Setup

Bloom Energy filed Q1 2026 10-Q on April 29. The print drove a +27% one-day move on first GAAP profitability at scale ($73.7M net income, +$73.6M operating cash flow). Three disclosures inside the same filing, less visible than the headline numbers, set up the next 90 days against the rally: the hyperscaler driving the inflection (Oracle), an Oracle warrant accounting headwind that flows ≈$248M through Q2-Q4 GAAP revenue, and a $19.7M warranty reserve for "identified product issues" with the failure mode undisclosed.

What the filing says

Q1 2026 revenue $751M (+130% YoY), product revenue $653M (+208%). MD&A: "increase in demand for our products, largely resulting from multiple projects executed through the joint venture with Brookfield, including a major hyperscaler project." U.S. revenue jumped 56% to 91%. Brookfield JV channel = 50% of Q1 revenue (43% FY2025).

The hyperscaler is Oracle. Cross-reference: Bloom April 13 PR (2.8 GW master agreement, 1.2 GW initial booking) plus Oracle April 27 "Project Jupiter" PR (incremental 2.45 GW NM microgrid). Combined ≈4 GW of Bloom SOFC committed. Peer fuel cell OEMs (FCEL, PLUG, BLDP) have zero comparable hyperscaler bookings — this is durable share gain, not sector beta.

Oracle warrant. Fair value escalated $55.9M → $183.6M → $261.3M at the April 9 grant date (locks the value). Q1 recognized $12.8M as contra-revenue. Remaining ≈$248M flows through Q2-Q4 2026 GAAP revenue as Oracle takes delivery — roughly $80M per quarter ratable.

Specific product warranty $19.7M for "identified product issues." Not in FY2025 10-K. Nature undisclosed. Total warranty $20.0M → $38.4M. Absorbed ≈3 margin points in Q1.

AEP $2.65B Devco/Cheyenne deal — the headlining strategic item in BE's February 2026 10-K — is completely absent from this filing. No CWIP update, no conditions language. Conditions due Q2 2026 (June 30). AEP CWIP only +$23M YoY through Dec 2025 on a $2.65B program. AEP Q1 10-Q (~early May) is the resolver.

Behavioral signal: six C-suite sales totaling $33M+ during the trailing-twelve-month +1,472% run, zero buys. Prior CFO Cameron departure unexplained; Edwards offer letter dated March 18 attached as exhibit.

What the market thinks

Analyst targets set after the print, relative to the April 29 post-print close:

Mean target sits 24% below the post-print close after six firms updated. Forward P/E 71.67 on $81.9B market cap. Forward 14d IV 114%, vol skew puts +53% over ATM (downside protection bid).

Reverse-engineering market probabilities against three states ($400 expands / $200 holds / $100 contracts), market-implied P(expands) ≈ 52%. Our P(expands) — once the warrant contra-revenue, warranty, and AEP silence are taken seriously — is 45%, not the 55% written into the long-horizon factor scenario. At 45/30/25, probability-weighted intrinsic value is $265, roughly 8% below post-print levels. The market is pricing a better outcome than the filing supports.

Why the gap exists

Three reasons sell-side has not priced the bear stack:

1. Warrant accounting is buried. The $248M GAAP revenue headwind sits in a warrant footnote, not the income statement narrative. Quarterly run-rate models extrapolate from headline revenue; the contra-revenue must be entered manually. After +27% on Q1 numbers, the impulse is to upgrade revenue, not to flag a 12-month accounting drag.

2. The warranty disclosure is undescribed. $19.7M for "identified product issues" with no nature given. Analyst Q&A on the Q2 call is the first chance to extract detail.

3. Silence is non-disclosure. Markets reprice disclosed bad news, not absent good news. The AEP Devco deal not being mentioned in a quarterly is interpretively ambiguous — and the impulse after a +27% print is to assume things are fine.

Risks (ranked)

1. Warranty reserve grows >$30M in Q3 10-Q — fleet reliability issue; thesis-breaker for the SOFC-for-AI bull case.
2. Q2 2026 GAAP revenue prints below consensus due to ratable Oracle warrant contra-revenue.
3. AEP Devco deal terminates or extends — $2.65B order risk; was a stated forward catalyst.
4. Insider exodus + unexplained CFO turn during peak run is a behavioral signal markets typically discount only on the way down.
5. Single-customer concentration — 50% of Q1 revenue from one channel; Oracle is THE customer.
6. Capital structure overhang — Oracle warrant 3.5M shares cashless-exercisable through Oct 9, 2026; SBC $228M annualized.

Catalysts

The next 90 days are the highest information-density window in the BE story since the Oracle warrant issued. Five resolvers stack:

What would change our mind

Bullish:

• Microsoft, Meta, or Google announces ≥500 MW Bloom SOFC commitment → P(expands) jumps to ≈70%, intrinsic value ≈$325.
• AEP closes deal or 8-K confirms condition satisfaction by June 30 → removes near-term bear catalyst.
• Warranty reserve flat at ≈$20M through Q3 10-Q → discrete batch confirmed; removes warranty from bear stack.
• Insider buying (Form 4 code P) at any level → reverses governance signal.
• Q2 contra-revenue ≥$50M AND headline revenue beats consensus → bear setup invalidated.

Bearish:

• Warranty grows to $40M+ → bear ratifies hard.
• AEP terminates or amendment disclosed → breaks multi-catalyst narrative.
• Oracle exercises warrant cashlessly at scale → supply overhang materializes Q3-Q4.
• CFO Edwards departs in <12 months → governance signal compounds.

Evidence