter](https://www.crusoe.ai/resources/blog/co-located-power-for-ai-data-centers),” which I’ll admit is catchy.
SemiAnalysis is already tracking [75 GW of equipment orders](https://newsletter.semianalysis.com/p/everyone-says-datacenter-moratoriums) for behind-the-meter assets, but every site I’m aware of intends to connect to the grid as soon as it can, chiefly because on-site power almost always costs more. So, once the grid is available, say in year five, you switch.
In the meantime, pairing on-site assets with even a partial connection is smart for the same reasons we built a grid in the first place. When one plant trips, the rest of ERCOT’s [1,460-plus generating units](https://www.ercot.com/files/docs/2022/02/08/ERCOT_Fact_Sheet.pdf) cover for it, but an islanded load (cut off from the grid) doesn’t have that luxury. The grid also provides things we take for granted, like [inertia](https://www.osti.gov/biblio/1659820), [fault current](https://www.osti.gov/biblio/1870461), steady voltage, and [black start](https://www.ercot.com/files/docs/2025/06/20/ERCOT-Grid-Insights-Black-Start.pdf). An island has to supply all of that itself — power systems folks know how hard this can be.
AI workloads make the job even harder. At xAI’s first Memphis site, swings of 10 to 20 MW several times a second were [wearing out turbine shafts](https://newsletter.semianalysis.com/p/what-is-so-hard-about-behind-the) until xAI added 150 MW of Tesla Megapacks. A grid ERCOT’s size dilutes swings like that, but on a private plant the turbines really feel it. Expect more batteries and other energy storage as rack-level power density grows and swings become more dramatic.
Then there’s fueling a site. [Winter Storm Uri’s](https://www.ercot.com/files/docs/2021/02/24/2.2_REVISED_ERCOT_Presentation.pdf) lesson is that gas plants can [fail together](https://www.nerc.com/globalassets/standards/projects/2021-07/ferc-presentation-phase-2.pdf). And since they tend to keep little fuel on site (NERC calls gas a [“just-in-time” fuel](https://www.nerc.com/globalassets/who-we-are/news/2025/10/2025-october-reliability-insights_final.pdf)), spare turbines don’t always help. [Sometimes the pipeline doesn’t even exist yet](https://sourcenm.com/2026/08/17/proposed-gas-pipeline-for-project-jupiter-data-center-delayed-to-2027-filings-show/). Solar trades that fuel risk for the sun and weather, and at gigawatt scale, you need a lot of batteries. Keeping a 1 GW campus running through one 14-hour winter night takes 14 GWh from batteries, [about half of all the battery storage on ERCOT’s grid](https://modoenergy.com/research/en/ercot-bess-buildout-16-gw-q2-2026-growth-slowing) as of June. For an island seeking 100% uptime, covering rare events like a cloudy week or another Uri gets expensive.
Whatever the fuel, you want an island that fails gracefully and predictably, with no single point of failure. Even nuclear, about as reliable as power plants get, runs only about [92% of the time](https://www.eia.gov/todayinenergy/detail.php?id=65904), mostly because each reactor goes offline for weeks to refuel. Redundancy in this case means effectively an entire second power plant. Thus, behind-the-meter setups favor modularity, like the [more than 500 gas engines](https://newsletter.semianalysis.com/p/what-is-so-hard-about-behind-the) of about 4 MW each planned for Shackelford, though hundreds of engines can be a pain to maintain.
Full reliability for an island is very, very hard, but some labs and hyperscalers have shown that, forced to choose between reliability and speed, they’ll pick speed. Meta has turned to tents with [no backup generators](https://newsletter.semianalysis.com/p/meta-superintelligence-leadership-compute-talent-and-data), and SemiAnalysis finds buyers [growing more willing to accept outages](https://newsletter.semianalysis.com/p/what-is-so-hard-about-behind-the), with some island designs aiming for as little as 99% uptime, or about 88 hours of downtime a year. With [GPU time this expensive](https://newsletter.semianalysis.com/p/the-great-gpu-shortage-rental-capacity), that still beats years of waiting, so for many buyers a temporary island makes sense even if it’s messy.
A “private grid” that ties several plants and campuses together can take back many of the benefits a lone island gives up, at least in theory. But today Texas, like most states, only allows building your own power within tight limits:
- **Supplying yourself:** You [aren’t a utility](https://tcss.legis.texas.gov/resources/UT/htm/UT.31.htm) if you supply only yourself, your employees, or your tenants, and nobody resells the power.
