Bloom Energy — Q2 2026 Earnings Preview

Growth trajectory — accelerating on every line, and that is now the problem. Q1 2026 printed revenue of $751.1 million, +130.4% on Q1 2025's $326.0 million, with product revenue $653.3 million (+208%), non-GAAP gross margin 31.5% (+280bps YoY), adjusted EBITDA $143.0 million (+468%) and non-GAAP diluted EPS of $0.44 against a ~$0.12–0.13 street. Management raised the FY2026 revenue guide by $400M at both ends — the low end of the new range above the high end of the old — and the stock reacted +27.2% the next session on 30% above-average volume, holding the gain three sessions later.

The set-up into Q2 is the mirror image. Six straight double beats did not stop a -47% drawdown from the high, because the debate moved from "will they beat?" to "is the backlog real and is the scandium supply chain real?" Consensus has also finally re-anchored: the Q1'26 bar was ~$551.6M and the Q2'26 bar is $849.7M on the widest aggregate — the bar moved up 54% in one quarter, with revenue estimates revised +16.85% in three months. A fifth 20%+ beat now requires roughly $1.02B, or +36% sequential, in what is not the seasonally strong quarter.

Key watch items:

• Guidance action (the event). FY26 stands at $3.4B–$3.8B revenue, ~34% non-GAAP GM, $600M–$750M non-GAAP operating income, $1.85–$2.25 EPS. Consensus is already above the midpoint on both lines. Precedent says reaffirm: at the equivalent Q2 call last July, management reiterated the FY2025 guide of $1.65B–$1.85B and then finished the year at $2.024B, +9.4% above the high end. A reaffirm this time reads as a cut.
• Backlog → RPO bridge. Last disclosed 2025-12-31: product backlog $6 billion, service backlog $14 billion. Daloopa carries no 2026Q1 backlog datapoint because management did not restate it. A third quarter of silence is itself the material event.
• Gross margin is the binding constraint, and nobody is watching it. Q1 delivered 31.5%; the FY ~34% guide requires ~34.7% blended across Q2–Q4, a 320bp step-up with no three-quarter precedent. Street models Q2 revenue up sequentially and EPS down — an embedded margin contraction in a year guided to 320bp of expansion.
• The unpriced positive. Current deferred revenue and customer deposits hit $194.1 million at 2026Q1 versus $59.96 million at 2025Q2 — +224% YoY and +92% sequentially. Cash prepayments are the one demand proof a short report cannot redefine, and no one is quoting them.
• Tariffs / trade: not the live issue. Tariff mentions fell to 1 on the Q1 call (from 2). The supply-side risk here is input-material and FEOC exposure on scandium oxide, not tariff rates.

Classification: CONSERVATIVE guider, but the easy asymmetry is gone. FY2025 was guided to $1.65–1.85B and delivered $2.024B; non-GAAP operating income was guided $135M–$165M and delivered $221M. The FY2026 initial guide was set at $425M–$475M of operating income in February and moved to $600M–$750M in April — a deliberately low initial bar. But the April raise spent the good news early, and the disclosure risk that now dominates is not a P&L risk. HOLD into the print: a beat on both lines with no incremental RPO, capacity or scandium disclosure is a bear-case outcome, not a bull-case one.

How to read Bloom "guidance." Bloom has issued no quarterly revenue, EPS or margin guidance in any of the last four quarters — Daloopa's quarterly-guidance series carry no values from 2024Q4 through 2026Q1. The company guides the full year only. Everything below therefore measures the print against (a) the standing FY2026 guide and (b) one qualitative sentence about Q2.

Guidance figures carry Daloopa citations in section 1 and section 4; EPS ranges are taken verbatim from the Q4'25 and Q1'26 8-K EX-99.1 Outlook sections.

Is the guide conservative or aggressive? Split verdict — and the split is the point.

The live management-vs-street contradiction is margin, not revenue. Four straight calls of "capacity, not demand, is the governor" have now been believed on the revenue line — consensus is +16.85% in three months and above management's own floor. That trade is done. What the street has not priced is the margin claim: $849.7M of revenue (+13.1% sequential) against $0.41 of EPS (-6.8% sequential) is a modelled sequential margin contraction in a year guided to 320bp of expansion. Either street Q2 margin is far too low, or the FY guide needs an implausible H2 step-function. It resolves on Tuesday, and notably it is not the thing the short reports attack.

