Why Google Chose a Reactor That's Never Been Built

Opening

On March 4, 2026, the U.S. Nuclear Regulatory Commission (NRC) granted a construction permit for a reactor to be built in Kemmerer, Wyoming. It’s the sodium-cooled fast reactor¹ from TerraPower, the company founded by Bill Gates. It was the first commercial non-light-water reactor approval in roughly 40 years, and a few weeks later, ground was actually broken.

At that same moment, the number of large nuclear reactors under construction in the United States was zero. This despite the fact that the design is finished, a licensing pathway exists, and Westinghouse’s AP1000 has already been built to completion twice.

Reader, here’s where things get strange. The companies most desperate for power—Google, Amazon, Microsoft, and the other hyperscalers—are choosing a reactor design that has never been built before, over a proven, large-scale plant. Let me give you the conclusion up front: this isn’t because they love risk. It’s because in the West, “proven technology” no longer means “technology you can build quickly.”

---

Right now, there isn’t a single large-scale nuclear construction site anywhere in the US

What created this situation was Vogtle Units 3 and 4 in Georgia. Final construction cost came to roughly $36 billion — more than double the original estimate — and it took 15 years to complete. Georgia Power’s share alone was $11 billion, and that burden got passed on as a 25% hike in residents’ electricity bills.

This single project froze decision-making across the entire US power industry. There’s no insurance product that protects against cost overruns. Any excess lands straight on the balance sheet, and utilities have to explain rate hikes to regulators. So everyone is standing there thinking, “I’d rather not be first.” There’s interest, there’s due diligence, there are letters of intent. But nobody makes the final investment decision.

The numbers make this even clearer. BloombergNEF tallied the cooperation announcements between data center operators and nuclear companies at 51 GW — more than 50 reactors’ worth. But most of these have no binding power purchase agreement², no financing announcement, and haven’t even filed a formal application with the NRC. That’s why BNEF barely factored this 51 GW into its forecasts.

It’s not that the government is sitting on its hands, either. Quite the opposite — over the past year, US policy support has been unprecedented. A wave of nuclear-related executive orders came out at once, and the NRC pinned down a deadline to reach a decision within 18 months of application receipt. The Department of Energy rolled out loan commitments to back the construction of 10 large reactors, and even set a goal of having 10 reactors under construction by 2030.

And yet nothing moves. What’s missing right now isn’t policy push — it’s market pull. No utility is raising its hand to say “we’ll build it.” A situation where the regulations and the financing are all in place but the first buyer hasn’t shown up — anyone who’s tried to sell a new product will recognize this picture. Without an early adopter, even the best terms stay on paper.

51 GW announced, zero reactors under construction. That gap is what the Western nuclear market actually looks like right now.

---

Big Reactors Aren’t Slow Because of Technology — They’re Slow Because of the Supply Chain

So why do large nuclear plants take so long? People usually blame regulation, but from what I can see, the more decisive factor is that there are only a handful of places on Earth that can actually build them.

A gigawatt-class reactor pressure vessel is an ultra-heavy forging. The equipment capable of producing these is concentrated almost entirely in Japan and South Korea, and total global capacity runs at only two or three units a year. There’s none of this capacity inside the US. The moment you try to build multiple reactors at once, the bottleneck isn’t the blueprint — it’s this forging press.

That’s why shrinking the size widens the pool of suppliers. For a 300–500MW SMR³, far more companies are capable of building the pressure vessel. And while components are being stamped out in a factory, site grading can proceed in parallel on the ground. Shorter construction time means lower interest during construction⁴. This matters more than it sounds — when you tie up trillions of won for over a decade, interest alone can be enough to wreck the project’s economics.

What makes China different isn’t technology — it’s this repeating structure. Building six units back-to-back on a single site means the same workforce simply carries over from Unit 1 to Unit 2 to Unit 3. Some 90–95% of components are domestically made, interest rates and labor costs are low, and the permitting path is simple. The result: reactors built for about a fifth of Western cost, in five to six years.

That rhythm was on display again just last week. On August 4, 2026, China approved eight new reactors in one go — six of them the domestic Hualong One⁵ design. This is the first batch under the 15th Five-Year Plan (2026–2030). As of the end of 2025, China had 59 reactors in operation and was building 35 more totaling 41.9GW — holding the world’s largest under-construction capacity for a 19th consecutive year. BNEF projects China’s nuclear capacity could reach 102GW by 2030, overtaking the US.

This gap turns into a political problem in export markets. For a country like Saudi Arabia weighing whether to adopt nuclear power, US technology runs around $15,000 per kW, while China’s offer comes in around $2,700. A gap of more than five times over stops being a technology choice and becomes a fiscal decision. If the West wants to persuade countries to “use ours,” security arguments alone won’t be enough — the cost gap itself has to close.