- **Running a private use network:** You can also run on colocated generation, [sell the surplus into the grid](https://www.jdsupra.com/legalnews/private-use-networks-self-generation-4515203/), and draw from it when you fall short. This is the model for the Armstrong County campus, and for the West Texas one once it connects.
- **Selling to a neighbor:** If you sell to the factory across the road, you need a [retail electric provider certificate](https://www.puc.texas.gov/agency/rulesnlaws/subrules/electric/25.107/25.107.pdf).
- **Stringing a wire:** Build your own line, and you’re probably running into the local utility’s [service territory](https://www.puc.texas.gov/agency/rulesnlaws/subrules/electric/25.101/25.101.pdf).
The Cato Institute’s [consumer-regulated electricity](https://www.cato.org/briefing-paper/case-consumer-regulated-electricity-private-electricity-grids-offer-parallel-path) proposal would loosen those limits by allowing private utilities to serve multiple customers across their own network. This isn’t an entirely new idea; Utah’s [SB 132](https://le.utah.gov/Session/2025/bills/enrolled/SB0132.pdf) lets loads of 100 MW or more contract for a fully off-grid system. Texas currently doesn’t let a network like this serve multiple customers, but if the demand for power remains insatiable, I’d expect the more permissive states to win larger chunks of the buildout with this “Wild West” utility structure. (However, you may also risk a utility “[death spiral](https://www.utilitydive.com/news/as-data-centers-go-off-grid-utilities-face-new-cost-and-planning-risks/811944/),” with the grid’s fixed costs falling on fewer and fewer customers.)
The flip side is a utility building the island itself. Outside ERCOT, El Paso Electric plans to put [813 small gas generators from ERock](https://interchange.puc.texas.gov/Documents/59076_1_1564187.PDF) (366 MW in all) beside Meta’s new campus and run them as an island, on Meta’s dime, for up to five years. This is an option because, unlike the transmission and distribution utilities inside ERCOT, it still owns power plants. After the island period, it would connect the plant to its grid and could seek to spread the cost across all its customers, though in September administrative judges [recommended](https://interchange.puc.texas.gov/Documents/59076_176_1686197.PDF) approval only if those customers are protected.
Anything that runs on fuel also needs an air permit that matches how it operates, so a diesel generator permitted only for emergencies can’t run all the time. Optimistically, permits can [come fast](https://newsletter.semianalysis.com/p/what-is-so-hard-about-behind-the) when things work. Sometimes they [don’t](https://news.constructconnect.com/oracle-drops-gas-turbine-plan-for-its-massive-project-jupiter-turns-to-bloom-fuel-cells), though. In Texas, the freeze now blocks them for data centers until the audit is done.
We should also ask what instances of behind-the-meter “bridge” gas are actually bridging to. xAI’s first two Memphis data centers answer that in different ways. The first ran temporary turbines off [an existing gas main](https://www.mlgw.com/images/content/files/pdf/new/xAI%202025%20Update.pdf) until the grid arrived, then [began removing them](https://memphischamber.com/blog/press-release/xai-phase-one-substation-63-providing-150mw-of-power-to-facility/). Along the way, it [ran dozens without air permits](https://www.selc.org/news/inside-memphis-fight-against-xai/). For the second, xAI built its plant [across the state line](https://www.wsmv.com/2026/04/14/naacp-sues-xai-alleging-unlawful-operation-gas-turbines-southaven/) in Southaven, Mississippi, but a July order [requires all 69 turbines to retire](https://tippahnews.com/mississippi-news/mississippi-sets-retirement-schedule-for-xai-turbines-allows-some-to-run-past-deadline/) by mid-2027 as a permanent 1.2 GW plant goes up in their place. One bridge led to the grid, and the other to a power plant of xAI’s own.
In August, though, the federal Tennessee Valley Authority (TVA) [agreed](https://www.aol.com/articles/tva-board-approves-agreement-spacexai-234602000.html) to serve that data center directly, too. Turns out it’s hard to stay away from the grid!
