BWX Technologies, the sole-source supplier of reactors and nuclear fuel to the US Navy's propulsion program, told investors on 3 August that its order book had reached $8.4bn and raised every line of its 2026 guidance. The shares have fallen 7.9% since, closing at $160.13 on 20 August, their lowest in a year.

Two weeks before that report, the companies that supply the atoms rather than the hardware had already stopped falling. Cameco, the Saskatoon uranium miner that also owns 49% of reactor designer Westinghouse, put in its low at $84.85 on 20 July and has recovered 12.9%. Centrus Energy, the only US-owned commercial uranium enricher and the only US-licensed producer of high-assay low-enriched uranium (HALEU), bottomed at $147.07 on 16 July and is up 19.9%. The habit of treating fuel, enrichment and reactor hardware as one "nuclear powers the data center" trade has broken down, and it has broken down in the direction that punishes the best operating results in the group.

The contractor whose numbers went the other way

BWXT's second-quarter revenue rose 18% to $901.6m. Its trailing-twelve-month book-to-bill ran at 1.7x, meaning it booked $1.70 of new work for every dollar delivered. Commercial Operations — reactor components, small modular reactor pressure vessels, tri-structural isotropic (TRISO) fuel — grew 72%, a third of that organically. Full-year guidance went up to roughly $3.8bn of revenue, $4.70-$4.80 of adjusted earnings per share and $345-$360m of free cash flow.

The entire decline is multiple, not results. At 33.7x forward earnings today, BWXT was worth 43.5x February's price and 42.7x May's on the same consensus estimate of $4.74 — about ten turns removed while orders grew two-fifths.

There is a genuine blemish. Operating income fell 12.2% to $89.9m and operating margin dropped to 9.97% from 13.41%, because BWXT is buying growth with capacity and hiring; the commercial segment's margin guidance was cut to about 13% from 14%, with recovery pushed into 2027. The TRISO line stays conditional — management called demand "still uncertain" pending a government award expected this year and will not commit full capex until the pipeline firms. Against that, the Navy's 30-year shipbuilding plan moves Ford-class carriers to a four-year cadence from 2028, removing the lumpiness that has long made this revenue base uneven.

For the fuel names, the problem is the price, not the order book

Cameco's reported quarter looks catastrophic — revenue down 7.2% to $814.1m, net income down 92% to $25.2m — and is mostly an artifact. The year-ago period carried a one-off Westinghouse payment tied to the Czech Dukovany reactor project; strip it out and the operating picture is intact. Cameco raised its realized-price assumption to C$91-96 per pound, left production guidance unchanged at 19.5-21.5m pounds and holds contracts for more than 28m pounds of average annual deliveries. The uranium long-term contract price, which is what utilities actually pay, reached $94 a pound in August, an 18-year high, above a spot market near $87. On 31 July the company disclosed that Westinghouse had confidentially filed a draft registration statement for a US listing — the event that marks Cameco's low.

What has not corrected is the price. Cameco trades at 161.6x trailing earnings, up from roughly 107x in May: the 21% six-month fall made the stock dearer, because earnings fell faster. Even the forward multiple is 62.2x, and it rests on 2027, when consensus has earnings per share stepping from $1.54 to $2.63.

Centrus is the same shape, sharper. Backlog nearly doubled to $4.5bn stretching to 2040, with $2.4bn of the enrichment book now under definitive rather than contingent agreements, including a supply contract with reactor developer X-energy and a letter of intent with Oklo, both carrying prepayments. The mechanism is a ban on Russian enriched uranium taking full effect within 18 months against no new Western capacity before 2028. The reported quarter was ugly regardless: separative work unit volumes fell 23%, gross margin compressed to 28.3% from 34.9% and operating income fell 69%. Centrus's forward multiple of 70.7x sits above its trailing 70.3x, because consensus has 2026 earnings 44% below 2025, and free cash flow yield is -6.6% during the Piketon build.

Rates amplify; they do not explain

The 30-year Treasury yield touched 5.323% on 18 August, a 19-year high, and BWXT's three worst recent sessions land exactly there, a 7.3% fall from 18 to 20 August. Oklo, Constellation and NuScale each dropped 4-5% that day while broader utilities held. But Brookfield, the Toronto asset manager that owns the other 51% of Westinghouse and reported record fundraising in the same quarter, is down only 8.3% over six months against Cameco's 21% and BWXT's 22.4%. Brookfield trades at 15.1x forward earnings. The cheaper the duration, the smaller the damage — which makes long rates the amplifier and the starting multiple the cause.

None of this was quiet. Since mid-February, Centrus has printed 45 sessions moving 5% or more in a day, Cameco 20 and BWXT 17; Centrus alone fell 20.7% on 11 February. Centrus's shares regained their footing enough for the 50-day average to stop falling away from the 200-day on 12 August. BWXT is the only one of the three still setting new lows.

The setup

Where it stands — Cameco and Centrus have recovered off July lows; BWXT keeps falling despite raised guidance and a 40% larger backlog.

Would confirm — BWXT books the commercial new-build reactor equipment order management expects by year-end 2026 while book-to-bill stays above 1x.

Would invalidate — Uranium's long-term contract price slips back below $90 a pound, or Cameco trims its 19.5-21.5m pound production guidance.

Watch next — Westinghouse's public S-1 filing and price range; BWXT's third-quarter report in early November.

Valuation — BWXT 33.7x forward against 43.5x implied in February; Cameco 62.2x forward, 161.6x trailing versus about 107x in May.