The math for powering an AI data center is simple, brutal, and getting more expensive by the quarter. A single large facility can draw as much electricity as a medium-sized city, and the tech giants building them are now signing 24/7 clean power contracts for gigawatts at a time. Into that gap steps Oklo, a company that has spent the last decade designing a nuclear reactor the size of a small warehouse. Its bet is that the only thing dense enough, and clean enough, to sit next to a hyperscale campus is a fission powerhouse that runs on recycled nuclear waste.
Oklo went public via a SPAC in May 2024, and its stock has since surged over 500% year-to-date, giving it a market cap that PitchBook pegged at $12.4 billion in August 2025 [PitchBook, Aug 2025]. The valuation reflects a staggering amount of forward-looking optimism for a company that has yet to turn on its first commercial reactor. The optimism is pinned to an order book. According to independent research and company announcements, Oklo has non-binding agreements for roughly 18 gigawatts of power supply, primarily with data center operators and an oil producer [Independent Research, Unknown]. That includes a massive, non-binding Master Power Agreement with Switch for up to 12 GW through 2044, and a 500 MW deal with Equinix [Datacenter Dynamics, 2026]. For context, 18 GW is more than the total nuclear capacity of Sweden.
The reactor in the parking lot
Oklo’s product is the Aurora powerhouse, a sodium-cooled fast reactor designed to generate between 15 and 75 megawatts of electricity. It is small enough to be largely factory-built and shipped by truck. The core innovation isn’t just the size, but the fuel cycle. The Aurora is designed to run on metallic HALEU (high-assay low-enriched uranium) and, critically, on recycled spent nuclear fuel from conventional reactors [Oklo, Unknown]. This addresses two problems at once: it offers a potential path to reducing existing nuclear waste stockpiles, and it aims to create a more secure, domestic fuel supply chain. The company is also pursuing fuel recycling services and radioisotope production, aiming to build a vertically integrated nuclear business [Oklo, Unknown].
Its first demonstration unit is slated for Idaho National Laboratory, with a target for commercial operation in late 2027 or early 2028 [ANS / Nuclear Newswire, 2025]. The company has secured a site use permit from the Department of Energy and, in a key step, received approval for its Preliminary Documented Safety Analysis under the DOE’s Reactor Pilot Program [InvestingNews.com, 2026]. It has also lined up industrial partners, with Kiewit Nuclear Solutions signed as lead engineering, procurement, and construction contractor and Siemens Energy contracted to supply the power conversion system [Neutron Bytes, 2025] [Oklo, 2025].
Founders with a long fuse
Oklo was founded in 2013 by Jacob and Caroline DeWitte, both MIT-trained nuclear engineers. This isn’t a pandemic-era startup chasing a trend; the DeWittes have been working on advanced reactor concepts for over a decade. Jacob DeWitte, the CEO, previously founded UPower, a Y Combinator-backed company developing nuclear thermal batteries [TechCrunch, 2014]. He has since been appointed to the U.S. President’s Council of Advisors on Science and Technology [LinkedIn, 2026]. Caroline DeWitte, a co-founder and former COO, has been a frequent spokesperson on nuclear technology and regulation. The company’s early association with Sam Altman, who was chairman until April 2025, provided a high-profile link to the AI and data center world that it now explicitly targets [ANS / Nuclear Newswire, 2025].
| Founder | Role | Key Background |
|---|---|---|
| Jacob DeWitte | CEO & Co-Founder | MIT PhD in nuclear engineering; founder of UPower; member of PCAST [Forbes, Unknown] [Fortune, 2026] |
| Caroline DeWitte | Co-Founder & former COO | MIT-trained nuclear engineer; former COO of Oklo [Oklo, Unknown] |
The data center wedge
Oklo’s commercial strategy is notably focused. While it lists industrial sites, communities, and defense as potential customers, its publicly announced pipeline is dominated by tech. The company has been described in research as “the most datacenter-focused nuclear developer” [Independent Research, Unknown]. This is a calculated wedge. Data center operators are desperate for firm, carbon-free power that they can site close to their loads, and they have the balance sheets to sign the long-term power purchase agreements Oklo needs to finance construction. The partnership with Vertiv to co-develop power and thermal management solutions for data centers underscores this focus [Datacenter Dynamics, 2026].