Every relevant power/AI peer that reported into BE's print beat and raised. That is analytically important: it means any disappointment on Tuesday is idiosyncratic — backlog quality, RPO, scandium, unit economics — and cannot be blamed on the end market. It also raises the bar. With GEV printing +134% orders, "demand is strong" is no longer differentiating commentary. Note that BE prints the day before MSFT and META — it goes without the two biggest capex datapoints of the quarter on the tape. Same-day, ENPH reports and can create sector noise.

4a. Current quarter — Q2 2026

Business type: industrials / capital equipment — product units × ASP through the Product line, with Installation and Service attached. % Diff is consensus vs the guide (or vs the implicit floor, where that is the only guide).

Q2 2026 consensus vs the same quarter last year:

Q2 2025 actuals: total revenue $401.2 million, product revenue $296.6 million, adjusted EBITDA $41.2 million, non-GAAP EPS $0.10, product non-GAAP GM 34.3%, service non-GAAP GM 12.2%.

Geographic mix is an unmodelled swing factor. US share of revenue ran 59% in Q2'25 → 92% in Q3'25 → 81% for FY2025 → 91% in Q1'26. The AI datacenter pull is almost entirely domestic and management said international "will lag behind a little bit." Expect US mix near 90%.

Revenue and margin run-rate into the print:

Citations: 2025Q3 revenue $519.0 million / product $384.3 million / EBITDA $59.0 million; 2025Q4 revenue $777.7 million / product $638.5 million / EBITDA $146.1 million; 2026Q1 revenue $751.1 million / product $653.3 million / EBITDA $143.0 million / GM 31.5%; FY2025 revenue $2,024.0 million / EBITDA $271.6 million.

4b. FQ+1 guide (expected) — Q3 2026

Bloom has not guided a quarter in any of the last four calls and there is no basis to expect a Q3 guide on 2026-07-28. What follows is the Q3 run-rate implied by the FY guide after backing out Q1 actual and Q2 consensus. This is arithmetic, not guidance.

4c. FY guide — old vs current

FY2025 guide citations: revenue $1.65B–$1.85B, GM 29%, op income $135M–$165M. FY2026 initial: revenue $3.10B–$3.30B, GM 32%, op income $425M–$475M, EPS $1.33–$1.48. FY2026 raised: revenue $3.40B–$3.80B, GM 34%, op income $600M–$750M, EPS $1.85–$2.25, adjusted EBITDA $650M–$800M.

4d. Balance-sheet demand signals — the unpriced tell

Citations: cash 794,751 / 574,764 / 595,055 / 2,454,108 / 2,491,433; current deposits 168,444 / 59,964 / 56,065 / 100,975 / 194,094; non-current 47,173 / 47,649 / 32,254 / 42,840 / 39,260; contract-balance deferred revenue 59,008 / 56,172 / 53,675 / 65,608 / 82,254.

Why this line is the whole argument in one row. Current deferred revenue and customer deposits nearly doubled sequentially (+92% Q4→Q1) and more than tripled off the 2025Q2 trough. Management attributed part of the record Q1 operating inflow to "customer prepayments to reserve capacity." Prepayments are cash — contractually recognised and immune to a definitional argument about what "backlog" means. If this line steps up again in Q2, the backlog critique loses most of its force regardless of what the RPO footnote says. If it stalls while backlog grows, the short thesis gets its best evidence. The non-current line drifting down while the current line explodes is consistent with near-dated deliveries, which supports the 2026 ramp.

Backlog, as the company defines it (2025Q4, last disclosed): product $6 billion + service $14 billion = the "$20B." Two things matter more than the headline: 70% of it is service backlog — the 10–15 year annuity attached at a 100% attach rate, real but long-dated and not a near-term product-revenue proxy; and the short-side "$492M audited binding RPO" figure and the $20B backlog figure are not measuring the same thing (RPO under ASC 606 excludes cancellable and option volumes). Management has never bridged the two. That omission is now the entire debate.

Bottom line. Management's tone into this print is the most confident it has been in four quarters — and the gap between that confidence and what management has actually disclosed is the entire set-up. Tone went disciplined-confident (Q2'25) → accelerating (Q3'25) → concrete-confident (Q4'25) → fully unhedged (Q1'26). The last step traded specificity for conviction. When conviction rises and disclosure falls in the same quarter, the short thesis writes itself — and three months later, it did.