### What flexibility can buy
A campus that can keep itself running can also be easier for the grid to accommodate, even welcome. It’s a large paying customer whose demand can “flex” when power is tight, whether by cutting its draw or exporting surplus power. This is how xAI got approved for grid power at its second Memphis site. What made its promise to flex credible was its ability to carry its entire load for four hours on its own power, and what the CEO of Memphis Light, Gas and Water called [“the world’s largest grid-connected battery system.”](https://www.canarymedia.com/articles/batteries/xai-massive-battery-memphis-data-center)
Batteries are only one way to flex. A campus can also [shift computing](https://cloud.google.com/blog/products/infrastructure/using-demand-response-to-reduce-data-center-power-consumption) to other hours or data centers, or switch to its own generators. So how much room could flexibility open up? [Tyler Norris and colleagues at Duke](https://nicholasinstitute.duke.edu/publications/rethinking-load-growth) estimated that, setting transmission limits aside, ERCOT could add about 10 GW of new load without new generation if that load gave up 0.5% of its yearly grid electricity. Since the average cutback lasts about two hours, it’s also conveniently battery-shaped. Building on this, a [study of PJM (the largest US power market) by Camus, encoord, and Princeton](https://www.camus.energy/flexible-data-center-report) found that for each GW of new data center load, making 20% of that load flexible would save other customers $78 million a year, while bringing its own capacity for the other 80% would keep another $326 million off their bills.
For flexibility to be valuable, it’s important that it’s always available when operators need it. In Texas, generators already work this way under [connect-and-manage](https://www.utilitydive.com/news/connect-and-manage-grid-interconnection-ferc-ercot-transmission-planning/698949/). Put simply, they can hook up early as long as ERCOT can cut them down when lines are congested. Generators can live with that because at worst they sell less for a while, but a data center that has promised its customers uptime is more challenging. Some loads, like [Bitcoin miners](https://www.cnbc.com/2023/09/06/texas-paid-bitcoin-miner-riot-31point7-million-to-shut-down-in-august.html), have made the trade anyway. [Batch Zero](https://www.ercot.com/files/docs/2026/06/18/ERCOT-Trending-Topic-New-Batch-Connection-Process-for-Large-Electricity-Users.pdf) gives large loads two optional paths here:
- **Bring your own power:** A campus can count its own power plant toward its size as long as it can cut back within [one minute](https://www.willkie.com/publications/2026/06/ercot-approves-implementing-new-batch-zero-process-for-large-load-interconnections) if the plant fails. So far, 11 Batch Zero projects have picked this path.
- **Agree to cuts:** A campus can draw up to the full amount it asked for, but ERCOT can automatically cut anything above its guaranteed share whenever lines are full.
A campus could also pay its neighbors to cut their load instead. In PJM, [Google is funding Voltus](https://www.utilitydive.com/news/google-virtual-power-plant-vpp-pjm-voltus/821838/) to pool up to 100 MW of batteries, thermostats, and other flexible devices across a territory. But PJM only counts what it trusts the pool to deliver, and it’s a [capacity deal](https://www.latitudemedia.com/news/google-is-voltus-first-bring-your-own-capacity-customer/) that helps the whole grid at its peak, not a fix for any one congested line. That’s a harder sell in Texas, an [energy-only](https://www.ercot.com/files/docs/2025/10/16/ERCOT-Grid-Insights-Energy-Only-Wholesale-Market.pdf) market where transmission is the main bottleneck. Texas has also [barred](https://interchange.puc.texas.gov/Documents/58881_105_1628395.PDF) colocated campuses from getting paid for [similar services](https://interchange.puc.texas.gov/Documents/59220_69_1668205.PDF), since under [SB 6](https://www.utilitydive.com/news/texas-law-gives-grid-operator-power-to-disconnect-data-centers-during-crisi/751587/) they already have to shut off when ERCOT tells them.
Even so, Texas has been a leader in distributed resources. One [Base Power](https://a16z.com/announcement/investing-in-base-power/) fleet, [run with the co-op GVEC](https://www.gvec.org/gvec-and-base-power-partnership/), passed ERCOT’s pilot tests to sell directly into the wholesale market on its first attempt, and is now expanding to 50 MW. These pilots are important; planners need that kind of proof before they’ll design around these fleets. But once trusted, distributed resources can rapidly add capacity without waiting for an expensive new “[peaker](https://en.wikipedia.org/wiki/Peaking_power_plant)” power plant, lines, substations, or a lengthy interconnection process.
