Oklo Inc. is turning spent nuclear fuel from a Cold War-era Idaho reactor into the clean power backbone of America's AI economy - one compact powerhouse at a time.
In the high desert outside Idaho Falls, a small A-frame building - less the size of a ski lodge than a traditional nuclear plant - is quietly taking shape on the grounds of Idaho National Laboratory. It is Oklo's Aurora powerhouse, and if the company's founders get their way, it will mark the first commercial deployment of an advanced fast-fission micro-reactor in American history. The company has no revenue yet, no operating reactor, and a stock that swung from $194 to $61 in the span of a few months. But in March 2026, with the U.S. Department of Energy signing off on its Nuclear Safety Design Agreement and the DOE's Reactor Pilot Program pressing hard toward a July 4 criticality target, the calendar is finally aligning with the ambition.
The origin story of Oklo Inc. is, in its own way, a creation myth built on physics. Jacob DeWitte and Caroline Cochran - now DeWitte - met in the nuclear engineering department at the Massachusetts Institute of Technology. He grew up in New Mexico, steeped in the nuclear culture of the Southwest; she came from Oklahoma's oil and gas country. Together, they wanted to rethink what a nuclear power plant could be: not the sprawling, multi-billion-dollar megaprojects that had defined the industry, but something small, self-contained, and deployable almost anywhere.
They named their company after the Oklo region of Gabon, where, nearly 1.7 billion years ago, a vein of uranium ore acted as a natural nuclear reactor - sustaining fission spontaneously, without human intervention, for hundreds of thousands of years. The name is deliberate. It carries a philosophy: that nuclear reactions, properly designed, are not explosive or chaotic but steady, patient, and self-governing.
The couple founded Oklo in 2013 and, that same year, crossed paths with Sam Altman, then still at the startup incubator Y Combinator. Altman had long believed that the future of energy and the future of artificial intelligence were inseparable. He became an investor, a champion, and eventually chairman of Oklo's board - a position he held from 2015 until April 2025, when he stepped down to clear the way for Oklo to pursue power supply agreements with companies that compete directly with his own OpenAI. That resignation, counterintuitively, was a signal of confidence. It meant Oklo's commercial pipeline had grown large enough to create actual conflicts of interest. Altman retains roughly 4% of the company.
What Oklo is building at Idaho National Laboratory is unlike anything else currently under construction in the American nuclear sector. The Aurora powerhouse is a fast neutron reactor - a design class that uses high-speed neutrons to drive fission, rather than the slow, moderated neutrons found in conventional light-water reactors. It uses liquid sodium as a coolant, runs on high-assay low-enriched uranium, and can operate on recycled spent nuclear fuel. The design ranges from 15 megawatts electric to 75 megawatts electric depending on configuration - a power output that matches almost perfectly with the discrete computing halls inside a hyperscale data center. Jacob DeWitte has made this point repeatedly to investors: data centers are built in modular halls of roughly 15 to 50 megawatts, and Aurora grows with them.
The engineering lineage is older and more proven than the startup wrapper might suggest. Oklo's Aurora draws directly on the Experimental Breeder Reactor II, known as EBR-II, which ran at Idaho National Laboratory from 1964 to 1994. A defining moment in EBR-II's history came during a 1986 safety test - the reactor was deliberately deprived of coolant flow while running at full power, and it shut itself down without operator intervention, without emergency systems, without external power. The laws of physics did the work. Oklo has built the Aurora on that same principle: strongly negative temperature reactivity coefficients mean the reactor reduces its own output as it heats up. There are no cooling pumps to lose, no pressure vessels to rupture.
The physical plant reflects that simplicity. There are no massive cooling towers, no redundant safety systems bolted onto a complex core. Solar panels on the roof power the control electronics. The entire structure can be sited in remote locations - military installations, Arctic outposts, mining operations, data center campuses - that have no access to the grid and no appetite for diesel generators running around the clock.
The path to March 2026 has not been clean. In January 2022, the Nuclear Regulatory Commission denied Oklo's first combined license application for a smaller 1.5-megawatt version of the Aurora reactor, citing insufficient information about safety classifications. The rejection - arriving on the same day Oklo had submitted its original application as the pandemic was declared - was a setback that pushed the company to rethink both its design and its regulatory approach.
In the years since, DeWitte repositioned Oklo's regulatory strategy. Instead of waiting for the NRC to adapt its processes, Oklo helped build the framework for those processes. The company submitted a Principal Design Criteria topical report that was accepted for review in just 15 days - less than half the normal 30-to-60-day window. It also filed a Product-Based Operator Licensing Framework, a novel approach that would allow operators to be licensed to the Aurora design itself, rather than to each individual site - a breakthrough that, if approved, could dramatically accelerate the path to multi-plant deployment.
