Verdict: Mixed

Key finding: The income statement was a clean beat and raise: Q3 revenue $1.24B (+138%), adjusted EPS $0.94 vs ~$0.69, and the 2026 revenue guide raised a third time to $4.4–4.6B. What the market sold on 10/23 was the cash statement. Operating cash flow turned negative (−$27M) and free cash flow was −$181M, because revenue is being booked faster than customers are billed: contract assets reached $859M, up $618M this year. On top of that, two customers made up 69% of Q3 revenue (one is a related party), and there was no backlog update.

Scorecard

Ticker Rev last Q ($M) Rev YoY last Q Rev YoY prior Q Gross margin (last Q, trend) Op margin EPS last Q vs yr ago P/S Mkt cap 1-mo price %
BE 1,236 +138.1% +165.5% 33.8% (up from 29.2%) 18.2% $0.78 vs −$0.10 14.3x $54.9B −26.6%
VRT 4,313 +61.2% +24.1% 38.1% (flat) 22.8% $1.95 vs $1.02 6.5x $84.6B −3.7%
GEV 12,642 +26.8% +21.9% 24.6% (up) 11.0% $4.81 vs $1.64 5.0x $219.4B −9.7%
FCEL 33 −29.4% −4.9% −74.3% (worse) −141.4% −$0.64 vs −$3.78 5.4x $0.8B −31.7%
PLUG 178 +2.5% +22.3% −0.9% −36.0% −$0.14 vs −$0.20 2.5x $1.8B −29.4%

GAAP diluted EPS; BE's adjusted EPS was $0.94. FCEL quarter ends 7/31, PLUG 6/30. P/S and market cap at the 10/30 close.

BE is the growth leader of the group by a wide margin (+138% revenue, 18% operating margin), and the most expensive (14.3x sales vs 5.0–6.5x for VRT and GEV). Growth slowed a little from +165% to +138% while margins kept rising. The fuel-cell peers (FCEL, PLUG) are not comparable businesses: they shrink or lose money.

Move

Day BE What else moved Read
10/22 (before the report) −8.7% GEV −10.6%, VST −9.8%, VRT −7.8%, SMH −6.0% Sector: AI-power selloff
10/23 (after the report) −11.2% to $193.30 VRT −0.6%, GEV −1.5%, SPY −0.5%, FCEL −4.1% Stock-specific: the report
10/30 −8.6% to $203.17 NBIS −10.4%, CRWV −9.7%, FCEL −7.9%, SMH −7.0%, QQQ −3.5% Sector: AI-infrastructure risk-off

Catalyst

The report (10/22 after the close, Q3 2026). From the 10-Q and press coverage:
- Revenue was $1,235.8M vs $519.0M a year ago. GAAP operating income was $225.4M vs $7.8M, and GAAP net income $249.0M vs a −$23.0M loss.
- Guidance for 2026: revenue $4.4–4.6B (third raise this year), normalized EPS $3.10–3.40 vs $2.71 consensus, gross margin about 34%.

What went wrong was cash, not profit:
- Q3 operating cash flow was −$27.1M, against +$226.4M in Q2. Capex was $153.5M, about 3x Q2. Free cash flow came to −$180.6M, ending four positive quarters. For the first nine months, operating cash flow is still +$272.9M and free cash flow +$41.6M.
- Working capital swallowed the profit. Over nine months, contract assets (revenue booked but not yet billed) rose $618.0M, inventories $285.5M and receivables $264.8M. Customer deposits rose $583.1M, which covered only part of it.
- Management's explanation: revenue is recognized when control of a system passes to the customer, but billing follows contract milestones and can land in a later quarter. CFO Simon Edwards said the quarter is "not the beginning of a trend."
- In July, Edwards set an operating cash flow target of $375M+ for 2026. He didn't repeat it on this call. Hitting it would take about $102M in Q4; the last two fourth quarters brought in $484M and $418M (TIKR's figures).

Concentration and disclosure:
- Two customers were about 43% and 26% of Q3 revenue, and the second is a related party. A related party is also the biggest receivable (about 31% of receivables). Related-party revenue was $321.6M in Q3 and $697.7M over nine months. The 10-Q doesn't name the customers.
- No formal backlog update. Jefferies (Hold) cut its target to $242 from $289, citing the missing backlog update and no update on delays tied to Bloom's largest opportunity, Oracle. Jefferies also said the customer wins Bloom highlighted were already known.
- GAAP remaining performance obligations were $661.2M at 9/30, up from $394.4M at year-end and expected to be recognized within 1–2 years. The guide implies Q4 revenue of about $1.35–1.55B (the $4.4–4.6B range minus $3.05B over nine months), which is roughly double the RPO. Inference: much of the "backlog" Bloom talks about ($6B product / $20B total, as of February 2026) isn't firm under GAAP rules, so investors can't check it from the filings.

Oracle. Bloomberg reported in September that Oracle invoked force majeure on Project Jupiter, the New Mexico AI campus that is expected to use up to 2.45 GW of Bloom capacity. Oracle isn't exiting; it wants to defer payments if the project slips. CEO KR Sridhar's answer on 10/1: Bloom boxes are "fungible" and can be sent to other sites.

Balance sheet is not the problem. Cash is $2.49B against $2.47B of recourse debt, almost all of it the 0% converts due November 2030. Management says liquidity covers at least the next 12 months.

Sector

AI-power and AI-infrastructure names sold off on 10/22, when Meta's results revived worries about AI spending, and again on 10/30. Both days hit BE about as hard as NBIS, CRWV and FCEL. On 10/23, when BE alone fell 11%, the group was flat. Since 10/9: BE −27.6%, FCEL −37.8%, OKLO −27.2%, GEV −14.6%, VST −14.3%, VRT −4.2%, SPY −2.9%.

Risks

Judgement

Mixed. The beat was real. BE is growing faster than any peer, with GAAP profit and rising margins. But the cash miss is a real, checkable flag, not noise: revenue ran ahead of billing, and two customers decide whether it turns into cash. Only the 10/23 drop (−11.2%) was a reaction to the report, and a discount for weaker earnings quality plus missing backlog and Oracle updates is reasonable. The 10/22 and 10/30 drops were the AI-power group. Q4 cash in February 2027 will show whether 10/23 was an overreaction. Until then, the stock trades on AI-capex sentiment.

Sources