Sam Altman Bet Millions on a Hot-Tub-Sized Nuclear Reactor Before Anyone Cared About AI. The Stock Is Up 500%.

A full decade before Sam Altman released ChatGPT and became the most famous human in technology, he was running Y Combinator — the world's most prestigious startup accelerator. As YC's president, Altman shepherded companies like Cruise, Flexport, and Scale AI from scrappy garage ideas into multi-billion-dollar juggernauts.

But there was one startup he believed would be bigger than all of them. So much so that after the company graduated from YC, Altman didn't just wave goodbye and move on to the next batch. He became chairman of the board. He led their next investment round. He put millions of his own money into it. And then, in 2024, he put his reputation on the line to take it public via SPAC — a financial vehicle so thoroughly associated with fraud and disappointment that doing it voluntarily is either an act of supreme confidence or clinical insanity.

That company is Oklo. Its stock is up 500% in the past year and 200% in 2025 alone. It has never generated a dollar of revenue. It has never built a product. It has not yet been approved to build that product. And its pitch to the world is this: we can construct nuclear reactors the size of a hot tub that are self-sustaining, physically incapable of melting down, require zero human operators, and will run for twenty years without refueling.

This is either the most important energy company of the 21st century or the most elaborately funded science fair project in financial history. Let's figure out which.

Oklo's CEO and co-founder, Jacob DeWitte, grew up in New Mexico — specifically, in the backyard of Los Alamos National Laboratory, where the first nuclear weapons were developed. If you're looking for an origin story that screams "this person was always going to end up in the nuclear business," growing up next to the birthplace of the atomic bomb is about as on-the-nose as it gets.

While doing his PhD in nuclear engineering at MIT, Jacob met his co-founder and COO, Caroline Cochran. They bonded over a shared obsession with a natural phenomenon that sounds like it was invented by a science fiction writer who'd given up on plausibility: a natural nuclear reactor that ran safely, by itself, for hundreds of thousands of years — two billion years ago in West Africa.

Both founders grew up in the shadow of Chernobyl and Fukushima — disasters that turned public opinion and government funding decisively against nuclear power. But Caroline and Jacob understood something that the broader public discourse often misses: the fear of accidents isn't the only reason nuclear hasn't scaled. The economics are equally catastrophic.

Traditional nuclear plants are financial nightmares. Projects routinely run billions over budget and years behind schedule. The industry's track record for delivering on time and on budget is, to put it charitably, non-existent. To Caroline and Jacob, the question wasn't "can we make nuclear safer?" It was: "can we make nuclear smaller, cheaper, and so simple that it doesn't require an army of PhD physicists to keep it from exploding?"

To appreciate what Oklo is attempting, you first need to understand exactly how absurd traditional nuclear power plants are. Let's use Plant Vogtle in Georgia as our case study, since it's the most recent major nuclear project completed in the United States and it went about as smoothly as a root canal performed during an earthquake.

The reactor vessel alone weighs 12,200 tons and requires specialized transportation and industrial cranes to install. The plant covers thousands of acres. It needs direct connection to major electrical grids. It consumes enormous amounts of water — up to three liters for every kilowatt-hour produced. The control rooms look like NASA mission control. There are thousands of safety systems, and backup systems for the backup systems, because when something goes wrong with a traditional nuclear reactor, the consequences range from "incredibly expensive" to "uninhabitable exclusion zone for decades."

This is the industry Oklo wants to disrupt. And their pitch is, essentially: what if none of that?

Oklo's Aurora reactor generates 15 to 75 megawatts of power — enough to run 10,000 to 60,000 homes, or more relevantly, enough to power a large data center or military base for two decades without refueling. It's designed to be buried underground, shipping-container sized, and operated entirely remotely.

The key technical differentiator — the thing that makes nuclear engineers either very excited or very nervous — is that Aurora uses fast neutrons instead of the slow (thermal) neutrons that traditional reactors depend on. This allows it to run on HALEU fuel: High-Assay Low-Enriched Uranium, which is more concentrated than standard reactor fuel, enabling a much smaller core while still producing significant power output.

Instead of water cooling — which is what ties traditional nuclear plants to rivers, coastlines, and enormous cooling towers — Aurora uses liquid metal coolant transported through advanced heat pipes. This means the reactor can theoretically be deployed in deserts, Arctic bases, remote oil fields, or anywhere else that doesn't happen to have a convenient river nearby. It removes the geographical constraint that has defined nuclear power for seventy years.

This isn't just theoretical handwaving. In 1986, scientists at Idaho National Laboratory deliberately sabotaged the safety systems of an experimental fast reactor in a test that was either the most courageous or the most insane experiment in nuclear history. They shut off the cooling, pulled out the control rods, and waited to see what happened. The reactor shut itself down. The physics worked. Nobody died. The lab wrote a paper about it, presumably after their heart rates returned to normal.