Other hardware can help, too. Unlike legacy steel units, a solid-state transformer uses semiconductor switches, so software can measure and steer the power flowing through it. Alongside network upgrades like reconductoring (restringing lines with higher-capacity wire) and dynamic line ratings (rating lines for actual weather instead of worst-case conditions), that greater visibility and control can squeeze more out of wires ratepayers already funded. It’s a big reason we backed [Heron Power](https://a16z.com/announcement/investing-in-heron-power/), which is set to install its solid-state transformers at a [West Texas battery site with RWE](https://www.globenewswire.com/news-release/2026/09/21/3365509/0/en/heron-power-and-rwe-sign-pilot-agreement-outline-path-to-multi-year-capacity-reservation.html).
The problem is that most of these tools help operators keep things running day to day, but planners don’t always count them when they size upgrades. Nothing in physics forces that, though. A September [study by Piq Energy](https://www.piqenergy.com/research/piq-energy-base-power---flexibility-without-curtailment.pdf), using [Base Power](https://a16z.com/announcement/investing-in-base-power/)’s data on potential fleets, found that about 80 MW of home batteries, strategically sited to relieve transmission constraints, could resolve all overloads triggered by a hypothetical new 100 MW data center near Fort Worth.
Batch Zero doesn’t consider things like this yet. It [still plans upgrades for a flexible campus’s full planned load](https://epeconsulting.com/epe-intelligence/news/new-pathway-and-deadlines-in-ercot-s-large-load-batch-zero-provisional-controllable-load-resource-pclr), since that path is a bridge to firm service, and there’s no option to stay flexible for good in exchange for smaller upgrades. Planners could instead size the upgrades smaller by crediting flexibility and other resources that relieve the same bottlenecks, assuming they’re measured in real time and perform reliably.
All of this saves time and money by getting more out of what’s already in the ground, but no amount of flexibility gets the grid out of building more generation and wires for all the demand coming down the pipe.
### Getting to hundreds of gigawatts
On-site power and flexibility will decide how the next few campuses energize, but the labs and hyperscalers I talk to worry most about scale. Their power teams tend to split in two: one picks sites and equipment for the next couple of years, and the other asks how to connect hundreds more gigawatts after 2030.
This is a lot! Run, say, 100 GW all year and it’s 876 TWh, [about a fifth](https://www.eia.gov/energyexplained/electricity/use-of-electricity.php) of what the country used in 2025. Here’s what power developers [told the Energy Information Administration](https://www.eia.gov/todayinenergy/detail.php?id=67205) (EIA) they planned to add in 2026 across the entire grid.
Every source helps, and much of it is headed to Texas anyway. But at last year’s average capacity factors (how much plants actually produce versus their maximum), the planned [solar, wind](https://www.eia.gov/electricity/monthly/epm_table_grapher.php?t=epmt_6_07_b), and [gas plants](https://www.eia.gov/electricity/monthly/epm_table_grapher.php?t=epmt_6_07_a) would make around 150 TWh a year, or about a sixth of that 876 TWh.
To be fair, that gas bar likely understates [what’s being built](https://www.distilled.earth/p/eight-new-gas-plants-will-power-amazons), since [EIA’s survey only counts plants tied to the grid](https://www.eia.gov/survey/form/eia_860/instructions.pdf). In this way, much of the 75 GW of on-site power equipment already on order could be ghost capacity that charts like this one will miss. Still, gas remains popular because it generally runs whenever you need it. And because of that, the constraint is mostly [getting the equipment](https://www.sec.gov/Archives/edgar/data/0001996810/000199681026000147/gevpressrelease2q26.htm) in the first place, so buyers are turning to alternatives that can be [easier to find](https://www.latitudemedia.com/news/is-the-gas-turbine-bottleneck-solving-itself/), like [reciprocating engines](https://www.powermag.com/data-centers-are-turning-to-gas-generators-for-prime-power-to-eliminate-long-lead-times-for-grid-connections/) and [fuel cells](https://investor.bloomenergy.com/press-releases/press-release-details/2026/Bloom-Energy-and-Oracle-Expand-Strategic-Partnership-to-Deploy-up-to-2-8-GW-to-Accelerate-AI-Infrastructure-Build-Out/default.aspx).
