Q2 2026 revenue
$1.07B
Bloom Energy · NYSE: BE · Post-earnings decision memo
The business accelerated. The stock still demands perfection.
Q2 2026 delivered a rare beat across revenue, margin, and EPS, followed by another guidance increase. But the quarter also included a tariff recovery, a specific warranty reserve, dilution, and 73% concentration in one contractual customer. The key question is no longer whether Bloom can grow—it is how much of this profit repeats and how much of 2028 success is already priced in.
Verdict: the business is much stronger; the stock's risk/reward remains unattractive.
Bloom demonstrated that data-center demand, product scale, and operating leverage are reaching the P&L. Revenue grew 166% YoY, GAAP operating margin reached 17.1%, and the midpoint of 2026 non-GAAP operating-income guidance rose 26% in one quarter. S1S2
That strengthens the company thesis, not necessarily prospective stock returns. At $163.75, a fully diluted proxy trades at 7.5× 2027 EV/revenue and 33.7× 2027 P/E. The 2028 base case is worth about $171 per share before discounting and $130 in present value; the current price therefore requires something between a very well-executed base case and the bull case. S7
Quality of print: the operating step-up is real; the run-rate is not clean
GAAP and non-GAAP improved in the same direction, which is encouraging. The caveat is that both include the tariff recovery and warranty cost; the published non-GAAP framework normalizes neither item. S1S2
$196.3MGAAP income
to common
to common
+$2.6MNoncontrolling
interest
interest
+$56.4MStock-based
compensation
compensation
–$8.6MEquity earnings +
asset gain
asset gain
+$1.5MRestructuring,
derivative, other
derivative, other
$248.2MPublished
adjusted income
adjusted income
What looks recurring
- Product revenue was $935.4 million, 88% of total revenue and +215% YoY; scale and mix explain a meaningful share of the leverage. S1
- Non-GAAP operating income reached $239.6 million, a 22.5% margin; adjusted EBITDA was $253.4 million. S2
- Service margin reached 22%, and management argued for a long-term level above 20%. This is a call claim that still needs repetition. S6
- Quarterly CFO of $226.4 million and derived FCF of $174.8 million show that the expansion was not only accounting profit. S2
What prevents a simple extrapolation
- A $37.4 million recovery of previously paid import tariffs added about 351 bps to total gross margin; excluding that credit alone, GAAP gross margin would be ~29.9%, not 33.4%. S1
- The warranty liability rose from $20.0 million at year-end 2025 to $77.8 million and includes a $58.3 million specific reserve. The filing does not allocate all of it to Q2. S1
- SBC of $56.4 million equaled 23.5% of non-GAAP operating income and added roughly $0.17 per share to the adjusted bridge. S2
- Average diluted shares were 323.3 million, +39% YoY and 12.5% above the basic average; per-share growth carries meaningful dilution. S1
Bridge conclusion: the operating improvement is unequivocal, but $0.78 of adjusted EPS is not a clean base. The tariff recovery alone equals $0.116 per diluted share before tax—a mechanical sensitivity, not normalized EPS. I did not net the tariff benefit against the warranty reserve because the filing does not provide the reserve's exact quarterly timing.
| Metric | Q2 2026 actual | Consensus mean | Beat | Read-through |
|---|---|---|---|---|
| Revenue | $1,065.4M | $827.0M | +$238.3M / +28.8% | Exceptional beat with strong demand content. S7 |
| Adjusted EPS | $0.78 | $0.407 | +$0.373 / +91.8% | Strong direction; magnitude affected by the operating items above. S2S7 |
Fiscal.ai consensus retrieved July 29, 2026. For Q2, the last pre-print data points are dated July 22, making the comparison clean; the provider does not preserve intraday timestamps and shows final historical consensus for older periods.
