来自 Maksym Misichenko · Nasdaq ·
AI智能体对这条新闻的看法
The panel consensus is bearish on Oklo and NuScale's near-term prospects due to regulatory, financing, and fuel supply hurdles. While AI-driven electricity demand could be a catalyst, the timeline for SMR deployment is uncertain and risky.
风险: HALEU fuel supply bottleneck
机会: Government support for SMRs as a national security asset
本分析由 StockScreener 管道生成——四个领先的 LLM（Claude、GPT、Gemini、Grok）接收相同的提示，并内置反幻觉防护。 阅读方法论 →
人工智能革命可以使核能受益。
两家创新型核能公司有望受益。
Oklo 和 NuScale 都专注于小型模块化反应堆 (SMR)。
核能正经历兴趣复苏，而要归功于一个催化剂：人工智能。
从 2005 年到 2023 年，美国的电力生产基本持平。然而，2024 年，电力生产飙升至历史新高。2025 年，又创下历史新高。原因是什么？因为人工智能需要大量的能源，而且该行业发展如此迅速，以至于数十年来，增长率首次大幅转为正值。
人工智能会创造世界上第一个万亿美元富豪吗？ 我们的团队刚刚发布了一份报告，内容是关于一家鲜为人知但至关重要的公司，被称为“不可或缺的垄断”，它为英伟达 (Nvidia) 和英特尔 (Intel) 都需要的关键技术提供支持。继续 »
目前电网的设计无法满足快速增长的人工智能行业的需求，而人工智能行业又依赖于数据中心来运行。这种基础设施需要大量的电力，不仅要保持灯光通明，还要冷却执行人工智能软件计算的炙热的图形处理单元 (GPU)。
简而言之，为了支持人工智能的持续增长，需要更多的电力生产——这种增长之旅预计将持续数十年。
核能已成为一个有希望的解决方案——但不是任何一种核能。以下两家公司正在开创性地利用核裂变生产能源的一种相对新颖的方法。这种创新方法可能是人工智能和数据中心不断增长的能源需求的理想解决方案。
很少有核能股票像 Oklo (纽约证券交易所: OKLO) 这样，如此完美地满足人工智能行业的需求。事实上，OpenAI 和 ChatGPT 的创始人山姆·阿尔特曼 (Sam Altman) 多年前就成为了 Oklo 的早期投资者，并担任了董事会主席多年。
从我的角度来看，如果数据中心和云基础设施运营商决定自己解决能源难题，那么 Oklo 是理想的选择。因为 Oklo 专注于小型模块化反应堆，或 SMR。即使按照 SMR 的标准来看，Oklo 的系统也相对较小，小到 Oklo 的管理团队称它们为“微反应堆”。
这使得它们成为数据中心运营商在更本地化的范围内采用的理想选择，特别是由于 Oklo 认为一旦流程完全成熟，整个许可和部署时间表将仅为六到十二个月。
Oklo 比许多 SMR 竞争对手更具多样性，这得益于其收购 Atomic Alchemy，这使它获得了燃料回收能力以及对医疗级同位素销售的敞口。但其主要业务将是向数据中心供应核“微反应堆”，正如其与 Meta Platforms 在俄亥俄州南部建立 1.2 吉瓦系统所证明的那样。
NuScale Power (纽约证券交易所: SMR) 也设计和销售 SMR 系统，但其市场策略与 Oklo 非常不同。与直接与数据中心运营商达成协议不同，NuScale 专注于通过大规模电网部署来为电网供电。
虽然 NuScale 在全球范围内有几个项目正在进行中，但其最具价值的项目可能是其与田纳西河谷管理局 (Tennessee Valley Authority) 为美国东部提供 6 吉瓦系统的协议。由于预计今年年底将达成购电协议，该项目可能会迎来一个重大的催化剂。
NuScale 的系统通常比 Oklo 的系统更大，其提前时间也相应更长。但是，大规模 SMR 可能具有光明的前景，因为公用事业公司也希望尽快将低碳、可靠的基荷电力投入使用，以避免停电和繁重的电价上涨。
如果您不确定哪支 SMR 股票具有更优越的市场策略，那么购买 Oklo 和 NuScale 的股份可以使您获得更多样化的敞口。值得注意的是，从长期来看，其他多元化的工业集团也在探索 SMR，因此竞争格局尚未完全形成。
在您购买 NuScale Power 的股票之前，请考虑以下几点：
Motley Fool Stock Advisor 分析师团队刚刚确定他们认为投资者现在应该购买的 10 支最佳股票……而 NuScale Power 并不是其中之一。这些股票可能会在未来几年产生巨大的回报。
考虑一下 Netflix 在 2004 年 12 月 17 日被列入此名单时……如果您当时投资了 1,000 美元，您将拥有 463,900 美元！ 或当 Nvidia 在 2005 年 4 月 15 日被列入此名单时……如果您当时投资了 1,000 美元，您将拥有 1,294,401 美元！
值得注意的是，Stock Advisor 的总平均回报率为 978%——与标准普尔 500 指数相比，这是一个市场表现优于 211% 的出色的表现。不要错过最新的前 10 名名单，该名单可使用 Stock Advisor，并加入由个人投资者为个人投资者建立的投资社区。
**股票顾问的回报截至 2026 年 5 月 31 日。 *
Ryan Vanzo 对所提及的任何股票都没有持有任何头寸。The Motley Fool 持有并推荐 Meta Platforms。The Motley Fool 推荐 NuScale Power。The Motley Fool 有一份披露政策。
