Oklo (OKLO) Q3 2026 earnings review
Oklo builds quickly, but its first plant may start later
Oklo's construction is moving quickly, but its first power plant now has a later start date. Steel is going up at Aurora-ID in Idaho, and chief executive Jacob DeWitte priced the Groves-1 test reactor at about $75 million. Quarterly spending on plants reached $154 million, funded by new shares sold at about $42 each, half the earlier price.
| Cash spent on plants and equipment | $154.0M $94.1M last quarter, by our math |
|---|---|
| Operating loss, excluding a land-sale gain | $82.4M $36.3M a year ago |
| Cash and securities | $3.13B after $402M of new share sales in the quarter |
| Full-year cash outlook | Unchanged $120–150M for operations, $400–500M for plants |
⚖️ Verdict: 🔴 Bearish
The story got worse because the filing moved the first plant to a 2028–2029 target and DeWitte accepted that reads as a delay. Each year of delay means more share sales at lower prices. The good news: Ohio's fuel moved from a letter of intent to a signed supply agreement, with Meta paying Centrus directly.
The question now is whether 2029 is a buffer or the real date. If the reactor components go in on schedule, the extra year is insurance; if the cost estimate arrives with another date, the delay is structural. The annual report in March will tell.
🐂 Bull Case
Oklo's First Reactor Cost About $75 Million
Groves-1 is a small test reactor in Texas that Oklo built to learn isotope production. Chief executive Jacob DeWitte priced it this quarter: "The build cost was approximately $65 million, plus approximately $10 million for commercially sourced fuel from Framatome."
The same team is now building Aurora-ID, the 75-megawatt plant in Idaho. "By November 4th, structural steel was going up," DeWitte said; at the fuel workshop, "Facility equipment installation is underway."
This matters because Oklo's plan rests on building faster and cheaper than the industry has. But a low-power test unit says little about a full power plant's cost.
What to watch: the total cost of Aurora-ID, which finance chief Craig Bealmear said last quarter would come once it firmed up. That figure shows whether the Groves-1 discipline carries over.
Meta's Money Now Pays Centrus for Oklo's Fuel
Oklo's reactors need HALEU, a more enriched uranium that few suppliers make. Until this quarter Oklo held only a non-binding letter of intent with Centrus, the Ohio enricher.
On September 30 Oklo signed a Centrus supply agreement, "subject to certain conditions", for the Ohio plants planned with Meta. On November 2 Meta's first payment left escrow and went "directly to Centrus as a credit towards fuel orders".
A customer paying the supplier commits harder than a letter of intent. But the first plant runs on material Oklo already holds; Centrus and plutonium decide whether Oklo can fuel the next ones.
What to watch: the Centrus conditions clearing, and the plutonium timeline, which DeWitte expects to "become a lot clearer in the next few months". Either gives Ohio's fuel a date.
Oklo's Acquired Engineers Now Work on Aurora-ID
In June Oklo bought two small engineering firms for $33.4 million: ARMEC, a precision machining shop, and Creative Engineers, which designs sodium-cooled equipment. DeWitte said ARMEC is already "supporting reactor module development at Aurora-ID".
The two supplied all of Oklo's $1.9 million of revenue this quarter, from work for outside customers. They also lost $1.1 million between them, after Oklo described both last quarter as cash-generating. Oklo also used AI to adapt Aurora models for Groves in under a week, which DeWitte said "compressed months of modeling work into days".
What to watch: any figure for savings on components Oklo now makes itself, and whether the two units earn a profit in the fourth quarter. A savings number matters more than the small sales side.
🐻 Bear Case
Oklo's First Plant May Start a Year Later
Aurora-ID is the plant everything else waits on: Oklo's first power revenue and the reference design for every plant after it. Oklo kept 2028 as the start date for a year. This quarter's filing changes that to "deploying our first powerhouse in 2028-2029".
Asked whether that meant a modest delay to 2029, DeWitte agreed and cited Groves-1 lessons about testing a fully assembled plant. "Electrical noise management is a huge thing in nuclear plants," he said; Oklo wants "that buffer in there because time is going to take time".
The slip matters because every year without power revenue is another year paid for by selling shares. The Ohio campus also leans on Aurora-ID's record for its NRC licence. The plant's total cost is still unpublished. DeWitte closed the call saying the site is "on pace to set a world record for fastest privately built fast reactor". The filing added a year in the same week.
What to watch: the Documented Safety Analysis, which DeWitte does not expect to submit until 2028. The start-date wording in March's annual report will show whether 2029 is a buffer or the new plan.
Oklo Sells Stock at $42 to Fund Spending
Oklo has almost no revenue, so every dollar it spends on plants comes from investors. This quarter it spent $154 million on plants and equipment plus $45 million on operations, by our math from the year-to-date cash flows.
To pay for that, Oklo sold 9.5 million new shares through its at-the-market programme. They brought in $402 million, about $42 a share after fees. That is under half the $96.95 from the programme completed in March. In September it opened a third programme. The first $87 million went out at $39.30 a share.
- Shares outstanding: 195 million, up 25% from a year ago
Bealmear said Oklo "proactively and intentionally raised capital" to commit to fuel and equipment early. But the lower the price per new share, the more of the company each plant costs existing holders. The same $402 million took more than twice as many shares as under the March programme.
What to watch: fourth-quarter cash used in operations, in March's annual report. The full-year plan Bealmear reaffirmed needs it to fall well below the third-quarter level, so a repeat would put the year above the range.