Even so, China’s domestic strength doesn’t translate directly into exports. The only country where China has actually built a nuclear plant is Pakistan. It was pushed out of the UK, and in Argentina, deals keep getting reversed with every change of government. Being cheap at home and being cheap in someone else’s country are two different problems. That’s why China has separately prepared the Linglong One, a 100MW SMR. Inside China, where power demand is exploding, a 100MW reactor has no real use case. It was designed from the start as an entry-level model aimed at markets like Africa and South Asia — places where the grid is too small to handle a large reactor.

---

The Real Competition Isn’t the Reactor — It’s the Pipeline

This is where the industry’s core logic comes into view. Every price that advanced reactor companies quote today is a first-of-a-kind cost⁶. Of course it’s expensive. The price point that can compete with gas or renewables only shows up around the 10th or 12th unit.

But to get to unit 10, you need orders for 10 units lined up first. In other words, what decides the winners in this market isn’t reactor performance — it’s whether you’ve locked in a continuous order pipeline. There are currently more than 50 SMR and advanced reactor developers. BNEF estimates that only two or three of them will end up holding both the pipeline and the capital at the same time. The rest won’t disappear because their technology is bad — they’ll disappear because they never rack up enough repetitions.

Seen this way, the hyperscalers’ behavior makes sense. Google is betting simultaneously on Kairos Power (based on TRISO fuel⁷) and Commonwealth Fusion (fusion). Amazon has put both equity and a power purchase agreement into X-energy’s high-temperature gas reactor. Neither company picked a single winner — they bought an options portfolio. After all, the moment they’ll actually need the power is the mid-2030s, and right now nobody knows which company will still be standing by then.

Utilities, by contrast, don’t have that luxury. They’re bound by a least-cost-supply obligation, and if something fails, they have to explain it away with a rate hike. That’s why everyone is watching Ontario Power Generation try building the GE Hitachi 300MW model first. If that works, the expectation is that TVA and Duke Energy will follow.

India has turned this exact principle into a procurement requirement. Its position: it won’t buy SMRs that lack a commercial operating track record at home. This holds even though India has set a target of 100GW by 2047, amended its Atomic Energy Act to open the door to private participation, and set an aggressive cost target of under $2,000 per kW. The message is clear: India won’t let itself become a testing ground for unproven technology. BNEF’s forecast puts India at 13GW by 2036 — still a long way from that target.

The one thing the U.S. can reliably scale up right now isn’t new construction — it’s revival. Restarting Palisades, the Crane Clean Energy Center (formerly Three Mile Island), and Duane Arnold — three units in total — would bring back 2.2GW, while roughly 30 power uprates⁸ at existing plants would add another 2.3GW by 2036. The fact that switching existing plants back on is faster than building new ones — that, too, captures exactly where things stand today.

Oz’s Lens

One piece of advice I give most often when building a GTM strategy is “use what’s proven.” With new technology, the cost of recovering from failure outweighs the cost of adoption. But the nuclear market has created conditions where that advice simply doesn’t hold.

Going through this material again, what I confirmed is that technological maturity doesn’t attach to the technology itself. Maturity attaches to supply chains, organizations, and repetition count. By design, the AP1000 is a fully proven technology. But the US effectively has no team that has built it on a fixed budget and schedule. The one case that exists took 15 years. China, meanwhile, is repeating the same kind of build more than thirty times over. That’s why the exact same technology is a gamble on one side and routine on the other.

In last Thursday’s issue, writing about ASML, I noted that handing over the complete blueprints still wouldn’t let you build the machine. Nuclear power has exactly the same structure. The designs are public and the licensing pathway is open, but the twenty years of on-the-ground judgment that fills the gap between them no longer exists in the West. Whether it’s semiconductors or nuclear plants, the real unit of industrial competitiveness today isn’t the blueprint — it’s how many times you’ve built it recently.

I see the same pattern on the ground in AI adoption. If you pick a solution that’s supposedly proven but nobody inside your organization has ever run it, then within that organization, it isn’t proven technology. Conversely, if a team has run an unfamiliar tool two or three times, it’s already mature technology for them. What you should look at when making an adoption decision isn’t the vendor’s reference list — it’s your own organization’s repetition count.

And I can’t leave out the Korea story. Korea stands on the opposite side of the bottleneck the West is now experiencing. Doosan Enerbility signed an agreement with X-energy to secure key SMR materials in advance, and announced a ₩800 billion (~$580 million) equipment investment in a dedicated SMR factory. There are more than 50 players competing to sell reactors, but the seat that sells components to all 50 of them is far less crowded. I think that’s the far higher-probability position. To be clear, this is a read on industry structure, not an investment call on any individual stock.

---

Closing

To sum up, three things.

First, large-scale nuclear plants are no longer the safe choice in the West. After Vogtle, utilities decided not to go first — which is why there are currently 0 large-reactor construction starts underway in the US.

Second, that’s why hyperscalers chasing speed are spreading bets across unproven reactor designs. They didn’t pick one — they bought options.

Third, the contest isn’t decided by reactor performance but by repeat orders. A company that can’t drive down the cost from its first unit to its 10th unit disappears, however good its technology is.

