Oklo Aurora: A Full Guide to the Sodium-Cooled Fast Reactor, DOE Authorization, Meta and Switch Deals, and Honest Limits

Oklo is the most-watched name in advanced nuclear, and almost all of that attention lands on one machine: the Aurora powerhouse. It is a compact, metal-fueled, sodium-cooled fast reactor that Oklo wants to build in fleets next to data centers, military bases and industrial sites. The first unit, Aurora-INL, is under construction at Idaho National Laboratory, and Oklo says it is scheduled to start up in 2028.

The hype around Oklo moves faster than the concrete, so this full guide sticks to what is public as of October 17, 2026: how the reactor works, the DOE authorization path Aurora-INL is on, what the Groves test reactor proved, the Meta and Switch deals, the fuel plan, the balance sheet, how Aurora compares with TerraPower Natrium and NuScale, and the honest limits.

For the wider landscape of nuclear options for AI power, keep Inside Deep Tech's small modular reactors guide open in another tab. This piece focuses on one design and one company.

Last updated: October 17, 2026.

Key Takeaways

• **Oklo Aurora** is a pool-type, sodium-cooled fast reactor rated up to **75 MWe**, with Aurora-INL planned to start near **45 MWe**.
• DOE approved the Aurora-INL Preliminary Documented Safety Analysis in June 2026 under the Reactor Pilot Program, and startup is scheduled for 2028.
• Oklo's Groves test reactor reached first criticality on August 5, 2026, less than a year after groundbreaking, the company's first proof that it can build and start a reactor.
• Meta signed to prepay for power from a planned **1.2 GW** Aurora campus in Ohio, and Switch holds a non-binding **12 GW** framework through 2044.
• Honest limit: Oklo reported **$1.2 million** of revenue and an **$81.6 million** net loss in the first half of 2026, and no Aurora unit has produced a single kilowatt-hour yet.

What the Oklo Aurora powerhouse actually is

Oklo describes Aurora as a metal-fueled, sodium-cooled fast reactor that can run on high-assay low-enriched uranium (HALEU), recycled fuel made from used nuclear fuel, and other qualified fuels. In plain terms, it is a small liquid-metal reactor in the lineage of the Experimental Breeder Reactor II (EBR-II), which ran at the same Idaho site for three decades.

Three design choices define it:

• Fast neutrons. There is no water to slow neutrons down. Fast-spectrum reactors can burn a wider range of fuels, including recycled material, which is central to Oklo's long-term fuel pitch.
• Liquid sodium coolant at near-atmospheric pressure. Sodium boils far above the reactor's operating temperature, so the primary system does not need a thick pressure vessel. DOE's environmental filing describes near-ambient-pressure operation, inherent reactivity feedback, passive safety features and air-cooled heat rejection.
• Metal fuel. Aurora-INL uses HALEU uranium-zirconium metal fuel enriched below 20% U-235, the same fuel family EBR-II used.

Aurora is also a business model, not just a reactor. Oklo plans to own and operate its powerhouses and sell power and heat under long-term contracts, rather than sell reactors to utilities. That keeps operating know-how inside Oklo, and it also keeps all of the construction and fuel risk on Oklo's balance sheet.

Oklo Aurora specs at a glance

Source: DOE Idaho Operations Office environmental determination for Aurora-INL and Oklo's PDSA announcement. Oklo has also quoted Aurora in a 15 to 75 MWe range in earlier materials; the 75 MWe figure is the current maximum.

How Aurora makes power

The core sits in a pool of liquid sodium. Fission heats the sodium, which carries heat to a power conversion system that drives a generator. Because sodium is a liquid metal with high thermal conductivity, the reactor can run hot, above 450 C in Oklo's own description, without pressurizing the primary loop.

The safety case leans on physics more than pumps. Metal fuel expands as it heats, which reduces reactivity, and a large sodium pool absorbs heat if forced cooling stops. EBR-II demonstrated this passive shutdown behavior in tests in 1986, and that history is the foundation Oklo builds on. The trade-off is that sodium reacts with water and air, so sodium handling, leak detection and fire protection become core engineering problems instead of afterthoughts.

