Oklo is a pre-revenue developer of 15–75 MWe sodium-cooled fast reactors (Aurora), pursuing a build-own-operate model — selling power under long-term PPAs rather than licensing reactor designs — with adjacent fuel-recycling and radioisotope businesses. Aurora-INL is the only commercial advanced reactor approved to proceed under DOE's authorization route, a separate track from full NRC commercial licensing, with first power targeted 2028, a 5 MT HALEU award, a $2.5B debt-free treasury, and announced relationships spanning Meta (1.2 GW Ohio campus), Switch (12 GW), Equinix and the U.S. Air Force. The demand backdrop is real: PJM projects a 50–60 GW shortfall, and hyperscalers have demonstrated ~$98/MWh willingness-to-pay for firm clean power. The problem is not the company; it is the price.
Probability-weighted fair value is $17.65/share (range $14–22): a cash floor of ~$12–14 plus a real option on first-of-a-kind (FOAK) costs breaking toward $4,000/kW. The reverse-DCF implies that today's ~$7.3B enterprise value requires roughly 9–15 GW of deployed capacity at ≤$4,000/kW with mature project financing near 8.5% — that is, the market is capitalising a cost outcome near the company's previously quantified but no longer emphasised aspiration, majority-to-full conversion of a non-binding customer book, and a project-finance market not yet proven for this asset class. Western FOAK datapoints (Vogtle ~$15,000/kW; NuScale's CFPP cancelled at ~$20,100/kW) sit 2.5–5x above that requirement, and announced nuclear capacity has historically converted to operating megawatts at under 10%. The market may be directionally right on demand; the price assigns too much certainty to cost, conversion and financing outcomes that remain unproven. We rate it Avoid at $56.44, watchlist with triggers.
I am wrong if the Aurora-INL build evidences all-in cost ≤ ~$6,000/kW and at least one multi-hundred-MW binding take-or-pay PPA with a creditworthy counterparty prints at ≥$95/MWh — because then the cost-breakthrough branch is demonstrated, project finance follows bankable contracts, and the $7.3B program value stops being narrative and starts being a discounted order book. Both legs are observable; neither has occurred.
1 · Investment thesis summary
Why it's interesting. Aurora-INL is the only commercial advanced reactor approved to proceed under DOE's authorization route, with first power targeted 2028, a 5 MT HALEU award, a $2.5B debt-free treasury, and announced relationships spanning Meta, Switch, Equinix and the U.S. Air Force (10-Q MD&A p.19–20). The demand backdrop is real: PJM projects a 50–60 GW shortfall, and hyperscalers have demonstrated ~$98/MWh willingness-to-pay for firm clean power (Crane/Microsoft PPA; via Utility Dive, 9/24).
What it's worth. The base-case DCF on my cost glide is negative: on my assumptions, shareholder-value-maximising scale deployment would not proceed without a major cost or financing breakthrough — which is why the equity is best read as cash-plus-option, not a claim on a 2.7 GW program. Probability-weighted fair value is $17.65 (range $14–22) — a cash floor of ~$12–14 plus a real option on FOAK costs breaking toward $4,000/kW.
What I'd pay / entry triggers. ≤$20 on price alone (the asymmetry flips long at the probability-weighted value); or pay up on information: a credible FOAK cost ≤$6,000/kW, a first exhibit-filed binding take-or-pay PPA, or a ≤9% project-finance print. Any two together: constructive at market.
What could go right (top 2). (1) The INL build delivers a credible cost print near aspiration — the option branch becomes the base case and fair value roughly doubles before the market finishes repricing. (2) A power-short decade prints $120+/MWh PPAs — my soft flank; bull value moves to ~$48–55. What could go wrong (top 2). (1) The FOAK cost print lands where Western FOAK projects have historically landed, and the never-binding pipeline loses its specifics the way the 14 GW figure, the LCOE and the Liberty Energy partnership already did — price converges on the $8–14 cash floor. (2) For any bearish expression: a squeeze — 19.7% short interest, ~95% IV, and a policy regime that has so far skewed favorably.