That said, gas feels like an incomplete answer to me. I’m no Greenpeace warrior, but running 100 GW around the clock at gas plants’ [average rate](https://www.eia.gov/tools/faqs/faq.php?id=74&t=11) would release nearly 8% of the country’s [energy-related emissions](https://www.eia.gov/environment/emissions/carbon/). Often the easiest equipment to get is even [less efficient](https://www.eia.gov/electricity/annual/html/epa_08_02.html), too. Moreover, operating such a fleet could seriously test our gas supplies.
Solar is compelling because it already has the production scale the labs are aiming for. The problem is that it’s mostly in China. The world [added more than 600 GW](https://www.solarpowereurope.org/press-releases/new-report-global-solar-market-hits-new-record-of-664-gw-installation-in-2025-as-global-solar-fleet-passes-3-tw) in 2025, but China alone makes [more than 80% of the world’s solar components](https://www.iea.org/reports/solar-pv-global-supply-chains/executive-summary) and [battery cells](https://www.iea.org/reports/global-ev-outlook-2026/manufacturing-and-trade). From what I can tell, Chinese suppliers don’t mind selling to us that much, at least partly because they see our scale-up as “cute.” However, China [reportedly weighed](https://finance.yahoo.com/sectors/energy/articles/exclusive-china-weighs-curbs-exports-074947904.html) curbing exports of specific solar manufacturing equipment. Washington has also put on pressure, with [forced-labor shipment holds](https://pv-magazine-usa.com/2022/11/11/due-to-forced-labor-concerns-u-s-ports-have-blocked-over-1000-shipments-of-solar-components/), [new tariffs](https://pv-magazine-usa.com/2026/08/07/us-announces-tariffs-minimum-import-price-on-polysilicon-imports/), and a [phaseout of wind and solar credits](https://www.irs.gov/pub/irs-drop/n-25-42.pdf). Despite this, Elon is aiming [for 200 GW a year](https://pv-magazine-usa.com/2026/01/26/elon-musk-at-wef-spacex-and-tesla-to-produce-100-gw-each-of-pv-per-year-in-the-u-s-this-decade/) of US solar manufacturing on his own, [obviously](https://x.com/elonmusk/status/1925872357781983522?s=20) solar-pilled.
Scaling firm power that isn’t gas is much more complicated. [Uprates](https://www.prnewswire.com/news-releases/georgia-power-agreement-with-google-to-provide-approximately-900-million-in-projected-benefits-for-customers-and-advance-nuclear-energy-in-georgia-302885197.html) and restarts can [squeeze](https://www.reuters.com/business/energy/google-enters-massive-36-gw-power-deal-with-constellation-energy-2026-10-06/) a little more from the existing nuclear fleet, but the real upside is new reactors, as [we’ve argued before](https://a16z.com/how-to-scale-nuclear-power/). [Meta](https://about.fb.com/news/2026/01/meta-nuclear-energy-projects-power-american-ai-leadership/) and [Amazon](https://www.aboutamazon.com/news/sustainability/amazon-smr-nuclear-energy) have signed big deals, but much of the capacity is still options and targets. So far, the military has been a stronger [buyer](https://breakingdefense.com/2026/08/army-taps-5-vendors-to-build-nuclear-microreactors-with-2-2b-pot/) to build microreactors on its bases, which is how factory-built reactors can learn to get faster and cheaper (blame EPC as much as the NRC). Geothermal could also leverage drilling (something Texas knows well) for repeatable power — [Google](https://fervoenergy.com/fervo-energy-and-google-sign-396-mw-ppa/) and [Meta](https://www.businesswire.com/news/home/20240826755596/en/Sage-Geosystems-and-Meta-Announce-Agreement-for-Next-Generation-Geothermal-Power-Generation) appear quite interested.
Regardless of the power source, it all still ends up waiting on other equipment like transformers and switchgear, and all the crews to install them. Large power transformers now take [more than two years](https://www.powermag.com/transformers-in-2026-shortage-scramble-or-self-inflicted-crisis/) to arrive, and the FCC has [limited new foreign-made inverters](https://docs.fcc.gov/public/attachments/DA-26-786A1.pdf) alongside an [August emergency order](https://www.whitehouse.gov/presidential-actions/2026/08/declaring-a-national-emergency-to-secure-the-united-states-bulk-power-system/) that could further bar Chinese-made equipment from the grid.