Guidance: the year's second increase shifts the burden of proof to H2 2026
| FY 2026 | Initial (Q4 2025) | After Q1 2026 | After Q2 2026 | Midpoint Δ QoQ |
|---|---|---|---|---|
| Revenue | $3.1–3.3B | $3.4–3.8B | $3.9–4.2B | +$450M / +12.5% |
| Non-GAAP gross margin | ~32% | ~34% | ~34% | Unchanged |
| Non-GAAP operating income | $425–475M | $600–750M | $800–900M | +$175M / +25.9% |
| Non-GAAP EPS | $1.33–1.48 | $1.85–2.25 | $2.55–2.85 | +$0.65 / +31.7% |
Sources: initial guide S5, post-Q1 guide S4, and current guide S2.
Implied H2 revenue
$2.084–2.384 billion, or $1.042–1.192 billion per quarter. The midpoint requires a $1.117 billion average, only 4.8% above Q2.
PlausibleThe volume hurdle is not heroic.Implied H2 gross margin
~34.7% non-GAAP to finish the year at 34%. That requires keeping margin above Q2 after removing the nonrecurring tariff benefit.
GateMix and warranty determine quality.Implied operating income
$430.6–530.6 million in H2. The midpoint requires ~$480.6 million, above H1's $369.4 million.
ExecutionThe guide assumes additional leverage.
Guidance read-through: the revenue hurdle is reasonable; margin is the harder test. Q3 + Q4 consensus averages about $2.26 billion, close to the $2.23 billion implied midpoint. The stock's next leg must come from margin quality, cash conversion, and further revisions—not simply achieving volume.
| Consensus revision | Last point dated July 28 | July 29 | Change |
|---|---|---|---|
| FY 2026 revenue | $3.761B | $4.119B | +9.5% |
| FY 2026 EPS | $2.196 | $2.710 | +23.4% |
| FY 2027 revenue | $6.522B | $6.752B | +3.5% |
| FY 2027 EPS | $4.602 | $4.862 | +5.6% |
Fiscal.ai supplies a date, but not a time, for the July 28 revision. It is the last point dated that day, not necessarily a freeze before 5 p.m. ET. S7
Capacity and data-center share: 2 GW is a target run-rate, not disclosed output
Annual manufacturing capacity
Best current reference: approximately 2 GW per year by year-end 2026. The 2025 10-K says Bloom is expanding Fremont from roughly 1 GW to 2 GW of annual production capacity by the end of 2026. On the Q2 call, an analyst framed the company as expanding “from 2 GW to beyond,” and management did not correct the premise. S16S6
Capacity targetActual yearly production
Not disclosed in GW. The Q2 release, 10-Q/A, and call do not state actual trailing-12-month output or shipments in megawatts. Revenue cannot be converted cleanly into GW because price, configuration, customer incentives, installation, service, and recognition timing vary.
Do not equate capacity with outputInstalled base
Bloom reported 1.8 GW deployed cumulatively across roughly 1,100 sites in nine countries as of its 2026 proxy materials. That is lifetime fleet deployment across all verticals—not annual production and not data-center-only capacity. S17
Cumulative, not annualWhat “very high 90s” actually means
In response to a question specifically about competing fuel-cell models targeting data centers, the CEO said Bloom's data-center fuel-cell share was in the “very high 90s.” No denominator, geography, period, or independent source was supplied. This is a management claim about the fuel-cell niche, not about all new data-center power. S6
Do not report “~98% of data-center power.” That would be a category error. Grid supply, gas turbines, reciprocating engines, renewables, batteries, and other sources dominate the broader market.
The only defensible all-market sizing bridge
Management cited 30–40 GW of new AI data-center capacity expected to turn on in 2027. Dividing the 2 GW annual-capacity target by that range yields a 5.0–6.7% theoretical capacity equivalent. S6
- This is not reported market share.
- It assumes full utilization of the 2 GW target.
- It overstates data-center share because Bloom also serves commercial and industrial customers.
- It ignores timing differences and any capacity added beyond 2 GW.