For holders: this is a level to harvest, not to add — at ~95% IV, covered calls against a reduced position are paid handsomely; collars are nearly self-funding. For the bearish view: no naked short — defined-risk structures only (put spreads financed by call sales), sized to survive ±50% single-quarter moves, with patience as the core position: the thesis has no forcing catalyst and the company has $2.5B of staying power. The asymmetric trade is conditional: if IV ever compresses below ~50% into a binary window (COLA decision, first power), long volatility is the cheap side. Invalidation in the same breath: both falsifier legs firing flips this constructive regardless of price. This describes how the thesis maps to instruments; it is not advice.
| Metric | Current | Target / fair | Note |
|---|---|---|---|
| Price / share | $56.44 | $17.65 ($14–22) | 52-wk range $44.88–$193.84 |
| Market cap / EV | $9.8B / $7.3B | — | EV nets $2,537M cash & securities (3/31/26) |
| Revenue | $0 | first revenue 2027E | zero in every reported period |
| EV/EBITDA / P/E / FCF yield | n.m. | EBITDA+ 2035E | UFCF negative every year through 2040E |
| Core operating burn † | ~$155M/yr | — | ex-interest, timing-normalised; reported CFO ~$72M/yr |
| Free cash burn (commitment) † | ~$286M/yr | — | incl. AP-parked capex |
| Net debt / EBITDA | net cash $2.5B | — | no debt; 8.95% lease IBR is sole Kd datapoint |
| Share count | 173.9M | ~230–290M by 2035E | +16.6% FY25, +8.3% Q1-26; $1.0B ATM live |
| Cumulative funding need to 2040E | — | ~$18.6B | cash exhausts ~2030 on base deployment |
| WACC (Ke) | 12.08% | 8.5% terminal/project | β 1.80 selected (regression R²=0.05) |
| Short interest / IV30 | 19.7% / ~95% | — | most-shorted utilities name; straddle ~22%/mo |
| Market-implied P(upside) | ≥93% | my P: 20% | price exceeds prob-1.0 bull blend on my values |
2 · Business quality & moat
Classify before comping: Oklo is an execution archetype — a capital-project developer maturing into an IPP — wrapped in platform language ("vertically integrated platform… recurring revenue," 1Q26 deck). The implied margin ceiling is power economics: ~40–45% consolidated EBITDA at maturity, ~25–30% operating margin after ~$21/MWh of depreciation. The moat is real but regulatory-positional: the only DOE-RPP commercial pathway, INL site rights, fuel awards, a PDC topical report approved — a 2–4 year head start with a political half-life, eroding by design as Part 53/57 licensing modernisation (which the company itself celebrates) lowers the barrier for everyone. The core technology is government-lineage (EBR-II); the scarce inputs (HALEU, plutonium, used fuel) are allocated by government, not owned. Switching costs arrive only with signed PPAs; the cost advantage is entirely unproven — no filing discloses $/kW, $/MWh, or LCOE. The strongest structural claim is efficient scale in the underserved 15–75 MWe co-location niche.
Governance and incentives. Insiders own 12.70% — principally the DeWittes' 12.16% (married co-founder CEO/COO team) — which creates clear economic alignment but concentrates key-person risk in one household. The trading record points the other way: both founders sell monthly under a 10b5-1 plan adopted March 31, 2025; the CFO routinely exercises and sells low-strike options ($2.11 weighted-average plan strike); and there has not been a single open-market insider purchase on record. None of that is unusual for high-beta growth stories — but it is hard to reconcile with a narrative that the stock is deeply undervalued. All NEO equity is time-based rather than milestone-based — nothing pays specifically for first power, license grant, or PPA conversion — and the CEO received a $250K bonus for signing a non-binding LOI (which the proxy reveals to be Equinix, unmasking the anonymous $25M ROFR counterparty): cash compensation tied to a marketing milestone, not a contracted revenue commitment. A director's affiliated firm earns $250K/quarter in advisory fees plus contingent transaction fees from this serial equity issuer — permissible and disclosed, but structurally tilted toward continued capital-raising. Sam Altman's 2025 board exit removed both the AI halo and a genuine customer-neutrality conflict; Meta arrived after. Netting it out: the ownership line-items screen as aligned, but the revealed behaviour — ongoing selling, time-based equity, issuance-geared advisory economics — all sits on the side of raising and monetising equity rather than maximising per-share value. A credible governance inflection would be: a first open-market insider purchase, an ATM pause while cash is ample, and a shift of NEO equity toward objective project milestones.