5a. The tone arc, measured

Keyword frequency across the four transcripts in the run folder — this is not impressionistic:

The single most important observation. At Q4'25 management built the entire equity story on backlog — 20 mentions, product backlog "increased 140% year over year to about $6 billion." One quarter later K.R. Sridhar actively disparaged backlog as a metric: "The traditional power industry has been the past 2 years, celebrating its backlog that is 4 and 5 years out… We want to rapidly build capacity… not just be satisfied with simply building backlog." Backlog mentions fell 20 → 6; capacity mentions rose 11 → 27. Management pivoted from stock of orders to rate of production one quarter before the short report attacked exactly the quality of that backlog. Read it charitably or uncharitably — the pivot is on the record and the Q2 call cannot dodge it.

5b. Where management is committed vs where it is hedged

The tell: the metric management guides most tightly (operating income) is the one most within its own control; the one it guides most loosely (revenue) is the one it has explicitly assigned to customer construction schedules.

Caveat inventory — strikingly thin. Across the entire Q1 2026 call the complete list of management-volunteered risk is: "barring any global shock or exogenous factors" (on GM), "the commercial landscape is fluid and dynamic" (used as an upside argument), supply-chain "speed bumps" (immediately dismissed — "are we worried about it or lose sleep… Absolutely not"), and international LNG delay (not in the FY26 guide in size). Critically, the mix-volatility gross-margin caveat that accompanied the 32% guide did not accompany the 34% guide. Fewer caveats attached to a harder target is not the direction you want. Never raised on any of the four calls: raw-material single-source risk, scandium, backlog-to-RPO mechanics, or any downside scenario. Sridhar closed Q1 with a dare rather than a caveat: "If you bet against any one of the statements that I made you can bet against [Bloom]."

5c. Post-guidance updates — everything since 2026-04-28

5d. Macro backdrop — every demand vector still supports the guide

Grid interconnection queues remain 5+ years in many states; gas turbine lead times run ~36–60 months (GE Vernova quoting ~3 years, capacity in 2029/2030), which directly validates Sridhar's core claim that "their supply to current orders arrive only in 2029 or later… ours arrive this year or the next"; CCGT pricing is up 10–20% YTD, widening Bloom's relative value stack without Bloom needing to take price; and FERC's 2026-06-18 §206 show-cause orders to all six RTOs/ISOs (Docket RM26-4) institutionalise the behind-the-meter path. The only genuinely new negative is supply-side — scandium concentration and FEOC exposure — and it is unresolved. That is an unusually clean bifurcation: the market Bloom sells into is fine; what is contested is whether Bloom can physically build what it has promised. A revenue beat cannot resolve it, because Q2 revenue was built from inventory and capacity that already existed.

Ranked by expected share-price impact on the print. "Consensus view" is what the street is modelled to believe for this quarter; "Confidence" reflects source quality, not conviction in the outcome.

The five questions that decide the call

1. What is RPO at 2026Q2, and how does it bridge to the $20B backlog? Everything else is secondary — a bridge ends the debate in one slide.
2. Did deferred revenue and customer deposits step up again from $194.1 million? The falsifiable version of question 1.
3. Jupiter: what is the revised delivery schedule for the 1.2 GW contracted, and are units being redirected to other Oracle sites under the MSA's portability clause?
4. Installed manufacturing capacity in MW at quarter-end, and MW added during Q2 — installed, not entitled.
5. Scandium: months of inventory on hand, and how many qualified non-China suppliers? Bloom has asserted sufficiency three times without a number.

Everything below post-dates the Q1 2026 report and precedes the Q2 print. Earnings mechanics are excluded — they live in sections 2 and 4. Content aggregators with no editorial standard surfaced repeatedly in search and were excluded; nothing here rests on them.

What the newsflow actually says

1. Demand-side news was uniformly strong; supply-side news was uniformly contested. In under four weeks Bloom added a $25B financing framework, a $1.7B project financing, and a favourable FERC posture. In the same window it absorbed two short reports, five plaintiff-firm investigations, and a competitive win by FuelCell. Those are not in tension — nobody disputes that customers want the product. The dispute is whether Bloom can physically build it.