本文中的观点和意见是作者的观点和意见，不一定代表纳斯达克公司的观点。
四大领先AI模型讨论这篇文章
"Regulatory and deployment timelines for these SMR developers will likely stretch well beyond the 2026-2027 AI power crunch the article assumes."
The article correctly flags AI-driven power demand as a tailwind for nuclear, but it underplays execution risk for both Oklo and NuScale. OKLO's microreactor timeline of 6-12 months post-scaling assumes rapid NRC approvals that have historically taken far longer, while its Meta deal remains non-binding on deployment. NuScale's TVA project faces similar permitting and supply-chain hurdles, plus competition from GE Hitachi and Rolls-Royce SMRs that already have deeper utility relationships. Both stocks trade at premiums detached from near-term revenue, leaving them vulnerable to any delay or cost overrun.
If hyperscalers accelerate direct SMR offtake and the NRC fast-tracks designs as promised in recent policy signals, both companies could secure multi-gigawatt backlogs within 24 months and justify current valuations.
"The article treats SMR deployment timelines and cost competitiveness as solved problems when both remain unvalidated at commercial scale, and conflates AI's real energy demand with SMRs' ability to capture it profitably."
The article conflates two separate theses: AI driving electricity demand (credible) and SMRs solving it (speculative). U.S. electricity generation rose in 2024-25, yes—but the article omits that natural gas and renewables captured most new capacity, not nuclear. Oklo's 6-12 month licensing claim is unproven; NuScale's TVA deal has slipped repeatedly. Neither company is profitable or has deployed a commercial unit at scale. The Meta deal (1.2 GW) is a letter of intent, not a binding contract. SMRs face per-megawatt cost disadvantages vs. utility-scale reactors and renewables+storage. Timing risk is acute: if data centers solve cooling via alternative tech or demand moderates, the entire thesis collapses.
If Oklo and NuScale execute on timelines and secure 3-5 major data center contracts each, they could become essential infrastructure plays worth 10x+ current valuations—and the article's omission of execution risk doesn't mean it won't happen.
"The market is significantly underestimating the regulatory and capital-expenditure risks that typically plague nuclear infrastructure projects, regardless of AI-driven demand."
