ANALYSIS, NOT A POSITION. THE Q2 2026 10-Q WAS FILED 2026-07-28, one day after the memo, and it changes two things materially. (1) TTM revenue is $3,090.8m, not the $1,975m the memo used - +56%. (2) THE ORACLE WARRANT IS RESOLVED AND THE EXPECTED CASH NEVER ARRIVES: Oracle completed a CASHLESS exercise on 2026-05-01, five months before the expiry the catalyst calendar was watching, issuing 2,154,231 shares (0.73% dilution) and NO cash - the calendar's $400.0m of expected proceeds do not exist. The unresolved $45.2m warrant gap is explained: a $72.3m net-settlement inducement, net of $17.9m already amortised, against a $251.6m grant-date fair value. $306.5m of contra-revenue remains queued. Quality Criteria FAIL on the gross-margin limb alone (31.5% TTM against the ~50% INFLECTION standard) - a hardware business measured against a software test, reported as FAIL because the input is present and the rule is the rule. Valuation Criteria FAIL: required 41.3% CAGR against 33.1% demonstrated, margin -8.2pp, at a GROWTH_MATCHED 33.2x exit multiple (n=33) implying 107.4 turns of compression from 140.6x. Flip point 44.7x - which independently corroborates the retired instrument's 40.1x break-even. The scandium record is preserved as corrected: scandium appears in one quarter of nine and in ZERO prepared remarks, fifteen months before Hunterbrook, and the sourcing allegation cannot be adjudicated from public filings.
How to read this
This is an analysis, not a position. The memo scores every Criteria and blocks on none of them. Whether an analysis justifies a position is a question about a particular book, and two books answer it differently — so this page carries no Long, Short, Watchlist or Avoid verdict. The previous verdict has been retired.
Every Criteria returns PASS / FAIL / INDETERMINATE, and carries a type. BINDING criteria are admission tests for a long-only absolute-return strategy. MEASURED criteria are always scored and stored, and never block — they inform timing, sizing or a future strategy. A missing input is INDETERMINATE, never FAIL.
Two valuation outputs, over two horizons. The implied-path test (reverse DCF) asks what today's price requires over five years and whether the business has demonstrated it; the 12-month target asks what the name is likely to trade at, on near-term consensus and the name's own multiple history. Neither replaces the other. Sensitivity is run over the exit multiple, never over scenario probabilities.
Momentum is entry timing only. It governs when to enter a position the thesis already justifies, never whether to own one.
Key findings
- THE ORACLE WARRANT - $306.5m of future contra-revenue, surfaced by tracing an unexplained new balance-sheet line rather than from any commentary. BE agreed on 2025-10-28 to issue Oracle a warrant over 3,531,073 shares at $113.28, accounted for under ASC 606/718 as consideration payable to a customer's customer, meaning it is recognised as a REDUCTION OF REVENUE as the underlying Energy Servers are delivered. Fair value marched $55.9m (Dec-25) -> $183.6m (Mar-26) -> $261.3m grant-date (Apr 9, 2026), and the June 30 balance sheet carries $306.5m ($90,967k current + $215,533k long-term). $12.8m was already taken against Q1'26 revenue. Roughly 7.6% of a full year of FY2026E revenue is queued as contra-revenue, landing precisely as the Oracle volumes - the growth story - are delivered.
- BACKLOG HAS BEEN QUANTIFIED EXACTLY ONCE. Q3'25 press release: zero mentions. Q4'25 press release (Feb 5, 2026): '~$20 billion total, ~$6 billion product, up ~2.5x YoY'. Q1'26 press release: boilerplate risk-factor mention only. Q1'26 10-Q: ZERO mentions of the word 'backlog'. Q2'26 press release today: boilerplate only. BE's own July 9 rebuttal to the short report expressly invoked 'our current fuel cell demand and backlog' as a defence, and the print three weeks later quantifies none of it. The same disclosure pattern flagged on TWST.
- AND THE BACKLOG DEFINITION INFLATES IT. Per BE's own footnotes to that single disclosure, product backlog 'reflects anticipated ITC and other tax incentives as applicable' - so the ~$6bn is not $6bn of Bloom product revenue. Service backlog covers contracted O&M for servers 'to be delivered' as well as installed, so the ~$14bn balance is long-dated annuity over the multi-decade fleet life. '~$20bn of backlog' is roughly 70% multi-decade O&M and 30% product-including-tax-credits.
- REVERSE-DCF: WHAT $183.90 REQUIRES. Even in the single most generous cell - a 50x exit multiple AND a 12% cost of equity for a stock with measured beta of 3.14-4.20 - the price needs FY2029 revenue of $8.8bn, a 29.6% CAGR for three years ON TOP OF a year that doubles. To clear the 4.7% hurdle, BE must still trade at 53-80x earnings in 2029 as a ~$9bn-revenue, ~20%-growth company. Chan/Karceski/Lakonishok 2003 find growth persistence beyond chance is not observable.
- THE STRONGEST FACTOR PROFILE IN COVERAGE - AND IT STILL FAILS. Top-decile 12-1 momentum (+386% to +800%), a +29.0% revenue and +90% EPS surprise versus consensus, an FY2026 EPS guidance raise landing 24.4% above prior consensus, and favourably negative accruals (-11.9% of assets). Offset by worst-decile asset growth of +122.4% YoY, driven by an external capital raise (cash $606m -> $2,667m), which is the bad kind of asset growth rather than the benign kind.
