· 供应链分析

长文拆解 SPEED Act 对 Neocloud 和 AI 数据中心的去风险作用

涉及标的:

中文翻译

刚刚,SPEED Act 在众议院取得进展。 这是今年 Neocloud 板块($NBIS、$CRWV、$IREN)最大的单一去风险法案/事件。 下面是原因和梳理: 美国政府正准备以美国对中国的国家安全为理由,支持从 Oracle 到 Nebius 的 AI 数据中心建设。 Oracle 和 CoreWeave 最近跌了 40%+(也把 $NBIS 从 140 美元打到 79 美元、$IREN 从 80 美元打到 35 美元、$CIFR 从 24 美元打到 14 美元),核心恐惧有三点: 1. 数据中心延迟和递延收入 2. 不可持续的 CapEx -> 没有 FCF 3. OpenAI 传染/积压订单风险 SPEED Act 和美国政府干预,修复了数据中心建设延迟的空头论点,也解决了利用率滞后带来的盈利问题(利润率)。 #1 数据中心延迟和递延收入 空头 thesis:多年许可延迟(NEPA、输电)把高价值合同变成了递延收入风险。$CRWV 明确把供应商延迟作为下调指引的原因,并在财报后大跌,因为相当一部分收入被推迟到 2026 年 Q1/Q2。 如果 SPEED Act 通过,alpha 在这里: 强制速度和诉讼保护。 - SPEED Act 要求联邦环境和监管审查遵守严格且不可协商的截止日期,通常是 1-2 年。 - 诉讼保护:该法案大幅缩短针对已批准许可提起诉讼的时效,比如缩到 150 天,并指示法院即使许可被临时挑战,也要允许数据中心建设继续推进。 结果:从签约到“GPU 上架 -> 收入流入”的时间线,被压缩了,并由联邦政府在政治上去风险。 递延收入被提前确认,修复了 $CRWV、$APLD 和 Neocloud 板块此前面临的延迟、递延利润/收入问题。 #2 不可持续 CapEx -> 无法从资产变现出 FCF 空头 thesis:公司在 GPU 和建设上花费数十亿美元($ORCL 的 capex 很巨大),但从购买 GPU 到变现之间的利用率拖累严重影响盈利能力和 FCF。 这也大幅影响 AI Cloud 供应商,因为它们缺乏足够电力来把 GPU/capex 变现。 公司因此面临巨大的减记风险,也就是利用率拖累:GPU 闲置时,折旧和通电成本的时钟仍在跑。 这个拖累对数据中心部门盈利能力影响巨大,The Information 关于 $ORCL AI 利润率极薄的报道也提到过这一点。 SPEED Act 和美国政府干预会直接降低 CapEx 风险,因为速度上的立法要求(修复点 #1)实际上保证电力基础设施会在一个确定且较短的时间线内到位。 这种确定性让 $NBIS、$CRWV 和 $IREN 可以更有信心安排数十亿美元 GPU 的采购和部署,知道资产到货后就能立即开始变现,同时也通过降低利用率拖累来加快 FCF。 这种结构性变化会流向整个行业。它会立即降低主要 AI Cloud 供应商($AMZN、$MSFT、$ORCL)的风险,因为它们现在能更确定地保证产能;同时也会保证 Colo/Infra/Energy 提供商($CIFR、$WULF 等)的需求,因为它们的核心业务就是提供电力容量。 关于 capex -> FCF 以及建设延迟时间线的空头论点,现在已经被 SPEED Act 直接处理。 现在美国政府准备加速 $NBIS、$CRWV 和 $IREN 这类公司,因为 AI 数据中心已经被放到美国和中国 AI 国家安全竞争的前线。 它能否在众议院通过,是每个投资者都该关注的事。但如果通过,这会是 Neocloud / AI 数据中心建设里最大的、尚未被充分讨论的顺风之一。

英文原文

Just now, the SPEED Act ADVANCES in the House. This is the single biggest de-risking bill/event for the Neocloud sector ( $NBIS, $CRWV, $IREN) this year. Here's why and a rundown: The U.S. GOVERNMENT is set to support the AI data center buildout from Oracle down to Nebius on national security grounds for US vs China. Oracle and CoreWeave recently dropped 40%+ (tanking $NBIS $140 -> $79, $IREN $80 -> 35, $CIFR, $24 -> $14 as well) on three core fears: 1. DC Delays & Deferred Revenue 2. Unsustainable CapEx → No FCF 3. OpenAI Contagion/Backlog. The Speed Act and US Government intervention fixes bear-case points for data center buildout delays and addresses utilization lag profitability issues (margins). #1 DC Delays & Deferred Revenue Bear Thesis: Multi-year permitting delays (NEPA, transmission) turned high-value contracts into deferred revenue risk. $CRWV explicitly cited vendor delays for lowering guidance and tanked on earnings shifting a large portion of revenue to from Q1 Q2 2026. The alpha if the Speed Act passes: Mandatory Speed and Litigation Shields. - The Speed Act mandates strict, non-negotiable deadlines (often 1-2 years) for federal environmental and regulatory reviews. - The Litigation Shield: The bill drastically shortens the statute of limitations for filing lawsuits against approved permits (e.g., to 150 days) and instructs courts to allow DC buildout to continue even if a permit is temporarily challenged). The Result: The timeline from contract signing to "GPUs on racks -> revenue flowing" is now compressed and politically de-risked by the Federal Government. Deferred revenue is pulled forward and fixes delays and deferred profitability/revenue that plagued $CRWV, $APLD, and the Neocloud sector. #2: Unsustainable CapEx -> No FCF from monetizing the assets Bear Thesis: Companies were spending billions on GPUs and construction ( $ORCL's capex is massive) with utilization drag (from the point of purchasing the GPUs to monetization) largely affecting profitability and FCF. This also largely affects AI Cloud vendors (lacking power to turn monetize the GPUs/capex). Again This forced companies to take a massive write-down risk due to Utilization Drag (the time the GPU sits idle while the clock runs on depreciation/power-up). This drag is HUGE for profitability on DC segments, as cited in The Information reports on $ORCL's razor-thin AI margins. The SPEED Act and the US Government intervention directly de-risks CapEx as the legislative mandate for speed (Fix #1) effectively guarantees that power infrastructure will arrive within a defined, short timeline. This certainty allows $NBIS, $CRWV, and $IREN to time the purchase and deployment of billions in GPUs with high confidence that the assets will begin monetizing immediately upon arrival as well as accelerates FCF from reducing utilization drags. This structural change flows down the entire industry. It instantly de-risks the major AI Cloud vendors ($AMZN, $MSFT, $ORCL) who can now guarantee their capacity, and it guarantees demand for the Colo/Infra/Energy providers ( $CIFR, $WULF, and others) whose core business is supplying that power capacity. The bear case on capex -> FCF + buildout delay timeline has now directly addressed with the Speed ACT. Now the US government is set accelerate companies like $NBIS, $CRWV, and $IREN as AI Datacenters is now placed on the forefront of the AI national security battle between the United States and China. Whether it passes legislation in the House is what every investor should be watching, but if does, this is one of the largest (not-talked about) tailwinds for the Neocloud /AI Decenter buildout.

在 X 上查看原推 ↗