Multiple informed or economically motivated constituencies are supplying stock or positioned skeptically — the company ($2.5B issued in 7 months, new $1.0B ATM; 8-K 5/13/26), the founders and CFO (monthly 10b5-1 sales; SEC Form 4s), and 19.7% of the float (short interest). These are different kinds of signal — treasury management, diversification, and positioning — but none of them is the behaviour of constituencies who consider the stock cheap, and jointly they put the burden of proof on the bull side of the reverse-DCF.
3 · The deployment model — there is no revenue to model
Zero revenue in every reported period. The model is cohorts × MWe × capacity factor × PPA price: INL 75 MW (2029, one-year slip applied), Eielson 15 MW (2030), Ohio/Meta 1.2 GW phased 2031–36, other commercial reaching 1.4 GW by 2040 — 2,715 MW cumulative by 2040, ~20% of the announced ~14 GW book, at $95/MWh (2028$, +2%/yr; anchored to Crane ~$98 and NuScale CFPP's $89 post-subsidy). Base revenue: $61M (2029) → $1.1B (2035) → $2.8B (2040).
| ($M) | 2029E | 2031E | 2033E | 2035E | 2037E | 2040E |
|---|---|---|---|---|---|---|
| Revenue | 61 | 204 | 521 | 1,108 | 1,801 | 2,814 |
| EBITDA | (391) | (412) | (286) | 51 | 536 | 1,195 |
| EBIT | (420) | (484) | (446) | (233) | 129 | 650 |
| UFCF | (1,520) | (1,920) | (2,301) | (2,780) | (791) | (789) |
| Cumulative funding gap vs cash | +176 | (2,410) | (6,585) | (11,733) | (15,813) | (18,571) |
Aurora-Ohio/Meta peaks at ~79% of revenue (2033) under an agreement never exhibit-filed; 2035 revenue is $1,108M with Ohio, ~$334M without. Deployment scale dominates price 6:1 in sensitivity — the thesis is won on units, not $/MWh. EBITDA turns positive in 2035, EBIT in 2037, and UFCF never inside the window.
4 · Quality of earnings — the numbers behind the numbers
Oklo publishes no non-GAAP measures, classifies the $25M Equinix prepayment as financing (never flattering CFO), and carries a clean Deloitte opinion — reporting integrity: high. The economic distortions are structural: interest income on the war chest masks 41.6% of the operating loss and converts a ~$155M core operating burn into a reported ~$72M CFO; management's "$80–100M cash burn" guide presents ~40% of the GAAP cost base. SBC fails the buyback-symmetry test in the no-buyback direction — the FY25 charge was $41.8M while the dilution-settled economic transfer ran an estimated $150–220M (options exercised at a $1.46 average strike into a $55–70 stock). A comparison of successive filings and decks shows a consistent pattern: the first-power target slipped twice (2026/27 → late-2027 → 2028); the 14 GW pipeline, the $40–90/MWh LCOE, and the Liberty Energy partnership were each quantified or featured once, then silently retired. Each of the specific, checkable forward numbers the company once volunteered has since been replaced by an adjective. The depreciation clock has not started ($89M of $95.6M PP&E sits in non-depreciating CIP) and the single largest accounting election ahead — the nuclear useful life — remains unmade; a 60-year election would flatter early margins ~6–7 points versus my 40-year base.
5 · Capital structure — the dilution treadmill
No debt; 2.4% liabilities/assets; the leverage question inverts to: how much of the $18.6B lands on the share count? The record: $2.9B raised in 27 months, share count 69.2M → 173.9M, the $1.5B December ATM fully exhausted at $88–97, and a new $1.0B ATM signed May 13, 2026 with the stock ~40% below January's $96.95 average print. Funding the 2035 gap 100% with equity at $40 implies 2.8x today's share count (breaching the 500M authorisation); even 25% equity at $100 implies +27% before SBC. The escape requires project debt that can only exist against binding PPAs — of which there are zero — plus federal credit (the Constellation $1B loan is the precedent) and customer prepayments. Flagged for the record: a $17.1M venture book of SAFEs in unnamed private companies (embryonic vendor-financing watch item), and a project-SubCo chassis (Oklo Power INL/SODI/Recycling TN LLCs) where off-balance-sheet complexity will accumulate — read the first project financing's consolidation and guarantee footnotes harder than its press release.