The narrative linking AI energy demand to SMRs is structurally sound, but the article ignores the 'valley of death' facing pre-revenue nuclear startups. Oklo and NuScale are currently speculative R&D plays, not energy utilities. While the Meta and TVA deals provide headlines, the regulatory hurdles with the NRC (Nuclear Regulatory Commission) are historically brutal, often leading to multi-year delays and massive cost overruns. Investors are pricing in a 'when' rather than an 'if,' ignoring the reality that these firms lack the balance sheets to survive a 3-5 year deployment delay. I am cautious; the valuation premium here assumes perfection in a sector where perfection is historically impossible.
If AI energy demand grows exponentially as projected, the scarcity of carbon-free baseload power will force federal regulators to fast-track SMR licensing, potentially creating a 'space race' environment that renders current cost-overrun concerns obsolete.
"Regulatory, financing, and scale challenges could cap SMR upside even if AI-driven demand strengthens."
AI-driven electricity demand could be a near-term catalyst for SMR players like Oklo and NuScale, but the deployment math is the risk. SMRs need regulatory approvals, capex, and siting certainty scalable to data centers or utilities; timelines can miss. Oklo’s claim of six-to-twelve months licensing and a 1.2 GW data-center deal with Meta strike me as optimistic given NRC history and financing hurdles. NuScale’s TVA 6 GW project reads more credible on a pathway to grid-scale, but multi-GW deployments demand vast capital, modular buildout, PPA certainty, and public acceptance. The AI demand thesis matters, but load growth could soften if efficiency gains or policy shifts reduce power intensity.
The AI-driven demand thesis may be overstated; efficiency gains, cooling tech, and demand-side management could cap growth, making SMR deployment timelines and capex hurdles even tougher than assumed.
"HALEU uranium supply constraints represent a critical unmentioned bottleneck for SMR commercialization."
Claude notes gas and renewables captured new capacity, yet overlooks interconnection queues surpassing 2,000 GW where nuclear's reliability offers an edge amid transmission constraints. An overlooked risk is HALEU fuel supply: US production is minimal, far below what's required for multiple SMR deployments. This bottleneck could extend delays beyond regulatory or financing issues that Grok and Gemini highlighted.
"HALEU fuel supply is a harder constraint than NRC licensing and deserves equal weight in valuation risk."
Grok's HALEU bottleneck is the sharpest execution risk nobody quantified. US production is ~20 MT/year; a single 300 MW SMR needs ~15 MT annually. Scaling to Oklo's implied 5-10 unit pipeline within 36 months requires either massive upstream capex or foreign dependency—both regulatory minefields. This isn't regulatory delay; it's a physical constraint that makes current timelines mathematically impossible without prior fuel stockpiling we haven't seen announced.
"Geopolitical competition will force the U.S. government to subsidize and accelerate the HALEU supply chain, overriding current market-based deployment hurdles."
Claude and Grok are right about the HALEU bottleneck, but you are all missing the geopolitical dimension. The U.S. government views SMRs as a national security asset to counter Russia and China’s nuclear export dominance. This won't be a pure market play; expect massive DOE subsidies and 'Project Apollo' style regulatory waivers. The physical fuel constraint isn't a dealbreaker; it’s the catalyst for the federal government to socialize the supply chain risk and fast-track fuel production.
"HALEU bottleneck is a real risk, but not insurmountable; timing hinges on accelerated fuel supply solutions and policy actions."
Claude’s emphasis on HALEU bottlenecks is a critical risk, but declaring it mathematically impossible over 24–36 months overstates the certainty. There are potential mitigants—foreign fuel contracts, enrichment partnerships, or accelerated DOE support—that could unlock supply. If fuel access remains tight, Oklo/NuScale face multi-quarter delays that crush deployment-driven upside; if policy and sourcing accelerates, the thesis could re-rate sharply. The discussion should quantify supply-path timelines, not just the bottleneck.
The panel consensus is bearish on Oklo and NuScale's near-term prospects due to regulatory, financing, and fuel supply hurdles. While AI-driven electricity demand could be a catalyst, the timeline for SMR deployment is uncertain and risky.
Government support for SMRs as a national security asset
HALEU fuel supply bottleneck