Someone also has to build the wires. In 2008, Texas regulators ordered the [CREZ lines](https://www.bakerinstitute.org/sites/default/files/2020-11/import/ces-pub-texascrez-111720.pdf) to carry West Texas wind, then spread their $7 billion cost across every ratepayer. Now they’re approving even bigger [765 kV lines](https://www.texastribune.org/2026/08/28/transmission-lines-texas-energy/), but it’s going [slower](https://thetexan.news/issues/energy/here-s-the-current-state-of-the-plan-to-build-765-kv-transmission-lines-across/article_b7e8c5bb-35c7-42d7-bed9-04f243de1927.html) than many would like. Some of that is just (unfortunately) typical construction, which is slow and expensive anywhere, but there’s also a myriad of additional regulatory hurdles on top. The [federal permitting deal taking shape in Washington](https://www.energy.senate.gov/press/rep/lee-capito-heinrich-and-whitehouse-introduce-bipartisan-american-affordability-and-jobs-act-of-2026/) could help move things along if Congress can pass it.
Admittedly, I’m more confident that we’ll need a lot of power than I am about the exact shape it takes. My bet is that a handful of setups, depending on geography and flexibility, get built over and over. Maybe on-site gas and batteries carry a campus until its grid connection shows up, then stick around as backup to flex when the grid is tight. Solar gets layered on now where it fits, and geothermal and reactors come in once they prove out.
Past 2030, it’s even harder to say who ends up building and owning all that power. One answer is that the same company builds both the plant and the campus, which Google’s [purchase of Intersect](https://abc.xyz/investor/news/news-details/2025/Alphabet-Announces-Agreement-to-Acquire-Intersect-to-Advance-U-S--Energy-Innovation-2025-DVIuVDM9wW/default.aspx) may signal. Another is that oil and gas companies, like Chevron or Williams, become [broader grid builders](https://travisfisher.substack.com/p/socrates-south-shows-the-benefits), and it isn’t hard to picture them, or “neo-utilities” like NRG and NextEra, building private grids that serve several campuses, assuming the law enables it.
I don’t know which way it goes yet, but they’ll likely all be buying from the same equipment makers. That market is huge and surprisingly ill-equipped to meet inflecting demand. Given that, I see two major ways for startups to break in:
- **Integration:** Some take familiar hardware and win on integration or business model, the way Base Power runs home batteries as a trusted aggregated resource.
- **Technology:** Others bring new, superior technology that early adopters will take a chance on in a constrained market, which is what Heron Power is aiming to do with solid-state transformers.
A third pitch, crudely put as “worse but faster,” sells well in a shortage and can be immensely profitable, but I’d ask what those profits are being reinvested in, because it might get hard to compete if broader supply catches up.
Indeed, shortages like these are an opening for startups, but it helps to understand why incumbents aren’t quick to fill them. The last bet on a turbine boom ended in a $22 billion GE Power write-down and [helped cost GE’s CEO his job](https://www.bostonglobe.com/business/2018/10/01/john-flannery-out-general-electric-ceo/7QRPwEHxACP1PON92Vq8WJ/story.html). A startup has to survive the busts incumbents are planning around, as well as compete globally with Siemens Energy, Mitsubishi Power, and all sorts of suppliers in places like India and China. Customers may pay for speed today, but keeping them will take reliable, competitively priced equipment and a service team that knows what it’s doing.
Every step of site development is hard. Vendors like GE Vernova now take [nonrefundable deposits](https://www.gevernova.com/sites/default/files/gev_webcast_transcript_10222025.pdf) just to reserve a manufacturing slot, leaving developers with a [chicken-and-egg problem](https://newsletter.semianalysis.com/p/what-is-so-hard-about-behind-the). Lenders want a long-term contract with a solid customer, who wants a credible timeline, and that timeline takes deposits the developer usually needs lenders to fund. Bring in an unproven vendor and the loop gets even harder to close (and markets notice fast when it [breaks](https://amarillotribune.org/2025/12/12/fermi-stock-drops-after-termination-of-150-million-construction-agreement/)). Once it closes, everything else still [has to go right](https://www.bloomberg.com/news/articles/2026-09-24/oracle-cites-force-majeure-to-shield-itself-on-controversial-data-center).