Bottom line: current public evidence supports “about 2 GW/year of target manufacturing capacity,” not “2 GW/year of proven production.” Bloom may control nearly all fuel-cell deployments into data centers, but its share of total new data-center power is not publicly disclosed. A rough capacity-based ceiling is only 5–6.7% of management's 2027 AI-power build estimate.
Five quarters: the operating curve turned before the narrative became comfortable
Revenue and GAAP operating margin
Fallback: revenue in $M / operating margin — Q2'25 401/–0.9%; Q3'25 519/1.5%; Q4'25 778/11.3%; Q1'26 751/9.6%; Q2'26 1,065/17.1%.
Adjusted EPS: actual versus consensus mean
Fallback: actual/consensus — Q2'25 0.10/0.018; Q3'25 0.15/0.102; Q4'25 0.45/0.299; Q1'26 0.44/0.128; Q2'26 0.78/0.407.
| Metric ($M) | Q2'25 | Q3'25 | Q4'25 | Q1'26 | Q2'26 |
|---|---|---|---|---|---|
| Revenue | 401.2 | 519.0 | 777.7 | 751.1 | 1,065.4 |
| GAAP gross profit | 107.1 | 151.7 | 239.9 | 225.5 | 355.6 |
| GAAP operating income | –3.5 | 7.8 | 87.5 | 72.2 | 182.2 |
| Net income to common | –42.6 | –23.1 | 1.1 | 70.7 | 196.3 |
GAAP operating margin is the profitability line because it best captures recurring operating leverage; net income is more volatile due to interest, equity-method income, derivatives, and taxes. S1S2
Cash and balance sheet: improved, but working capital still tells a project story
Cash
$2.67B
High flexibility after convertible issuance. S1
Q2 2026 FCF
$174.8M
$226.4M CFO less $51.6M capex. S2
Inventory
$758.2M
+$114.9M since December. S1
AR + contract assets
$886.4M
Rapid growth still absorbs working capital. S1
Cash conversion
- CFO was $300.0 million in H1 2026; derived FCF was $222.2 million. S1
- Customer deposits + deferred revenue rose to $445.0 million from $143.8 million at year-end 2025.
- Even with $211.7 million of Q2 inflow from deposits/deferred revenue, other operating lines produced an approximate $49.9 million net use.
- The CFO offered a “$375 million-plus” CFOA baseline on the call, but that is CALL-ONLY, not formal guidance. S6
Debt, dilution, and customer economics
- Cash exceeds recourse + nonrecourse debt by about $188.9 million, before financing obligations and leases.
- The Oracle warrant/inducement had $324.4 million of aggregate value; $17.9 million had reduced revenue cumulatively, including $5.0 million in Q2. S1S10
- $306.5 million remains on the balance sheet to reduce future revenue as the contract is executed.
- The consideration does not invalidate the order, but it makes the customer's economics less comparable with an unincentivized sale.
Stock reaction: the beat encountered an even larger stock of expectations
Daily price since January 2025
Fallback: peak close of $345.85 on June 22, 2026; $302.70 on June 30; $254.29 on July 8; $166.84 before the reaction; $163.75 on July 29.
Reading the tape
June 22
$345.85 at the period's peak close. The market was already pricing power scarcity and data-center dominance. S8
June 30
Brookfield expands its framework from $5 billion to $25 billion. It is financing capacity, not backlog or contracted revenue. S9
July 8–9
A short report challenges supply chain/scandium; Bloom responds in an 8-K and claims visibility for 25 GW/year without disclosing suppliers, inventory, or cost. S12S14
July 29
After the print, BE opened at $183.50 (+10.0%), touched $185.66 (+11.3%), and closed at $163.75 (–1.9%). S8
Inference: the gap reversal suggests that valuation, positioning, and credibility dominate the near-term headline. But this was not an isolated rejection: QQQ fell 2.0%, VRT 17.2%, and GEV 4.6% in a broad AI/power selloff; BE roughly matched QQQ and materially outperformed the adjacent names. Even after a 52.7% decline from the June 22 peak, forward multiples remain premium. S15
Concentration and disclosure: the filing's biggest risk arrived the next day
What the 10-Q/A corrected
On July 29, Bloom amended its 10-Q to correct a transposition between “three months” and “six months” in customer risk. The correct figure is one unrelated contractual customer = ~73% of Q2 revenue; for H1, two customers represented ~44% and ~21%, with the second a related party. The financial statements did not change. S1
Three customers also represented 36%, 34%, and 17% of accounts receivable. U.S. revenue was 90% of total.