It's no coincidence that Oklo chose Idaho National Laboratory as the site for its first commercial reactor, targeted for operation by 2027.

Then there's the nuclear waste angle, which is the part that sounds too good to be true (because it might be). Traditional nuclear plants produce radioactive waste that remains dangerous for thousands of years. The United States currently has roughly 90,000 tons of the stuff sitting in storage with no permanent disposal solution. Oklo claims its reactors can actually consume existing nuclear waste as fuel, converting a massive national liability into an energy source. If this works at scale, it doesn't just solve Oklo's fuel problem — it addresses one of the longest-running political quagmires in American energy policy.

Here's where Oklo diverges from every nuclear company that came before it, and where you can see why Sam Altman — a man whose entire worldview revolves around platform economics and recurring revenue — fell in love with this company years before the AI boom made it obvious.

Most nuclear companies try to sell you a reactor. Oklo wants to sell you the electricity.

Instead of customers buying, licensing, and operating their own nuclear plants (a process roughly as fun as it sounds), Oklo builds, owns, and operates the reactors themselves. Customers sign long-term power purchase agreements — PPAs — buying Oklo's clean electricity at a fixed price for 20 to 30 years. It's the SaaS model applied to nuclear fission. "Reactor-as-a-Service," if you will. (Please don't actually call it that.)

And who are these customers? To Oklo's considerable advantage, they're some of the most cash-rich and power-desperate entities on the planet: data centers, military installations, and industrial operations like oil and gas fields. These are facilities that are often geographically isolated from traditional power grids, willing to pay premium prices for reliable electricity, and constitutionally incapable of tolerating downtime.

Let's run the back-of-the-napkin math that makes venture capitalists hyperventilate. A 50-megawatt Aurora reactor running at roughly 80% capacity generates around 350,000 megawatt-hours per year. At Oklo's target price of $80-$130 per megawatt-hour — let's call it $100 — a single reactor produces about $35 million in annual revenue, or $700 million over its 20-year lifespan.

Oklo has signed letters of intent for 14,000 megawatts of capacity. That's approximately 280 reactors. If they could deliver on all of it — a galactic "if" — that's $9.8 billion in annual revenue and $200 billion in total lifetime revenue. Oklo's multi-billion dollar market cap reflects investors' belief that the company can capture even a fraction of this demand.

Early customers include Switch (data centers), Diamondback Energy (oil and gas), and — here's the one that makes defense analysts perk up — the United States Air Force, which awarded Oklo a contract to power the extremely remote Eielson Air Force Base in Alaska. When the Department of Defense starts writing checks for your pre-revenue nuclear startup, you've either built something real or pulled off one of history's great cons.

For all of Oklo's technological elegance, visionary business model, and 500% stock chart, there is one fact that the bulls need to sit with in uncomfortable silence: the Nuclear Regulatory Commission already rejected them once.

In January 2022, the NRC denied Oklo's application for a combined construction and operating license, citing that the company "repeatedly failed to submit substantive technical information." That's not "we have some minor concerns about your paperwork." That's "you didn't give us enough to even evaluate this properly." For a company whose entire thesis depends on regulatory approval, getting rejected by the only agency that can grant it is — to use a technical term — very bad.

Oklo has since re-engaged with the NRC and is preparing to submit a new proposal. There's also been meaningful political tailwind: nuclear energy expansion is one of the rare genuinely bipartisan priorities in Washington. Biden's Inflation Reduction Act included billions in nuclear incentives. Trump signed executive orders in May 2025 directing the NRC to reduce regulatory burdens and expedite licensing. Both parties want more nuclear. The question is whether the bureaucracy can move fast enough for Oklo's timeline.

But regulation is only one of several existential risks. Let's catalog them, because pretending they don't exist would be journalistic malpractice.

The fuel problem. Oklo's reactors run on HALEU — a type of enriched uranium that barely exists as a commercial product. The U.S. Department of Energy has made conditional commitments to provide it, but the infrastructure to produce HALEU at scale is still in early development. Oklo is essentially building a car and hoping the government can figure out how to produce the gasoline before the car rolls off the assembly line. This is not an exaggeration. It's almost literally their situation.

The prototype problem. Oklo is targeting late 2027 for its first reactor at Idaho National Laboratory, and — in a move that's either brilliantly aggressive or dangerously arrogant — they're skipping the prototype phase entirely. They plan to go straight to a commercial reactor. In the nuclear industry, where "move fast and break things" is not a philosophy anyone wants to see in practice, this is raising eyebrows among engineers and regulators alike.