The more significant shift came in May 2025, when President Donald Trump signed an executive order on nuclear reactor testing, launching the DOE's Reactor Pilot Program. The program - with its explicit goal of achieving criticality in at least three test reactors by July 4, 2026 - offered an accelerated path outside the traditional NRC framework. Oklo secured three of the eleven program slots, more than any other company. In September 2025, it broke ground at INL as the third participant to do so, with an event that drew Idaho and Utah governors, two U.S. senators, the EPA administrator, the Secretary of the Interior, and a senior NRC commissioner.
In March 2026, the DOE Idaho Operations Office approved the Nuclear Safety Design Agreement for the Aurora powerhouse at INL. Oklo immediately requested DOE commence review of its Preliminary Documented Safety Analysis - the formal next step under the Reactor Pilot Program authorization pathway. For a company that had spent more than a decade in pre-application conversations with regulators, these were not incremental steps. They were the beginning of real execution.
Oklo's business model has never been to sell reactors. It intends to own them, operate them, and sell power directly to customers under long-term contracts - sometimes for 20 years, sometimes for 30. This "Power as a Service" structure shifts capital risk onto Oklo and regulatory burden away from the customer. It also creates recurring revenue that, once flowing, is highly predictable.
The order book has grown accordingly. In June 2025, the Defense Logistics Agency, acting on behalf of the U.S. Air Force, named Oklo as the intended awardee for a 30-year, fixed-price contract to build and operate an Aurora powerhouse at Eielson Air Force Base in Alaska. Eielson, home to the 354th Fighter Wing and its F-35A aircraft, sits in the sub-Arctic and runs on diesel power that is expensive, exposed to supply disruption, and fundamentally incompatible with operational independence. A reactor that requires no grid connection and can run autonomously for up to a decade without refueling is not an incremental improvement - it is a different category of energy security.
On January 9, 2026, Meta Platforms announced an agreement with Oklo to develop 1.2 gigawatts of nuclear capacity in Pike County, Ohio - about 85 miles from Meta's planned "Prometheus" data center campus in New Albany. The deal sent OKLO shares up more than 20% in a single session. The first phase is expected online around 2030, with the campus scaling to full capacity by 2034. Meta provided upfront funding to help Oklo secure fuel and begin early site work - a structure that effectively transfers some development risk to the hyperscaler and signals a level of conviction unusual in pre-revenue energy partnerships. Oklo also holds letters of intent with Diamondback Energy for 50 megawatts of power in the Permian Basin, and with Wyoming Hyperscale for 100 megawatts at a data center campus, both over 20-year terms. Master Power Agreements with Switch data centers and others have contributed to a cumulative pipeline Oklo now describes as approximately 15 gigawatts of interest.
No account of Oklo's trajectory can skip the fuel question, because it is the most serious structural risk to the company's ambitions at scale. The Aurora powerhouse runs on high-assay low-enriched uranium, known as HALEU - uranium enriched between 5% and 19.75% U-235, far above the 3-5% used in conventional reactors. Domestic HALEU supply is, at present, almost nonexistent. Centrus Energy operates a small enrichment cascade at Piketon, Ohio, but production levels remain far below what a commercial deployment of Aurora reactors would require at scale.
Investment firm Kerrisdale Capital Management released a critical short-seller report in late 2024 warning that Oklo's HALEU dependency represented a structural constraint its cost projections and timelines did not adequately account for. The report captured a real tension: the company's 15-gigawatt order book implies future capacity that domestic HALEU production plans across all producers could not plausibly support by 2040. Community groups near the planned Ohio campus, citing the site's proximity to the contaminated former Portsmouth Gaseous Diffusion Plant, have raised separate concerns about concentrating unproven reactor technology alongside expanding uranium enrichment operations.
Oklo's answer to this problem is its fuel recycling program - the feature of its technology that, if it works at scale, changes the economics entirely. The company's fast reactor design can run on metallic fuel recycled from existing spent nuclear fuel, of which the United States holds more than 80,000 tons in temporary storage. The Aurora Fuel Fabrication Facility at INL - whose Nuclear Safety Design Agreement received DOE approval in November 2025 - will fabricate the first fuel assemblies for the Aurora powerhouse using recovered material from EBR-II. In parallel, Oklo is developing an Advanced Fuel Center in Tennessee and has completed fast-spectrum plutonium criticality experiments supporting plutonium as a bridge fuel source.
"Advanced fuel fabrication and recycling technologies represent a significant unlock for our business, addressing fuel-supply challenges while transforming fuel economics and creating new revenue opportunities," Jacob DeWitte said at the time of the A3F approval. Recycled fuel at meaningful commercial scale is not expected before 2029, and the pathway remains technically demanding. But the logic is sound: Oklo's technology was designed from the beginning to close the fuel cycle - turning the nuclear sector's most intractable waste problem into a competitive advantage.
While the Aurora powerhouse remains in construction, Oklo's firs…