Backlog is not one measure
Filed performance obligations include $442.4 million of product/installation and $51.7 million of service. They exclude short contracts and right-to-invoice arrangements, so they are not the same as disclosed commercial backlog. S1
In Q4 2025, Bloom cited ~$20 billion of total backlog and ~$6 billion of product backlog, 2.5× YoY. Q2 did not provide a fresh quantified backlog; the call only said backlog was growing faster than revenue. S13S6
Correct reconciliation: a potentially diversified backlog can coexist with extremely concentrated current revenue. Contractual protection becomes credible to shareholders only through deposit, contract-asset, inventory, and revenue conversion without slippage.
Call Q&A: what analysts tried to break
Quartr/Fiscal.ai provider transcript; not company-certified. An official replay exists, but no official textual transcript was found. Answers below are paraphrased and identified as management claims where appropriate. S6
Mark Strouse · JPMorgan → K.R. Sridhar
Does “validated” mean customer, order, or deployment?
The CEO declined to break out how many hyperscalers operate, purchased, or only signed agreements. He said capacity would not constrain the visible order book and pipeline.
Why it matters Technical validation is not revenue. Gate: MW/GW delivered, deposits, and converted backlog.
Chris Dendrinos · RBC → K.R. Sridhar
Is the Brookfield framework contracted demand?
Management called it a “financial shelf” and gave no utilization schedule. Third-party diligence helps validate the technology and execution, but it does not create revenue on its own.
Falsifier No new JVs, funded capital, or purchased assets.
David Arcaro · Morgan Stanley → Edwards / Sridhar
What happens when a data center slips?
Bloom claims protected contracts, redeployable equipment, and a financier still obligated to take delivery; guidance would not depend on one project.
Tension Terms were not quantified, and one customer generated 73% of Q2. Watch inventory, contract assets, and deposits.
David Arcaro · Morgan Stanley → K.R. Sridhar
Does scandium constrain expansion?
The company says supply is sufficient, visibility supports 25 GW/year, and it is not dependent on China; inventory, contracts, and supplier details are proprietary.
Unverified Treat this as a company claim until quantitative evidence appears in a filing.
Nick Amicucci · Evercore ISI → Simon Edwards
How much incremental profit converts to cash?
The CFO indicated a $375 million-plus CFOA baseline and expects 100% of the $175 million increase in operating-income midpoint to convert.
CALL-ONLY This is not formal guidance. AR, contract assets, inventory, and deposits test the promise.
Nick Amicucci · Evercore ISI → K.R. Sridhar
Does training or inference sustain demand?
Management sees demand in both and cited $12–24 billion of annual customer revenue per GW plus 35–40 GW of AI capacity in 2027.
No methodology These are management estimates, not independent data; hyperscaler capex is the falsifier.
Sunaina Ocalan · Bernstein → K.R. Sridhar
What is the real competition in data-center fuel cells?
The CEO claimed a “very high 90s” share of fuel cells in data centers without supplying a denominator, period, geography, or source.
Claim This is not total data-center-power share. Track competing MW/GW and losses in new projects.
Colin Rusch · Oppenheimer → K.R. Sridhar
Do pricing and service economics improve structurally?
Management emphasized value pricing, a 22% service margin, and claimed that 80% of 2025 orders came from repeat customers.
Proof Sustain service above 20% and contain performance-guarantee payments and reserves.