The unit economics, at least on paper, are compelling. A single 75 MW Aurora reactor could power roughly 60,000 homes, or one very large data center campus, around the clock for years with a single fuel loading. The company’s reported customer pipeline of 2,100 MW for its Aurora reactors suggests a baseline of demand, while the broader 18 GW of non-binding agreements points to the scale of ambition [World Nuclear News, Unknown] [Power-eng.com, Unknown].
Where the concrete meets the coolant
The risks here are not subtle. Oklo’s entire business case rests on executing a first-of-a-kind nuclear project on time and on budget, a feat that has eluded much larger players in the industry. The company’s initial license application was denied by the Nuclear Regulatory Commission in 2022 due to gaps in safety information, a stark reminder of the regulatory mountain it must climb [Private candid take]. It is now pursuing licensing through the DOE’s alternative pathway, but final regulatory approval for commercial deployment is not guaranteed.
The other looming challenge is fuel. While using recycled waste is a brilliant narrative, establishing a supply chain for HALEU or processed spent fuel at commercial scale is itself a monumental task. The U.S. industrial base for advanced nuclear fuel is nascent. Oklo’s plans for a fuel recycling facility in Oak Ridge, Tennessee, are a necessary part of the plan, but they add another capital-intensive, regulatory-heavy project to the critical path.
- Regulatory execution. The NRC denial was a setback, and the new DOE pathway, while progressing, remains unproven for a commercial power plant. Every month of delay is expensive.
- First-of-a-kind costs. The history of new nuclear designs is a history of cost overruns. Oklo’s factory-build approach aims to avoid this, but the proof will be in the first poured concrete.
- Fuel chain reality. Securing enough HALEU or recycled fuel to feed even a fraction of its 18 GW pipeline requires building an entire industrial sector that does not yet exist.
The next twelve months
For Oklo, 2025 and 2026 are about turning paper into steel. The critical milestones are visible: breaking ground on the first Aurora at Idaho National Laboratory, securing the remaining regulatory approvals for that build, and demonstrating progress on its fuel fabrication facility. The company raised $440.6 million in a post-IPO equity round in June 2025, which should fund these near-term steps [Crunchbase, 2026]. The market will be watching for any of those non-binding gigawatt agreements to convert into firm orders, which would provide the revenue visibility needed for the even larger financing rounds required for serial construction.
A back-of-the-envelope calculation puts the scale in perspective. If Oklo converts just its 500 MW agreement with Equinix into built reactors, and charges a conservative $60 per megawatt-hour on a 20-year PPA, that single deal represents over $5 billion in lifetime revenue. The capital required to build that capacity, however, would be several times that amount. The company must prove it can build its first reactor for something close to its projected cost, which it has not publicly disclosed.
Oklo’s ultimate competition isn’t another advanced reactor startup. It’s the combined might of natural gas peaker plants, grid-scale batteries, and next-generation geothermal,all of which are also racing to sell 24/7 clean power to data centers. To win, Oklo doesn’t just need to be safe and approved. It needs to be cheaper, and it needs to be built. The clock, and the data center power demand, is ticking.
Sources
- [PitchBook, Aug 2025] Oklo stock price and market cap data | https://pitchbook.com
- [Independent Research, Unknown] Oklo described as most datacenter-focused nuclear developer |
- [Datacenter Dynamics, 2026] Oklo signs agreements with Switch and Equinix | https://www.datacenterdynamics.com
- [Oklo, Unknown] Aurora powerhouse technical specifications | https://www.oklo.com
- [ANS / Nuclear Newswire, 2025] Oklo timeline and Sam Altman step-down | https://www.ans.org
- [InvestingNews.com, 2026] DOE safety analysis approval | https://investingnews.com
- [Neutron Bytes, 2025] Kiewit EPC agreement and fuel fabrication progress | https://neutronbytes.com
- [Oklo, 2025] Siemens Energy power conversion system contract | https://www.oklo.com
- [TechCrunch, 2014] Profile of Jacob DeWitte and UPower | https://techcrunch.com
- [Forbes, Unknown] Jacob DeWitte educational background | https://www.forbes.com
- [Fortune, 2026] Jacob DeWitte appointed to PCAST | https://fortune.com
- [World Nuclear News, Unknown] Oklo customer pipeline of 2,100 MW | https://www.world-nuclear-news.org
- [Power-eng.com, Unknown] Oklo 18 GW order book details | https://www.power-eng.com
- [Crunchbase, 2026] Oklo post-IPO equity round | https://www.crunchbase.com
- [LinkedIn, 2026] Jacob DeWitte biography and appointments | https://www.linkedin.com