- A QUALITY-OF-EARNINGS POINT IN THE COMPANY'S FAVOUR THAT THE BEAR CASE HAS NOT CREDITED. Q1'26 contained $373.3m of related-party revenue - 49.7% of the quarter - sold into the Brookfield JVs in which BE holds equity. Q2'26 contained $2.8m, i.e. 0.3%. Q2's growth is almost entirely third-party and arm's length, making it materially higher-quality revenue than Q1's.
- THE H2 MATH. H1'26 revenue was $1,816.4m against FY guidance of $3.9-4.2bn, implying H2 of $1,041.8-1,191.8m per quarter versus Q2's $1,065.4m - so the LOW END of guidance implies a sequential decline. Against H2'25 of $1,296.7m, however, the H2'26 midpoint is still +72.3% YoY. Both framings are honest; the bear uses the first and the company the second.
- ALPACA'S REPORTED IMPLIED-VOLATILITY FIELD IS UNRELIABLE. For BE it returned call IV ~110% and put IV ~155-170% at the same strike and expiry - impossible if put-call parity holds, and it does: measured C-P against S-Ke^(-rT) agrees to under $1 at every strike tested. A Black-Scholes recomputation from mid prices gives call and put IV agreeing within ~2 points (146.2% vs 147.1% at the Sep-18 $160 strike). True term structure is 146% / 137% / 129% at 52 / 80 / 171 days, versus 122.5% three-month realised - so implied is ~1.19x realised and premium is expensive, the exact inverse of GOOGL at 0.81x.
- A MODEL ERROR CAUGHT ONLY BY EXCEL READBACK. The first pass used an 18% tax rate in the bull case and 21% elsewhere, which double-counts optimism. Standardising on 21% cut bull FY2029 EPS from $7.64 to $7.37 and the probability-weighted value from $73.76 to $72.43. The nine-line tie-out block to the filed statements returned TIE on every line both before and after - which is precisely the point of the C8 rule: a model that balances but does not tie is confidently wrong.
Sections
Disclosed limitations
- TRANSCRIPT SERIES IS INCOMPLETE - 4 of 9 quarters. The shared Alpha Vantage 25-requests/day cap was exhausted mid-run by six concurrently-running agents, yielding only 2024Q1, 2024Q2, 2024Q3 and 2026Q1. The missing 2024Q4-2025Q4 window is exactly when the Oracle deal was signed (Oct 2025). NO 'first-ever mention' claim is made anywhere in this memo; the strongest claim made is 'first in the four quarters sampled'. A background job is queued to complete the series when the quota resets at 00:00 UTC, and this memo will be refreshed.
- THE Q2 2026 EARNINGS CALL TRANSCRIPT IS NOT YET INCORPORATED. The call was held at 14:00 PT on 2026-07-28 and this memo was assembled while it was in progress. Management commentary on the H2 guidance shape, capacity constraints, backlog, and any response to the short report is therefore NOT reflected. This is the single largest open input.
- BE DISCLOSES NO GEOGRAPHIC REVENUE SPLIT. Attempted and returned nothing: the Q1'26 10-Q disaggregation note (four product categories only - product, installation, service, electricity - with zero occurrences of 'revenue by geography'), the segment note (BE reports as a single segment), and the related-party notes (which reveal a Republic of Korea VIE and the SK ecoplant relationship but no revenue split). The mandatory revenue-by-geography chart is therefore replaced by revenue-by-category, with this absence stated rather than papered over.
- THE $45.2m WARRANT GAP IS UNRESOLVED. Grant-date fair value is disclosed as approximately $261.3m but the carried customer-consideration asset at June 30 is $306.5m. The Q2 10-Q, not yet filed, should reconcile this. No explanation has been assumed.
- THE SCANDIUM-SOURCING ALLEGATION CANNOT BE ADJUDICATED FROM PUBLIC FILINGS - but the original phrasing of this limitation was FALSIFIED and is corrected here. The completed 9-quarter transcript series shows scandium discussed 3x and china 6x in 2025Q1, fifteen months BEFORE the Hunterbrook report. The corrected and narrower claim: scandium appears in ZERO prepared remarks in any of nine quarters - never volunteered by management, only surfaced under analyst questioning, in the single tariff-driven quarter. BE does not name input suppliers. The company's 25GW/yr figure is a supply-chain VISIBILITY claim, not a demand or revenue-capacity claim.
- FY2029 EXIT MULTIPLES AND THE COST OF EQUITY ARE ASSUMPTIONS, NOT MEASUREMENTS. They are the two dominant levers, so both are reported as a full sensitivity grid rather than as point estimates. The Gate 4 conclusion holds across every defensible cell, but the grid is the honest presentation.
- PRE-PRINT CONSENSUS FOR THE PEAD CHECK IS WEB-SOURCED AND THE FIELD DISAGREED WITH ITSELF: one source gave Q2 revenue consensus of $766.9m with $0.39 EPS, another $826.1m with $0.41. The beat is large on either basis, and the $826.1m/$0.41 pair is used throughout as the more conservative.