Read-throughs: real validation, but different contracts carry different economics
Oracle · up to 2.8 GW
The master agreement adds visibility; an initial 1.2 GW is contracted and is expected to deploy through 2027. The $324.4 million warrant/inducement shows that this demand carried economic and dilutive consideration. S10S1
DemandConcessionBrookfield · $25 billion
Project financing lowers friction and signals bankability. Without a draw schedule or committed projects, however, it is not revenue, backlog, or guaranteed conversion. S9S6
ValidationOptionalityNebius / IDF / Oaktree · $1.7 billion
The structure combines Oaktree, Morgan Stanley tax equity, and MUFG senior debt, reinforcing the ability to finance a behind-the-meter alternative for neoclouds. It also exposes Bloom to those customers' funding cycle. S11
Project financeCounterpartyValuation: the market is still paying today for a large portion of 2028
Basic market cap
$48.23B
Adjusted diluted EV
$50.49B
Proxy: 323.3M shares; converts treated as equity.
2026 EV / revenue
12.3×
Against $4.12B consensus. S7
2026 / 2027 P/E
60.4× / 33.7×
Mean adjusted EPS of $2.71 / $4.86. S7
Per-share value scenarios
Fallback: Bear $76.9 nominal / $58.4 present value; Base $171.1 / $130.1; Bull $292.6 / $222.4.
Assumptions—not consensus price targets
| 2028 | Bear | Base | Bull |
|---|---|---|---|
| Revenue | $8.0B | $9.98B | $12.0B |
| EBITDA / margin | $1.60B / 20% | $2.94B / 29.4% | $3.84B / 32% |
| EV / EBITDA | 14× | 18× | 24× |
| Nominal value/share | $76.9 | $171.1 | $292.6 |
| PV at 12% / 2.42 years | $58.4 | $130.1 | $222.4 |
| Company · NTM | Logic | EV / revenue | EV / EBITDA | P/E |
|---|---|---|---|---|
| BE | SOFC / AI onsite power | 9.83× | 41.70× | 48.77× |
| VRT | AI power and cooling | 5.85× | 23.68× | 32.62× |
| GEV | Power and grid equipment | 4.66× | 27.33× | 42.61× |
| CEG | Power / nuclear producer | 3.25× | 13.36× | 22.49× |
| VST | Power producer | 2.70× | 9.40× | 15.61× |
NTM multiples from Fiscal.ai on July 29, 2026. These are adjacent names, not pure comparables; BE's premium reflects faster growth but measures multiple-compression exposure. S7
Valuation conclusion: the $163.75 share price is close to the nominal value of the 2028 base case before discounting for time and risk. An attractive return requires something close to the bull case—$12 billion of revenue, a 32% EBITDA margin, and a still-premium multiple—or a materially lower entry price.
Method: the fully diluted proxy uses 323.331 million average diluted shares, treats converts as equity, adds financing obligations + nonrecourse debt, and subtracts cash. Scenarios add roughly $2.458 billion of adjusted net cash and divide by the same diluted count. Fiscal.ai 2028 consensus is $9.98 billion of revenue and $2.94 billion of EBITDA. This is analyst modeling, not company guidance or personalized advice. S1S7
Next gates: what proves it, what kills it, and when to look
What proves the thesis
- H2 2026 revenue near the $2.23 billion midpoint.
- ~34.7% H2 non-GAAP gross margin without another tariff benefit.
- At least $430.6 million of H2 non-GAAP operating income.
- At least $375 million of annual CFOA—remembering this is a call framework.
- Service ≥20%, warranty stabilization, and conversion of new customers.
What kills the thesis
- Margin returns toward 30% after removing tariffs.
- Concentration or slippage forces another inventory/contract-asset build.
- Warranty reserves and payments continue to escalate.
- Supply-chain/scandium evidence contradicts claimed visibility.
- AI capex slows or project financing fails to convert into equipment.
Catalyst windows
- Late October 2026, estimated: Q3; date not yet announced.