For a new vendor, even getting into a campus’s first phase is difficult. As far as I know, hardly anyone has placed a large, firm order for data center power equipment without thousands of hours of operation. For example, [FTAI’s big order](https://www.globenewswire.com/news-release/2026/07/22/3331360/35538/en/ftai-announces-1-465-billion-gas-turbine-generator-set-order-through-j-f-power-systems.html) rests on the CFM56, a jet engine that has logged [more than a billion flight hours](https://www.safran-group.com/pressroom/cfm56-engine-fleet-surpasses-one-billion-engine-flight-hours-2019-06-04), while Crusoe [stepped back from Boom](https://techcrunch.com/2026/09/25/crusoe-abandons-1-25b-plan-to-use-boom-turbines-at-ai-data-centers/) as Boom’s first engine core was still [gearing up for tests](https://aviationweek.com/business-aviation/aircraft-propulsion/boom-supersonic-gears-engine-prototype-core-testing). So even credible teams building awesome technology face a cold-start problem. Thus, getting your product designed into a real project matters far more than early deposits or revenue.
### After the pause
Let’s get back to Texas. I think it’s fair to check that projects in Batch Zero are [what their developers swore](https://www.ercot.com/files/docs/2026/08/20/Batch-Zero-Verification-and-Audit-PUCT-Presentation.pdf) they are, and credible projects should move ahead soon. The broad permit freeze is harder to defend, though I see the state’s logic if the goal is to approve nothing until the audit is done. That said, I don’t like that it makes a developer that has funded its first phase, even one bringing its own power, wait like a speculator. Texas should narrow it now by exempting generators permitted only for emergencies and projects that have funded their first phase. The October 19 update to the governor from the state’s environmental regulator is the obvious place to start this conversation.
Once the audit is done, the ongoing tests can be more straightforward. A developer that posts its security deposit, pays for the capacity it reserves, and hits phased milestones should get a connection date it can plan for. It should also connect sooner if it agrees to cut back when the grid is tight (Batch Zero [already does some of this](https://epeconsulting.com/epe-intelligence/news/new-pathway-and-deadlines-in-ercot-s-large-load-batch-zero-provisional-controllable-load-resource-pclr)). Going further, the utility building upgrades should probably answer for delays much like a developer does. Perhaps Texas [could](https://capitol.texas.gov/tlodocs/86R/analysis/pdf/SB01938F.pdf) [open lines to competitive bids](https://www.texasenergyandpower.com/p/texas-competes-on-everything-except) with cost caps and penalties.
Whatever Texas decides, power will likely stay tight for years while the bottleneck keeps moving. Importantly, data centers are not that unique in the equipment they need; they just hit these limits first. In this sense, they’re the perfect rehearsal, because much of what we want to (re)build in this country will run on the same stuff.
That also makes it an opportunity I’d hate to waste. AI companies will pay nearly any price for power and can build almost anywhere, so they can help fund upgrades the grid needs anyway. Analyst Hans Royal estimates inference could pay an absurd [$5,600 per MWh](https://computeheatrate.com/chr-index) for power and still earn a decent return, nearly 60 times what [the average US industrial customer pays](https://www.eia.gov/electricity/monthly/update/end-use.php). That can strain local prices for gas, power, labor, and materials, but the infrastructure it funds can be worth far more. Admittedly, that argument can be a hard sell right now. In an August poll, [56% of Texas voters](https://texaspolitics.utexas.edu/blog/new-ut-texas-politics-project-poll-finds-talarico-leading-paxton-abbott-leading-hinojosa-continued-resistance-to-data-centers-2) said more data centers would hurt local energy bills.
The worry is fair since the honest answer is that it depends on who pays. In a perfect world, developers cover all related upgrades and pay for the capacity they reserve, supporting shared infrastructure construction that improves the grid for all of us. Even just by buying a lot of power, a campus spreads a service territory’s fixed costs over more sales, shrinking everyone else’s share. [Berkeley Lab found](https://eta-publications.lbl.gov/sites/default/files/2026-03/retail_price_trends_2026_edition.pdf) that from 2019 to 2025, the states with the most load growth generally saw average prices fall after inflation. Imagine all the money racing into AI helping the next factory connect and the next household electrify without an unaffordable bill (or a grid too tight for an EV, or a home robot). That household may never open a chatbot and still come away far better off. We’re already seeing early versions of this:
- **Indiana:** The utility serving Amazon’s New Carlisle campus [proposed using data center revenue to cut household bills by about $100 a year](https://www.wvpe.org/wvpe-news/2026-08-26/i-m-announces-electric-rate-cut-plan-enabled-by-data-centers).