6 · Valuation
WACC. Ke = 4.47% (10Y) + 1.80 × 4.23% (Damodaran implied ERP) = 12.08%; β selected on reasoning because the regression is noise (R² = 0.05; published betas 0.77–3.00). No size premium at $10B. Execution risk is priced once: probability-weighted cash flows at 12.08% with no CSP; single-path management-plan flows at 17.08% (+5% CSP, disciplined by ~95% IV). Terminal/project WACC ~8.5%. The bridge is explicit: I use corporate Ke (12.08%) for development-stage equity cash flows, and 8.5% only for mature, contracted project economics and terminal value. With D/V = 0, current WACC = Ke. Mechanically this is among the most rate-sensitive equities conceivable; behaviourally the price is set by flows ($97 → $56 with the 10Y unchanged). The rate binds fair value, not the tape — that gap is the thesis.
DCF and reverse-DCF. The base-case DCF is negative (EV −$7.1B perpetuity / −$6.6B exit): $8.7B of discounted construction-era outflows against $1.6–2.1B of terminal PV. Properly read: on my assumptions, shareholder-value-maximising scale deployment would not proceed without a major cost or financing breakthrough — so the equity is best understood as $2.5B of treasury assets plus a compound real option on costs breaking toward $4,000/kW. The reverse-DCF grids price the option exactly: a kW earns ~$520/yr after tax at maturity, worth $3,925 at 12.08% or $5,423 at 8.5%, against capex. The market's ~$7.3B program value requires ~9–15 GW at ≤$4,000/kW at 8.5% — no corner of the 12% grid reaches it, even 20 GW at $3,500/kW ($3.8B). Capex/kW is the dominant valuation variable: $6,500 → $4,000 swings 15 GW program NPV by ~$17B; rate and terminal-growth sensitivities are second-order by an order of magnitude.
At $56.44 you are not buying $56 of discounted cash flow. You are buying $14.60 of treasury assets plus $41.84 — 74% of the price — of faith in the conjunction {costs ≤$4,000/kW} × {12+ GW converts} × {cheap leverage} × {limited dilution}. On diluted counts, even the bull corners land at $29–48/share — below today's price.
Brighthedge — Equity researchComps and transactions. Oklo trades at $519M EV per pipeline-GW vs NuScale's $368M — a 41% premium to the only peer holding an NRC-certified design and an advancing sovereign project; the BOO archetype justifies some premium, but Oklo's numerator contains zero binding capacity. Constellation's 14.3x EV/EBITDA applied to Oklo's 2040E EBITDA discounts to ~$3.3B of EV today — under half the current EV before any of the $18.6B need. And the anchor that stings: Brookfield + Cameco paid $7.9B EV for Westinghouse — the world's largest operating nuclear services franchise — essentially the figure the market assigns to Oklo's pre-revenue plan. The same asset has traded between ~$850M (de-SPAC, 2024) and ~$33B (the $193.84 high) with the same fundamental status throughout.
| Marker | Value | vs $56.44 |
|---|---|---|
| Cash floor | $12–14 | −77% |
| Probability-weighted FV | $14–22 | −69% |
| Operating-anchor comp (CEG 14.3x on 2040E) | ~$20 | −65% |
| Bull — 9–15 GW @ $4k/kW, 8.5% | $29–40 | −39% |
| Current price | $56.44 | — |
| Hyper-bull / $120-MWh branch | $48–62 | ±0% |
| 52-week range | $44.88–$193.84 | — |
7 · Scenarios & probability weighting
| Parameter | Bear | Base | Bull | Hyper-bull |
|---|---|---|---|---|
| 2040 deployment / achieved capex | ~0 GW / >$15k | ~2.7 GW / $5–6.5k | 9 GW / $4k | 20 GW / $3.5k |
| Ohio converts / financing | no / equity only | partial / blended | yes / 8.5% | yes+ / 8% |
| Terminal EBITDA margin / exit | n.m. | 42% / 7x | 45% / 9x | 48% / 10x |
| Fair value / share | $8 | $14 | $35 | $60 |
| Probability (ours) | 30% | 50% | 15% | 5% |
PW fair value $17.65. At $56.44, the price cannot be reconciled with my scenario grid: it exceeds my probability-one upside blend ($41.25), so no probability mixture of my values reproduces it. Even treating the $60 hyper-bull as the market's only upside state, the implied probability is roughly 93%, against my 20%. The distribution itself is lopsided by construction — base sits only $6 above bear because at base-case costs a kW adds nothing below the $5,400/kW value threshold — so for the price to be right my scenario values must be wrong, not my probabilities; that is what the sceptic's pass tests. Reporting-risk asymmetry: ±30–50% single-quarter moves on one headline are normal here ($44.88–$193.84 in twelve months).