- Across 2026–27: initial 1.2 GW Oracle deployment.
- Next filings: customer mix, warrants, deposits, warranty, capacity, and supply chain.
- Early 2027, estimated: FY 2026 close and final proof of the guide.
Action discipline: for new money, the better posture is to wait for proof or price. An entry near the base-case present value (~$130) improves the skew; alternatively, revisions that lift 2028 EBITDA above ~$3.8 billion with clean margin and lower concentration could justify paying more. Without one of those two conditions, business growth and stock returns are not the same thesis.
Sources, evidence, and limitations
Priority: official filings/releases/decks for numbers; a provider transcript for call language; Fiscal.ai for consensus; Yahoo Finance for price. Management statements are identified as claims when they cannot be independently verified. No portfolio model, prior internal note, or user position was available in the workspace for reconciliation.
S1
Bloom Energy 10-Q/A, July 29, 2026
P&L, balance sheet, cash flow, corrected concentration, Oracle warrant, warranty, tariffs, and performance obligations.
P&L, balance sheet, cash flow, corrected concentration, Oracle warrant, warranty, tariffs, and performance obligations.
Primary · SEC
S2
Q2 2026 earnings release, Exhibit 99.1
Results, non-GAAP reconciliations, quarterly cash flow, and guidance.
Results, non-GAAP reconciliations, quarterly cash flow, and guidance.
Primary · company
S3
Primary · company
S6
Quartr/Fiscal.ai transcript, July 28, 2026
Prepared remarks and Q&A. Provider transcript; not company-certified.
Prepared remarks and Q&A. Provider transcript; not company-certified.
Secondary · call
S7
Fiscal.ai, NYSE-BE — overview and estimates, retrieved July 29, 2026
Revenue, EPS, EBITDA, FCF, and price-target consensus. Historical intraday estimate timestamps unavailable.
Revenue, EPS, EBITDA, FCF, and price-target consensus. Historical intraday estimate timestamps unavailable.
Market data
S8
Yahoo Finance chart API
Daily price from January 1, 2025 through July 29, 2026; observed at Jul 29, 2026, 4:00 PM ET.
Daily price from January 1, 2025 through July 29, 2026; observed at Jul 29, 2026, 4:00 PM ET.
Price · secondary
S9
Primary · partnership
S10
Bloom / Oracle agreement up to 2.8 GW
Commercial agreement and deployment; warrant economics in the 10-Q/A.
Commercial agreement and deployment; warrant economics in the 10-Q/A.
Primary · partnership
S11
Project announcement
S12
Bloom 8-K response on scandium, July 9, 2026
Company response; supply-chain claim not independently verified.
Company response; supply-chain claim not independently verified.
Primary · claim
S13
Primary · company
S14
Hunterbrook short report on Bloom
Declared short party; allegations are not treated as facts and are paired with the company's filed response.
Declared short party; allegations are not treated as facts and are paired with the company's filed response.
Short seller · conflict
S15
Associated Press market close, July 29, 2026
Broad selloff context; security-specific returns checked against market data.
Broad selloff context; security-specific returns checked against market data.
News · market
S16
Bloom Energy 2025 10-K
Fremont annual-capacity expansion from approximately 1 GW to 2 GW by year-end 2026.
Fremont annual-capacity expansion from approximately 1 GW to 2 GW by year-end 2026.
Primary · SEC
S17
Bloom Energy 2026 proxy materials
1.8 GW cumulative deployed base, roughly 1,100 sites, and capacity-expansion context.
1.8 GW cumulative deployed base, roughly 1,100 sites, and capacity-expansion context.
Primary · SEC
Material limitations: no official textual transcript; no disclosed actual annual output in GW, total data-center-power share, backlog by customer, count of customers in operation, quantified contractual protections, Brookfield draw schedule, scandium inventory/supplier/cost detail, or exact quarterly allocation of the warranty reserve. Consensus may include post-print revisions and should be read as a snapshot.