- **Alabama:** A Tuscaloosa County campus will pay [$270 million in community benefits](https://yellowhammernews.com/tuscaloosa-county-lands-3-2-billion-data-center-first-in-alabama-to-pay-community-benefits-270-million-over-20-years/) over 20 years. In return, it gets about $314.5 million in tax breaks, though school taxes (about $131.5 million) aren’t abated.
- **Pennsylvania:** One developer [offered every household in Hazle Township $10,000](https://www.yahoo.com/news/us/articles/data-center-developer-offers-10-114000653.html) if its campus gets approved, which some residents (perhaps correctly) called a bribe.
I would not be surprised to see hyperscalers covering a whole town’s power bills as part of hosting a local campus. Most states, [Texas included](https://texas.public.law/statutes/tex._util._code_section_36.003), don’t let a utility single out one town, but at a small utility where data centers use most of the power, like [Oregon’s Umatilla Electric](https://www.cobank.com/knowledge-exchange/earning-the-business-by-demonstrating-network-value), covering a [typical household’s](https://www.eia.gov/electricity/sales_revenue_price/pdf/table_5A.pdf) bill for each of its [roughly 17,000 meters](https://www.umatillaelectric.com/about-us/who-we-are/) would only cost about $30 million a year. Such an agreement could pencil because that’s less than 0.5% of what a 1 GW campus [costs to own and run](https://epoch.ai/data-insights/ai-datacenter-cost-breakdown).
It’s also important to read the fine print when deals are made. In Arkansas, Google agreed to pay [$443 million](https://www.arkansasonline.com/news/2026/sep/03/googles-contribution-was-at-issue-in-arkansas/) up front toward an Entergy solar plant for its data center. The catch is that Entergy counted it as [prepayment for power](https://www.arkansasonline.com/news/2026/sep/03/googles-contribution-was-at-issue-in-arkansas/), so it can still seek the plant’s full cost plus a return from all its customers. And Entergy [sued two newspapers](https://www.cnbc.com/2026/09/02/google-data-center-arkansas-entergy-court.html) to stop them from reporting on the contract, then dropped the suit after a judge refused. Even if the deal is legal, this isn’t how you build trust with the community.
A campus that levels with its neighbors and actually follows through keeps people on its side, and the next one gets easier to welcome. Get that wrong and everyone else can end up paying for it. In PJM, [the market monitor says data centers account for 38% of the latest capacity bill](https://www.utilitydive.com/news/pjm-data-centers-capacity-auction-imm-bowring/825626/), some $6.3 billion. Let me be clear: none of that justifies a blanket pause. But it does mean getting the rules and incentives right as we scale up development. AI companies have [pledged to pay their way](https://www.powermag.com/hyperscalers-sign-white-house-pledge-to-fund-data-center-power-grid-upgrades/), and most would rather connect to the grid than avoid it. It seems ideal for everyone that we let them do so instead of driving that spending into private islands, or pushing the infrastructure upgrades that we’ll need anyway onto everyone else’s bills.
Texas created the energy fast lane first; now it has the opportunity to show us a better one, and I expect it will. Power is where America finds out whether it can still build.
If you’re working on energy problems, I’d love to learn from you. Please reach out.

[Subscribe now](https://www.a16z.news/subscribe?)

_This newsletter is provided for informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. Furthermore, this content is not investment advice, nor is it intended for use by any investors or prospective investors in any a16z funds. This newsletter may link to other websites or contain other information obtained from third-party sources - a16z has not independently verified nor makes any representations about the current or enduring accuracy of such information. If this content includes third-party advertisements, a16z has not reviewed such advertisements and does not endorse any advertising content or related companies contained therein. Any investments or portfolio companies mentioned, referred to, or described are not representative of all investments in vehicles managed by a16z; visit [https://a16z.com/investment-list/](https://a16z.com/investment-list/) for a full list of investments. Other important information can be found at [a16z.com/disclosures](http://a16z.com/disclosures). You’re receiving this newsletter since you opted in earlier; if you would like to opt out of future newsletters you may unsubscribe immediately._