2. The bear case is specific, testable, and therefore resolvable on 2026-07-28. The reports do not allege a soft valuation problem; they allege a hard input constraint and an overstated backlog. Both are answerable with disclosure. Management's willingness to provide the reconciliation is a higher-signal observation than the revenue beat.

3. Sentiment inverted from "good" to "questioned" with no fundamental deterioration. Nothing in the period is a downward revision, a lost customer, a guidance change, or a delayed filing. The entire narrative reset came from two self-disclosed short sellers. Under the sentiment-inversion principle that is the interesting direction — consensus enthusiasm broken by an argument rather than by results. Whether it is an opportunity depends entirely on whether the scandium claim is true, which no outside party can currently verify.

4. The genuine red flag is not the short report — it is the definitional gap. A ~$20B headline backlog against ~$492M of audited binding RPO is a 40x gap. There are legitimate reasons for it (framework MSAs, capacity reservations and financing commitments are not enforceable performance obligations under ASC 606). But management quotes the large number publicly and the small number is the one in the filings. That is precisely the "conflicting statements across filings or transcripts" pattern in the red-flag framework, and it needs resolving from the 10-Q rather than accepted from either side's framing.

Pattern: regime-change beater, not a "consistent beater." Over 12 quarters BE beat non-GAAP EPS 7 times and missed 5 (58% hit rate), and 7 of 9 assessable quarters on revenue (78%) — a 66.7% combined rate. But the record is not random: there is a hard break at 2024Q4. Before it, BE missed 4 of 6 EPS prints and 2 of 3 revenue prints. Since it, BE has gone 6 for 6 on EPS and 6 for 6 on revenue. Anyone quoting "100% beat rate" is quoting a six-quarter window that begins exactly at the AI/datacenter demand inflection.

Revenue: consensus vs actual

Actuals with citations: $235.3 million, $335.8 million, $330.4 million, $572.4 million, $326.0 million, $401.2 million, $519.0 million, $777.7 million, $751.1 million. Non-GAAP diluted EPS actuals: -$0.17, $0.15, $0.07, -$0.17, -$0.06, -$0.01, $0.43, $0.03, $0.10, $0.15, $0.45, $0.44.

Is beat magnitude improving or deteriorating?

Dollar revenue beats compound rather than plateau: +$25.0M → +$91.0M → +$137.6M → +$220.6M. Each quarter the street has been wrong by more than the last, in the same direction, without exception.

How to read that — carefully. Trajectory-over-absolutes says improving beat magnitude is a positive signal. It is, but only when the improvement comes from execution against a model the street broadly understands. A +41.6% revenue surprise is not that. It says the sell-side has no working framework for BE's revenue recognition, which means the beat is a function of estimate error, not a durable operating gap. Estimate error mean-reverts. Six quarters of one-way surprise is exactly the condition under which analysts over-correct — and the Q2 2026 consensus shows they just did: the bar moved from $551.6M to $818–850M in one quarter, +54%.

What the record implies for Tuesday

Three conclusions. (1) The revenue bar is genuinely high for the first time in six quarters — consensus sits ~9% above the only quantitative Q2 statement management has made, and every prior beat in the streak started from a consensus below management's framing. The $51M spread between Zacks and Investing.com means "beat or miss" is partly a question of which consensus gets quoted in the headline. (2) The EPS bar is soft — $0.402 implies EPS falls sequentially in a year guided to 320bp of margin expansion; on the 12-quarter record EPS has beaten in every quarter revenue beat, so an EPS beat is the base case. (3) The record cannot resolve the actual controversy. Six straight double beats did not stop a 47% drawdown from the high or a -14.9% session on 2026-07-24, because the debate moved from "will they beat?" to "is the backlog real?" A beat on both lines with no incremental RPO, capacity or scandium disclosure is a bear-case outcome, not a bull-case one. The asymmetry is in the disclosure, not the print.

HOLD into the print. The guidance history says the quarter beats — a conservative guider with escalating beat magnitude, a soft EPS bar, a demand backdrop where every peer beat and raised, and an unpriced prepayment signal running +224% YoY. But three things blunt the usual asymmetry: the bar has migrated from the printed consensus to the whisper; the April raise pulled the good news forward so the FY guide already embeds the H2 ramp; and the dominant risk is a disclosure risk that no revenue number can retire. The guidance history says nothing about whether they refresh the backlog, and that is what this print is actually about. Buy the disclosure, not the beat.