If your organization is evaluating a new technology right now, I’d suggest asking this question before you ask for vendor references: “How many times has our own team actually run this?” Maturity isn’t a property of the technology — it’s a number attached to you.

How about you, Reader? If you’ve ever picked the “proven choice” only to find it took even longer, tell me in the comments whether the bottleneck was the technology, the people, or the supply chain. If enough stories come in, I’ll pull them together next issue into a pattern on “where maturity breaks down inside an organization.”

---

💬 Tell me in the comments about a time the “proven choice” turned out slower — and what the bottleneck was. I’ll fold it into a future issue. 📨 If you know a colleague wrestling with tech procurement or infrastructure investment decisions, send this their way.

---

📎 References & Further Reading

Primary sources

• Tom Rowlands-Rees, Stephanie Diaz, Chris Gadomski, “Nuclear Power Market Outlook 1H 2026: Approaches Diverge”, BloombergNEF Switched On, 2026. ··· This is the backbone of today’s piece. The key points are the near-total lack of real progress against the 51GW of announcements, and the forecast that “only two or three of fifty projects will survive.”
• “NRC Approves TerraPower Construction Permit”, ANS Nuclear Newswire, March 2026. ··· Lays out how the first non-light-water reactor approval in roughly 40 years came about, and how long the review took. The fact that it took 2 years from application to approval is the starting point of this piece.
• “What Was Learned from Building New Nuclear Reactors?”, POWER Magazine. ··· The source for Vogtle Units 3 and 4’s final cost of $36 billion and 15-year construction timeline. Also covers the up to 25% rate hikes.
• “8 new nuclear reactors given full approval”, China Daily, August 4, 2026. ··· The first approved batch under China’s 15th Five-Year Plan. Bear in mind this is state media, but the figures — 59 reactors operating, 35 under construction — come from China Nuclear Energy Association tallies.

Background

• Jane Nakano, “China’s Nuclear Energy Priorities Under Its 15th Five-Year Plan”, CSIS, May 2026. ··· A report that dissects where China’s speed actually comes from, at the policy-structure level. Start here if you want to dig deeper into today’s “repetition structure” argument.
• “China is building half of the world’s new nuclear power”, Global Energy Monitor. ··· Lets you confirm, with data, just how geographically concentrated capacity under construction really is.
• “India will need Rs 25 lakh crore to achieve 100 GW nuclear target by 2047”, Down To Earth. ··· A summary of a TERI report that stress-tests the realism of India’s target from a capital-requirement angle. Worth reading alongside BNEF’s more conservative forecast of 13GW by 2036.

Related past issues worth reading

• The Machine You Can’t Build Even With the Blueprint: ASML ··· Covers, from the semiconductor angle, why having a blueprint isn’t enough to actually build the machine. Exactly the same structure as today’s “proven design ≠ ability to build” argument.
• The Containment That Built China ··· Covers how constraints ended up building China’s own domestic supply chain — the backstory behind China localizing 90–95% of its nuclear components.

---

📝 Glossary

The author, Kwangseob Ahn, is a professor of business administration at Sejong University and lead consultant at OBF (Oswarld Boutique Consulting Firm). He teaches statistics and data analysis, including business data management and business analytics, while leading GTM and AI strategy consulting in the field, designing the seam between technology and business. He has published academic research on a memory architecture for AI dialogue systems (HEMA) and runs Daily Arxiv, a daily curation of global AI papers. He holds a master's from Korea University's Graduate School of Technology Management and a KMBA. He is the author of Homo Brainless: The People Who Outsource Their Thinking.

Footnotes

1. Sodium Fast Reactor (SFR): A reactor cooled with liquid sodium instead of water. Sodium transfers heat better than water, which allows for a smaller design, but it reacts violently with air and water, making it tricky to handle. ↩

2. Power Purchase Agreement (offtake): A contract, signed before a plant is even built, that locks in who will buy its electricity and at what price. Banks won’t lend without one, which effectively makes it the document that decides whether construction happens at all. ↩

3. SMR (Small Modular Reactor): A reactor that works on the same principles as conventional plants but scaled down to the 300–500MW range. The goal is to manufacture components in a factory and assemble them on-site. ↩

4. Interest During Construction (IDC): Financing costs that accrue while a plant is being built. Since the project can’t sell a single kWh during that period, the longer construction drags on, the faster the economics deteriorate. ↩

5. Hualong One: A roughly 1,000MW pressurized water reactor designed in-house by China. It’s also the flagship model for China’s nuclear exports. ↩

6. First-of-a-kind cost (FOAK): What it costs to build a given design the first time. Repeated builds bring costs down through learning effects, and the point they settle at is called the nth-of-a-kind (NOAK) cost. ↩

7. TRISO fuel: Fuel made of uranium particles wrapped in multiple layers of ceramic shell. The shell itself acts as a containment barrier, keeping radioactive material from leaking out even at high temperatures. ↩

8. Uprating: Getting more electricity out of an existing reactor by upgrading surrounding equipment rather than building a new plant. Much faster and cheaper than new construction. ↩