For how on-site generation plugs into an AI campus, see Inside Deep Tech's behind-the-meter generation guide and the 800 VDC power distribution guide.

The licensing path: why Aurora-INL runs through DOE, not the NRC

Oklo's first attempt went through the Nuclear Regulatory Commission. It filed the first custom combined license application for an advanced reactor in 2020, and the NRC denied the application without prejudice in January 2022, citing missing information. That denial still shapes how investors read Oklo.

Aurora-INL now uses a different door. In 2025 DOE launched the Reactor Pilot Program, which lets selected projects build and operate under DOE authorization on a faster track. The milestones are a Nuclear Safety Design Agreement, a Preliminary Documented Safety Analysis (PDSA), a final Documented Safety Analysis (DSA), a readiness review and startup authorization.

Oklo has cleared the first two. DOE's Idaho Operations Office approved the Aurora-INL PDSA, a review of the project's hazard analysis, accident analysis, safety controls and design commitments. On the Q2 2026 call, management said the remaining steps are DSA approval, the readiness review and startup authorization, with startup still scheduled for 2028.

The NRC track has not gone away. The NRC approved Oklo's Principal Design Criteria topical report for Aurora on an accelerated schedule, and Oklo says it will pursue an NRC commercial license for the Ohio powerhouses. In short: DOE authorization gets the first unit running and generating data; NRC licensing is still required for the commercial fleet.

Aurora-INL milestone tracker

Source: Oklo Q2 2026 company update and the Q2 2026 earnings call.

What Groves proved, and what it did not

Groves is not an Aurora unit. It is a low-power isotope test reactor in Lockhart, Texas, also authorized under the Reactor Pilot Program. It matters because it is the first reactor Oklo has actually built and started.

Oklo says Groves reached first criticality less than a year after groundbreaking, on a greenfield site on private land, with components and fuel commercially sourced or manufactured by Oklo. The Q2 deck dates first criticality to August 5, 2026 and DOE startup authorization to July 23, 2026, and says substantial construction took 229 days.

What Groves proves: Oklo can run the full DOE authorization sequence, procure nuclear-grade components and bring a reactor critical on schedule. What it does not prove: anything about a 75 MWe sodium power plant. Groves runs at low power and makes isotopes, not electricity. The gap between a critical test reactor and a grid-connected powerhouse is where most first-of-a-kind nuclear projects lose years.

Customers: Meta, Switch and the Ohio campus

Oklo's customer list is the main reason Aurora gets so much attention. Two agreements carry most of the weight.

Meta and the 1.2 GW Pike County campus

In January 2026, Oklo and Meta announced an agreement to support a 1.2 GW power campus on 206 acres in Pike County, Ohio. Meta can prepay for power and fund early project work, and Oklo says it will use the money to secure fuel and advance the first phase. Pre-construction starts in 2026, with the first phase targeted for as early as 2030 and full capacity by 2034.

The structure is worth reading closely. Oklo's project page says the powerhouses will sell power into the PJM market, with Meta paying Oklo for that power and drawing from the grid for its data centers. That is a grid-connected financial arrangement, not a reactor bolted to a server hall. The Oklo deal was one of three Meta nuclear agreements announced together, which support up to **6.6 GW** of new and existing capacity by 2035.

Switch and the 12 GW framework

In December 2024, Oklo and data-center operator Switch signed a non-binding Master Power Agreement to deploy **12 GW** of Aurora powerhouses through 2044. Binding contracts are expected only as individual project milestones are reached. Read it as a statement of intent and a pipeline, not as contracted revenue.

Other projects

Oklo is also developing an Aurora application at Eielson Air Force Base in Alaska for both electricity and heat, and it signed an MOU with Kiewit for engineering and procurement planning on the Ohio campus.

Fuel: the constraint that decides the fleet

Fast reactors need HALEU, uranium enriched between 5% and 20%, and almost none is commercially available in the US. Oklo's first core is covered: INL granted it access to 5 metric tons of HALEU recovered from EBR-II, under a cooperative agreement awarded in 2019, and DOE approved the conceptual design of the Aurora Fuel Fabrication Facility that will turn that material into fuel.

