NuScale's AI Hype Cycle: The SMR Race Isn't About Design Speed—It's About Who Survives the First Build
The press release landed at 9:47 AM EST. NuScale Power, the first company to earn U.S. Nuclear Regulatory Commission design approval for a small modular reactor, announced it was deploying AI to accelerate SMR design. The stock popped 12% within the hour. Crypto Briefing ran the story. The narrative was clean: AI plus nuclear equals the future of clean baseload power. But here's what the headline didn't say—NuScale has zero operating reactors, zero new commercial orders since its Idaho project collapsed in 2023, and its AI announcement is a marketing pivot, not a technical breakthrough. Volume spikes lie; liquidity flows tell the truth. The same applies to press releases.
Context: Why this matters now. The 2023-2026 window has seen Microsoft, Google, Amazon, and OpenAI pour billions into nuclear—especially SMRs—to power AI data centers. Google signed a power purchase agreement with Kairos Power. Amazon invested in X-energy. Microsoft is evaluating SMRs for its data centers. NuScale, riding this wave, has seen its market cap swing from $500 million in 2024 to over $10 billion in 2025 on AI-related announcements. But the company's core problem hasn't changed: it has no paying customer for a first-of-a-kind reactor. The CFPP project in Idaho was canceled in November 2023 after costs ballooned from $57/MWh to $89/MWh, and the utility walked away. Since then, NuScale has signed MOUs with Romania, Poland, and Kazakhstan—but none have reached final investment decision. The AI narrative is a lifeline, not a strategy.
Core: The technical reality behind the AI claim. AI can accelerate design optimization, simulation, and documentation—but it cannot solve the fundamental bottlenecks of SMR commercialization. Let's break down what AI actually does in nuclear design. First, parameter optimization: AI can explore fuel arrangement and thermal-hydraulic configurations faster than manual iteration. This is real, but it's been used in traditional nuclear design for decades. Second, simulation acceleration: AI surrogate models can compress weeks of physics simulations into hours. But these models require validation against physical experiments, and the NRC won't relax safety standards just because AI was involved. Third, safety analysis: AI can generate probabilistic risk assessment trees, but human experts must review every branch. Fourth, document generation: LLMs can draft licensing documents, but the NRC still requires traceable physical verification. The net effect? AI might reduce engineering time by 20-40%—useful, but not revolutionary. The real bottleneck is the supply chain. HALEU fuel—enriched to 5-20% U-235—is in critical shortage. The U.S. has only one domestic producer, Centrus Energy, which started operations in 2025. Russia remains a major HALEU supplier, creating geopolitical risk. Reactor pressure vessels and steam generators are manufactured by a handful of companies in the U.S., Japan, South Korea, China, France, and Russia. Building a dedicated SMR production line requires billions in upfront capital—a risk no one will take without a confirmed order. AI cannot manufacture a pressure vessel. AI cannot shorten the NRC's six-year licensing process. AI cannot convince a utility to pay a 50% premium for unproven technology. The chart doesn't lie: NuScale's revenue in 2025 was a few million dollars from DOE contracts, while R&D and administrative costs exceeded $200 million. The company burns through cash at a rate that gives it 1.5-2 years of runway. If no new orders materialize by 2027, NuScale will need to dilute shareholders again or face a liquidity crisis.
Contrarian: The real race is between China, Russia, and the U.S.—and the U.S. is losing. While NuScale talks AI, China's ACP100—the world's first commercial land-based SMR—started construction in 2021 and is scheduled for grid connection in 2026. Russia's KLT-40S floating reactor has been operating since 2019. The U.S. has no SMR under construction. The policy support is there: the Inflation Reduction Act includes nuclear in clean energy tax credits, and the DOE's ARDP program has allocated over $3 billion to advanced reactor developers. But the regulatory pathway remains a nightmare. The NRC took six years to approve NuScale's design—and that was just the design certification. Construction permits, operating licenses, and emergency planning zones add years more. Meanwhile, China's regulatory process is streamlined, and its supply chain is vertically integrated. The U.S. is betting on a market-driven approach, but the market won't move without a successful first build. And here's the uncomfortable truth: NuScale's AI announcement is a distraction from the fact that it has no first build. The company's SPAC listing in 2022 raised $380 million, but the subsequent stock collapse and management churn have eroded institutional trust. The AI narrative is designed to attract retail and quant funds—which now dominate its shareholder base—not to solve engineering problems. Speed is safety when the exploit is already live, but in this case, the exploit is a narrative, not a vulnerability. We don't need AI to design a reactor; we need a customer willing to pay for the first one.
Takeaway: Watch the order book, not the press releases. The next 12 months will determine whether NuScale survives as a going concern. If it announces a new commercial order with a credible utility or tech giant, the AI story becomes a footnote. If it announces another MOU or a partnership with an AI company, the story is just marketing. The real signal will come from the NRC's docket—any new construction permit applications—and from the DOE's cost-share decisions. The SMR race is not about who has the best AI; it's about who can build a reactor without going bankrupt. China and Russia are already there. The U.S. is still waiting for its first concrete pour. The question isn't whether AI accelerates design—it's whether NuScale can accelerate a customer.