Eleven companies that mine uranium, enrich it, or design and build reactors reported second-quarter results between 31 July and 7 August. Read together, the prints separate the group into three businesses with almost nothing in common except a shared narrative about electricity for artificial-intelligence data centres — and the shares of all eleven bottomed on the same day, 29 July, and rallied through the reporting week regardless of what they said.

The developers are financed by share sales, not customers

Oklo, a developer of 15-to-75-megawatt "Aurora" fission powerhouses with 215 employees, recorded $1.21m of revenue in the quarter against a $73.2m operating loss. It ended June with $3bn of cash, having raised $1.9bn during 2026 through an at-the-market programme — a drip-feed sale of new shares — and it raised its own spending guidance twice, to $120-150m of operating cash use (from $80-100m) and $400-500m of capital expenditure (from $350-450m). The engineering is moving: its Groves isotope reactor reached first criticality in under eleven months from groundbreaking. The revenue is not. Oklo's first Idaho plant is slated for commissioning by late 2027 or early 2028, with commercial deployments as early as 2029 and Google eyeing 2030, and its agreements with Meta, Switch, Equinix and Wyoming Hyperscale are non-binding letters of intent or frameworks — Equinix's 500-megawatt commitment, the first of its kind, is a procurement agreement, not a plant under construction. Consensus puts Oklo's 2030 revenue at $286m, in a range spanning $72m to $514m, with a net loss still forecast that year. At 6,619x trailing and 4,317x forward sales — the only applicable lens for a pre-revenue company — the price embeds the 2030s in full.

NuScale, whose 77-megawatt module is the only small reactor design certified by the Nuclear Regulatory Commission (NRC), reported revenue of $75,000 against $8.05m a year earlier and a $64.0m operating loss. It holds $1.9bn of cash, up $900m in three months, and management described talks with the Tennessee Valley Authority and its partner ENTRA1 as active but with no signed power-purchase agreement; Romania's RoPower notice to proceed is roughly twelve months away. At 0.75x book value it is the only member of the group trading below its stated equity. NANO Nuclear, a $1.03bn company with 36 employees and four unbuilt reactor concepts, has never recorded revenue and is not forecast to be profitable in 2030. Lightbridge, a 13-person developer of metallic reactor fuel, raised $44.4m through its own share programme against $8.3m of first-half cash burn; its supply memorandum with Centrus carries no price, quantity or exclusivity.

For these four, the business CONFIRMS the year-long de-rating.

The fuel cycle: better contracts, worse earnings

Cameco, which mines and mills uranium and converts it into reactor fuel, is the mirror image. Reported revenue fell 7.2% to $814.1m, gross margin narrowed from 29.3% to 21.1% and net income fell 92.1% to $25.2m — mostly the absence of a one-off Westinghouse payment from the Czech Dukovany project booked in 2025, plus currency costs and spring disruptions at Key Lake, McArthur River and Cigar Lake. Production guidance was unchanged at 19.5-21.5m pounds. The contract book improved: deliveries averaging more than 28m pounds a year, with market-related floors in the high-$70s. The long-term price indicator reached $93 a pound in March, its highest since 2008, with spot at $86.48 on 8 August, and Kazatomprom cut 2026 output about 10%, roughly 5% of world primary supply. Cameco also disclosed that Westinghouse, 49% owned by Cameco and 51% by Brookfield, has filed a confidential draft registration for a US listing. Valuation CONTRADICTS the bounce: the trailing multiple went from 106.8x in May to 168.8x now, and price-to-gross-profit from 38.6x to 62.7x, as the market value fell from $50.3bn to $43.2bn. The shares got dearer, not cheaper. Forward earnings put it at 60.5x.

Centrus, the only US commercial enricher, grew revenue 14.0% to $176.1m but saw operating income fall 69% as separative-work volumes dropped 23%, prices rose 3% and unit costs rose 13%. Backlog reached $4.5bn extending to 2040, and a $900m Department of Energy task order moves its Piketon cascade from demonstration to commercial operation. Commercial production starts 2029; X-energy deliveries begin 2030. At 75.0x trailing and 73.6x forward earnings, the market prices no growth into next year — consensus 2026 earnings per share of $2.56 are down 42.9%. INCONCLUSIVE on the business, unsupportive on price. Among the miners, Energy Fuels grew revenue six-fold to $25.1m at a 57.4% gross margin yet lost $33.4m and is not forecast profitable until 2027; Uranium Energy booked no revenue at all in its April quarter against a $40.8m operating loss.

BWX Technologies is the outlier

BWXT, which builds reactors, fuel and missile launch tubes for the US naval propulsion programme and components for commercial utilities, grew revenue 18.0% to $901.6m — its fourth consecutive quarter above 18% — with net income up 13.6%, backlog of $8.4bn up 40%, a book-to-bill ratio of 1.7x, and guidance raised on revenue, earnings and free cash flow. Commercial operations revenue rose 72%. Management expects at least one new-build equipment order by year-end. Its multiple compressed from 55.0x trailing earnings in May to 43.7x, and 35.8x forward, while 2027 consensus earnings rose to $5.29 a share. The business CONTRADICTS its tape, and the valuation confirms the contradiction. Yet BWXT is the only member down over the past 30 days: shares slipped after hours on 3 August despite an earnings beat, on a $4m revenue shortfall and higher planned capital spending.

The tape agrees with none of this yet. Every name has traded with its 50-day average below its 200-day since early July at the most negative reading the measure produces, and BWXT and Cameco stepped down through June — before the prints — so the break is price, not a corporate event. Ninety-day returns run from -12% to -32%. It is also not a nuclear-specific reset: AI-power and data-centre names fell together in late July. And the dates the companies themselves disclose sit outside this decade — Centrus 2029, X-energy 2030, and construction on the reactors backed by the Department of Energy's conditional $17.5bn loan package beginning by 2030, for service in 2033-34.

The setup

Where it stands — Eleven nuclear names rallied off a 29 July low through earnings week; only BWXT's results and multiple moved in opposite directions.

Would confirm — BWXT books the new-build reactor equipment order management promised by year-end, and Cameco's next quarter restores gross margin above 29%.

Would invalidate — Oklo or NuScale launches another equity raise before signing a definitive power-purchase agreement, or uranium term price slips below $86.

Watch next — Third-quarter results in late October and early November; the Westinghouse listing terms, still undetermined; the Department of Energy's five site selections.

Valuation — BWXT 43.7x trailing and 35.8x forward earnings, against 55.0x in May; Cameco 168.8x trailing, 60.5x forward, versus 106.8x in May.