Everything after the first core depends on new supply. On the Q2 2026 call, Oklo described a three-part strategy of commercial HALEU, plutonium blending and recycling. Specifics include a Centrus letter of intent for HALEU for up to five powerhouses, with deliveries expected from 2029, and advanced negotiations for allocations from a 20-ton DOE plutonium tranche as a bridge fuel. Recycling, using pyroprocessing based on INL technology, is the long-term plan and is still at the planning stage.

The balance sheet behind Aurora

Oklo is pre-revenue in any meaningful sense and well funded. At the end of Q2 2026 it held **$3.0 billion** in cash and marketable securities, including **$1.9 billion** raised through at-the-market offerings in 2026.

Source: Oklo Q2 2026 company update, Q2 2026 results release and the Q2 2026 earnings call. Management said the higher guidance reflects first-of-a-kind costs at Aurora-INL, grid interconnection build-out and pulling procurement forward to protect the schedule.

The cash pile buys time, but it does not prove unit economics. Oklo has not published a firm cost per kilowatt or a levelized cost of electricity for Aurora, and the first-of-a-kind unit will almost certainly cost more than later ones.

Oklo Aurora vs TerraPower Natrium vs NuScale

Aurora is often lumped in with every small modular reactor. The designs and regulatory positions are very different.

Source: DOE on the Natrium construction permit, Federal Register notice of the NuScale US460 approval and the Aurora-INL DOE filing.

The comparison cuts both ways. NuScale and TerraPower hold NRC approvals that Oklo does not, but neither has a unit running. Oklo is the smallest of the three, which lowers the cost of each unit and raises the number of units needed for gigawatt-scale deals. A 1.2 GW campus is roughly 16 Aurora units at full rating. If fusion is on your comparison list too, Inside Deep Tech's Commonwealth Fusion Systems SPARC guide and Helion Energy guide cover the fusion timelines.

Inside Deep Tech's take

Inside Deep Tech's take: Oklo Aurora is a credible reactor concept with a real engineering heritage, a working authorization path and more cash than almost any advanced nuclear developer. Groves turned Oklo from a company that writes license applications into one that has started a reactor.

But the market prices Oklo as if Aurora were already a product. It is a first-of-a-kind sodium plant with a 2028 startup date, a first core that cannot be repeated, and customer agreements that are either prepayment mechanisms or non-binding frameworks. The checkpoint that matters is Aurora-INL generating power at its initial rating, followed by a binding contract with disclosed pricing.

Honest limits

First, nothing is generating. Aurora-INL is under construction, and startup is scheduled for 2028. Every revenue projection depends on that date holding, and first-of-a-kind nuclear schedules rarely do.

Second, the customer numbers are softer than the headlines. The Switch **12 GW** agreement is explicitly non-binding, and the Meta campus targets first power as early as 2030, not today.

Third, grid access is its own risk. On September 24, 2026, FERC denied Oklo's complaint against PJM in docket EL26-101, leaving a 750 MW mixed-technology Oklo project in Virginia, which pairs a small nuclear block with other generation, out of PJM's current interconnection study cycle. Oklo can resubmit in the next cycle or use PJM's expedited track, but the episode shows that a reactor permit does not guarantee a grid connection.

Fourth, fuel beyond the first core is unsecured at scale. The Centrus supply is a letter of intent, and plutonium allocations are still in negotiation.

Fifth, the cost is unknown. Oklo lost **$81.6 million** in the first half of 2026 and raised its 2026 capex guidance to **$400 million** to **$500 million**, but has not published an all-in cost for an Aurora unit.

FAQ

What to watch next

Four signals will tell you whether Aurora is on track. First, Oklo's Q3 2026 results and any update to the 2028 Aurora-INL startup date. Second, DOE approval of the final Documented Safety Analysis. Third, startup of the Aurora Fuel Fabrication Facility, planned for 2027. Fourth, a binding power contract with disclosed pricing from Meta, Switch or another buyer.

For the wider power picture, read Inside Deep Tech's small modular reactors guide, the behind-the-meter generation guide and the 800 VDC power distribution guide.

Small Modular Reactors: A Full Guide to Nuclear Power for AI Data Centers

Which SMR designs are licensed or under construction, the hyperscaler deals, and why most of this power arrives after 2030.

Inside Deep Tech