8 · The sceptic's pass — the strongest case against this report
What the adversary wins. Demand is realer than my bear sketch: PJM's 50–60 GW shortfall is utility math; electrification and reshoring stand independent of AI; and hyperscalers have demonstrated the firm-clean premium with signed money (Crane ~$98/MWh). The free-option pipeline is stickier than a pure bullwhip read implies, because power-queue positions are harder to re-enter than chip orders. My $95/MWh could be $120+ in a genuinely short decade — that moves bull value to ~$48–55 and the implied probability from ~93% toward ~75%: still rich, no longer absurd. Operationally this team executes what it controls (Groves built in 229 days; DOE milestones on schedule). And the regime concession is total at the expression level: 19.7% short interest, ~95% IV, and a policy environment that has so far skewed favorably make aggressive bearish positioning unsound even where the bearish valuation is sound.
The conversion base rate: the 2008 renaissance turned ~30 COLA applications into roughly two builds, one abandoned with ~$9B written off — under 10% historically, 0% currently binding here. The analogy is imperfect — COLA applications and commercial LOIs are different instruments — but both show announced nuclear capacity converting poorly into operating megawatts. The cost prior: every Western FOAK datapoint sits 2.5–5x above the price's requirement. The dilution treadmill. The de-quantification pattern. Insiders selling monthly while the company sells stock. The re-point: the fragility is not in the accounting (clean) nor primarily in demand (real) — it is that the price pays roughly 93% for an outcome history funds at well under 10%, a gap closable only by two observable prints that have not occurred.
9 · Entry triggers, catalysts, risks & monitoring
Triggers: (1) price ≤$20 — the asymmetry flips long at the probability-weighted value; (2) credible FOAK cost ≤$6,000/kW — pay up on information; (3) first exhibit-filed binding take-or-pay PPA ≥100 MW; (4) project-finance/federal-credit print ≤9%; (5) first code-P insider buy plus an ATM pause. Any two together: constructive at market. Catalysts: Groves criticality (~July 2026), opening the path to the company's first revenue — possible late 2026, booked 2027E in my model, COLA acceptance for review (2026–27), INL first power (2028–29 — the cost print arrives with it, cutting either way), Meta conversion, rolling federal awards.
Risks, ranked (probability × impact): FOAK cost landing where Western FOAK projects have historically landed; pipeline non-conversion/Meta walk; dilution treadmill; policy reversal or an industry accident anywhere; second COLA failure; fuel supply (HALEU $24–33k/kg, capacity nascent); key-person concentration; and — against the Avoid itself — squeeze risk.
Nothing here is existential this decade ($2.5B, no debt, cancellable cost base); the high-probability cluster collapses the valuation gap to the $8–14 floor, not the company. Existential risk emerges only past ~2030 if cash exhausts mid-build — which is why financing prints outrank quarterly losses. The equity can lose two-thirds of its value without the business breaking; that is precisely why the stance is Avoid-on-price, watchlist with triggers, and the expression harvests volatility rather than betting on impairment.
Monitoring (footnotes before prices): the ROFR resolution at its ~Feb 2027 lapse; CIP-vs-progress each 10-Q (the only real-time FOAK cost meter); the useful-life election in the first post-COD 10-K (read it before the EPS); Pike County language drift, read against the company's established pattern of dropping specific figures; 424B5 cadence and prices; Form 4 regime (plan changes, acceleration, first purchase); IV regime (below ~50% into a binary window = the cheap-hedge asymmetry); the 2027 proxy (any move to milestone-vesting equity would be a real governance signal).