Oklo Still Has No Binding Power Contract
Oklo plans to own its plants and sell power under long-term contracts, so a signed power purchase agreement would be the first sign of revenue. It still has none. Asked whether the Meta agreement for Ohio obliges Oklo to build, DeWitte said "it is not binding as is to build the reactor".
Bealmear added there are "some binding elements to it": Meta's capital reserves capacity, and Oklo must hit milestones, starting with fuel. PJM, the regional grid operator, withdrew one of Oklo's Virginia connection applications, and the federal energy regulator dismissed Oklo's complaint. The Ohio study is "going to take place over the next 24 months", Bealmear said.
- Customer prepayment on Oklo's books: $25 million since 2024; the buyer's right of first refusal ends around February 2027 by our math
- Capacity in PJM's 2026 study: 900 megawatts
A two-year study means Ohio cannot connect before late 2028. The agreement, as DeWitte describes it, does not yet oblige either side to a power sale.
What to watch: a signed power purchase agreement with Meta, with a price and a delivery date. That would be the first revenue line anyone can model.
Risks this quarter didn't answer
Several standing questions got no answer in the filing, the slides or the call. Each has a specific disclosure that would settle it.
- Aurora-ID total cost: still unpublished; Bealmear said last quarter a number would come once it firmed up
- Surplus plutonium: no volume or date; DeWitte expects the timeline to "become a lot clearer in the next few months"
- Ohio first phase: the 150-megawatt-by-2030 target from the annual report has not reappeared in three quarterly filings
- Eielson Air Force Base: ground investigations due in summer 2026; no update
- First isotope revenue: now a target to "complete our first contracted isotope delivery" within twelve months, with no dollar figure
- Policy: asked about the election and state limits on data centres, Bealmear said customers have not pulled back; nothing quantified
👓 Other Themes
NRC Will Accept Oklo's DOE Safety Evidence
On October 26 the Nuclear Regulatory Commission finalised a rule for reactors the Department of Energy has already authorised, which is Aurora-ID's route. A licence application can now draw on their safety analyses and operating data. Oklo said this "could reduce duplicative technical reviews and improve licensing efficiency". The NRC keeps full review authority, and no reactor has yet made that conversion.
Groves-2 and a Parts Deal for 27 Reactors
Two new programmes carry no revenue yet. Groves-2 Forge, a larger isotope plant that would irradiate, separate and package products, won a spot in the DOE's Nuclear Energy Launch Pad. Oklo has bought its fuel but is "still finalizing the site location". Oklo also reserved capacity for 27 reactor modules at U.S. Forged Rings through 2035, for what Bealmear called "an initial small investment".
💲 Other KPIs
Interest on Oklo's cash pile rose for a fifth straight quarter, to $27.9 million from $7.1 million a year ago. It now covers about a third of the operating loss before one-offs, against a fifth a year ago, because share sales enlarged the balance sheet. By our math the money earns about 3.6% a year.
Share-based pay, a non-cash cost, doubled from a year ago to $19.2 million and made up 23% of operating expenses. It has stayed near that level every quarter this year. Oklo still has $187 million of granted awards to expense over the next three and a half years.
Running costs more than doubled from a year ago, to $84.3 million. Research and development rose 179% and overhead 95%. Pay for about 240 more employees explains most of it, along with outside engineering consultants. Oklo expenses all research and development as it occurs, the filing says.
A new line on the balance sheet: Oklo has paid $68.7 million in advance for enriched uranium, against nothing at the end of 2025. A further $58.0 million sat in restricted cash at quarter-end as a letter of credit for another uranium purchase. The bank released it on October 13.
🔮 Guidance
Unchanged. Bealmear said Oklo still expects to use $120–150 million of cash in operations this year. That leaves only $9–39 million for the fourth quarter, by our math. The third quarter alone used $45 million, so a repeat would put the year at about $156 million, above the range. Oklo also exceeded last year's range on a reported basis, so the plan assumes a slowdown or a working-capital swing.
Unchanged. Oklo still plans to spend $400–500 million on plants and equipment this year, mainly Aurora-ID, the Idaho fuel workshop and fuel purchases. By our math, that leaves $119–219 million for the fourth quarter, in line with the $154 million spent in the third. Spending so far this year is already more than eight times last year's total.
Pushed back. The quarterly filing now targets "deploying our first powerhouse in 2028-2029", where the previous one said 2028. Two years ago the target was late 2027. DeWitte accepted on the call that this reads as a modest delay and tied it to commissioning lessons from Groves-1. No cost estimate accompanies the new date.
❓ Key Questions
What will Aurora-ID cost in total?
Oklo has now published a cost for Groves-1 but not for the 75-megawatt plant that carries its revenue plan. Bealmear said last quarter a figure would come once it firmed up in 2026, which leaves the fourth-quarter report.
What did Oklo receive for the land it sold?
The quarter's $11.4 million gain came from selling land with a book value of about $4.6 million, by our math. The cash flow statement lists the sale as non-cash, and the filing names neither the buyer nor the consideration.
What conditions remain on the Centrus supply agreement?
The September 30 agreement is "subject to certain conditions" that the filing does not spell out. The first Meta payment to Centrus has no amount attached. Both decide whether the Ohio campus has fuel on a date.
What happens to the $25 million prepayment in February?
A data-centre customer paid $25 million in 2024 for a right of first refusal on Oklo's power. By our math that right ends around February 2027 with no plant running, and the filing does not say whether the money stays with Oklo.
Is 150 megawatts in Ohio by 2030 still the plan?
The annual report targeted a first phase of about 150 megawatts around 2030. Three quarterly filings since then refer only to an initial phase, and the grid study now runs two years. Both the size and the date need restating.