The patience problem. It will take a decade or more for Oklo to prove this business has legs from a fundamentals perspective. Overruns, delays, and regulatory slowdowns aren't risks in the nuclear industry — they're certainties. The question is whether investors who bought the stock at 500% above its SPAC price will still be around when the first reactor finally switches on.

Oklo isn't the only company racing to reinvent nuclear energy. But the competitive landscape is more nuanced than a simple horse race.

NuScale Power is Oklo's most direct public-market comparison. It already has NRC approval for its small modular reactors — a significant lead — but its design uses traditional light-water technology, which is proven but comes with all the water-dependency and scale limitations that entails. NuScale has 12 units in production and is targeting 2030 for an operational reactor.

TerraPower, Bill Gates' private nuclear venture, is chasing a similar fast-reactor design as Oklo but at much larger scale — 345 MW modules intended to serve major grids and cities. Like Oklo, TerraPower runs on HALEU and hasn't received NRC approval.

But comparing these companies directly is a bit like comparing a cargo ship to a speedboat to a submarine. They're all in the water, but they're doing fundamentally different things. NuScale and TerraPower are building reactors for massive electrical grids serving entire cities. Oklo's mini-reactor strategy is uniquely focused on powering specific commercial sites that are underserved by — or entirely disconnected from — traditional power infrastructure. Remote data centers, military bases in Alaska, oil platforms in the middle of nowhere.

Here's where all the threads come together, and where Sam Altman's decade-old bet starts to look less like a lucky guess and more like genuine foresight.

Altman understood something before almost anyone else in Silicon Valley: there cannot be abundant AI without abundant energy. Now that AI is in everyone's hands and the scale of compute demand is clear, this thesis has been proven correct with almost embarrassing precision.

According to McKinsey, data centers already account for 5% of total U.S. energy demand. That share is projected to nearly triple to 12% by 2030. Tech giants like OpenAI, Google, Amazon, and Meta are hitting an energy wall they will pay essentially any price to break through. There simply is not enough clean, reliable power to run the AI revolution at the scale these companies are planning.

In March 2025, a coalition of the world's largest companies — including Amazon, Google, and Meta — along with 14 banks and 31 countries, pledged to help triple global nuclear capacity by 2050. The mandate for nuclear investment isn't coming from governments and utilities alone anymore. The world's most valuable companies and biggest financial institutions have a mandate to put capital behind nuclear. And it's increasingly clear that traditional reactors can't be built fast enough or cheaply enough to make the tripling target achievable.

This is Oklo's strategic moat — if they can execute. While competitors are building reactors for city-scale grids, Oklo is specifically targeting the customers who need power most urgently and are willing to pay the most for it: AI data centers. A single Aurora reactor could power a hyperscale data center for two decades with zero carbon emissions, no water consumption, and no connection to an overstretched electrical grid.

For a company like OpenAI — whose CEO happens to be Oklo's chairman — that's not just appealing. It's existential.

Let's be honest about what Oklo is right now. It's a pre-revenue company trading at a multi-billion dollar market cap on the basis of technology that hasn't been built, regulatory approval that hasn't been granted, and fuel supply that doesn't yet exist at scale. Its stock price has increased 500% in a year. The last time a SPAC went up 500% on theoretical revenue, the company was usually being investigated by the SEC six months later.

But let's also be honest about what Oklo could be. If the physics works at commercial scale — and the physics has been proven in laboratory settings — Oklo could be the company that cracks the most important infrastructure problem of the 21st century. Not just for AI, but for energy independence, military readiness, industrial decarbonization, and the 90,000 tons of nuclear waste rotting in storage facilities across America.

The honest assessment is this: Oklo is a genuine, high-conviction bet on the intersection of two of the century's most consequential trends — AI and clean energy. The technology has theoretical backing. The demand is real and growing exponentially. The business model is elegant. The political environment is favorable. And the guy who saw ChatGPT coming a decade before anyone else has been writing personal checks for this company since before most people had heard of it.

But the gap between "theoretically possible" and "commercially deployed at scale" has swallowed nuclear companies whole for seventy years. The NRC has already said no once. The fuel supply chain doesn't exist yet. The company is skipping prototyping. And the stock is priced for a future that is, at minimum, five to ten years away from validation.

Oklo is simultaneously the most compelling and the most terrifying investment thesis in the energy sector. It's a bet that two engineers inspired by a two-billion-year-old natural reactor in West Africa can do what the entire nuclear industry has failed to do for half a century: make nuclear power small, cheap, safe, and fast enough to matter.

Two billion years ago, the laws of physics ran a reactor in Gabon without any human involvement at all. Caroline and Jacob are betting those same laws will work again — this time, on purpose, in Idaho, powering the servers that run the AI revolution.

If they're right, this stock is a steal at 500%. If they're wrong, well — the graveyard of nuclear energy promises has plenty of open plots.