Built by Grok 4.6 High

Reports Tuesday, October 27, 2026. Prepared Thursday, September 24, 2026. This is the Q3 2026 print (quarter ending September 30). It has not been reported. The date is FMP and the repo calendar, not a Bloom scheduling release. Time of day is not confirmed. The historical pattern is an after-close call. Last reported quarter is Q2 2026, released after the close on July 28, 2026. Bloom is a calendar-year company. The homepage calendar labels this date 2026Q4. That label is the announcement quarter. This page is 2026Q3.

BE | Earnings Preview

Bloom reports Q3 on Tuesday, October 27, if the estimated date holds. There is still no company release locking the hour. On July 28 the company printed revenue of $1,065.4 million, up 165.5% versus $401.2 million, and non-GAAP diluted EPS of $0.78, up 680% versus $0.10. Non-GAAP gross margin was 34.3%, up 610 bps versus 28.2%. The same call raised the year again. That range is still the only guide.

Growth trajectory. Annual revenue slowed from +23.3% in 2022 ($1,199.1 million) to +11.2% in 2023 and +10.5% in 2024, then reaccelerated to +37.3% in 2025 ($2,024.0 million). Quarterly, the reacceleration is sharper and is already the thing the year guide assumes will stop. Revenue growth went from +57.1% in Q3 2025 to +35.9% in Q4 2025, then +130.4% in Q1 2026 and +165.5% in Q2 2026. Product did the work: +208.4% in Q1 and +215.4% in Q2, and it was 87.8% of the last quarter ($935.4 million of $1,065.4 million). Service sped up too, from +15.6% to +26.8%, off a small base. Electricity is still shrinking.

The street's Q3 revenue of $1,058.3 million is +103.9% versus $519.0 million. That is a deceleration of about 6,160 bps from the last print, back toward — and still well above — the year-ago growth rate. It is also only about $7 million under the Q2 dollar level. The harder compare is part of it: Q3 2025 was already $519.0 million against a Q2 2025 base of $401.2 million. The rest is the year itself. First-half 2026 revenue is $751.1 million plus $1,065.4 million, or $1,816.4 million, +149.8% versus a first half of 2025 of $326.021 million plus $401.242 million, or $727.3 million. The $4.05 billion midpoint leaves $2,233.6 million for the second half, +72.2% versus second-half 2025 of $519.048 million plus $777.683 million, or $1,296.7 million. That is about 7,760 bps slower than the first half. Two more quarters at the Q2 revenue level would finish near $3.95 billion, just above the $3.9 billion floor and about $100 million short of the midpoint.

Margin is expanding year over year and is not what the guide asks to expand further. Full-year non-GAAP gross margin went from 25.8% in 2023 to 28.7% in 2024 and 30.3% in 2025. Product gross margin did not: FY2025 was 36.1%, −160 bps versus FY2024 37.7%. Q4 2025 consolidated gross margin was the air pocket, 31.9% versus 39.3%, −740 bps, which management tied to project mix. Q2 2026 then printed 34.3% and product margin 37.2%, +290 bps versus Q2 2025 product 34.3%. One quarter is not a five-year trend. The July hold at ~34% says keep the Q2 rate. First-half blended margin was about 33.1%. Landing 34% on the $4.05 billion midpoint needs a second-half rate near 34.7%. A second half that simply repeats 34.3% lands the year near 33.8%, which is the "approximately 34%" they are guiding. Simon Edwards said he will give up a point of quarterly margin to expedite for a customer who stays. A Q3 print a bit under 34% can still fit the year.

What consensus is underwriting. FMP, pulled September 24, has Q3 revenue of $1,058.3 million (10 estimates, $999.0–$1,107.8 million) and EPS of $0.679 (10 estimates, $0.622–$0.792). Versus Q3 2025 that is +103.9% on revenue and +353% on non-GAAP EPS against $0.15. FY2026 FMP revenue of $4,089.8 million is +1.0% versus the $4.05 billion midpoint and inside $3.9 billion–$4.2 billion. FY2026 FMP EPS of $2.699 sits on the $2.70 midpoint of $2.55–$2.85. FMP does not label EPS as GAAP or non-GAAP. The company guide is non-GAAP, and FMP's historical actuals match non-GAAP diluted, not GAAP. Internal models, Visible Alpha, and Bloomberg were not available. There is no quarterly company number to beat. The October 27 bar is the street.

The contradiction management keeps repeating. KR Sridhar says demand is compounding, the constraint is the customer's site, and 2026 revenue does not depend on any single project. The guide agrees with the second half of that sentence and not the first: the year they just raised still slows from about +150% in the first half to roughly +61% to +84% in the second. The street's FY2026 number agrees with management. It never left the range. The street's FY2027 revenue of $6,834.4 million, +67.1% versus the FY2026 street number, does not have that protection. Oracle's September 24 force majeure on Project Jupiter is the first public fact that can test the "already in the algorithm" line. It threatens the 2027 conversion, not this quarter, unless the call drops the no-single-project sentence.

What to watch on October 27

• Whether Q3 lands near $1.06 billion or slips. The FY low can absorb a lumpy quarter. If Q4 later matches the street's $1,223.8 million, Q3 could print near $860 million and $3.9 billion would still be intact. KR's standing line is that slips will not change the annual guide.
• Non-GAAP gross margin versus 34.3% and versus the ~34% year. One point given away to expediting is allowed. A run of quarters under 34% is not. Profit is the easier bridge: two more quarters at Q2 non-GAAP operating income of $239.6 million would land on the $850 million midpoint. Revenue is the one that still needs a step-up from a flat Q2.
• Customer concentration. The July 29 10-Q/A says one customer, explicitly not the related party, was about 73% of Q2 revenue. The six-month note is milder: two customers at about 44% and 21%, the second a related party. Validation of "all the major U.S. hyperscalers" is not the same as recognized revenue.
• Any walk-back of "not dependent on any single project," and whether backlog dollars are updated. Product backlog of $6.0 billion, +140% versus $2.5 billion, and service backlog of $14.0 billion, +45.8% versus $9.6 billion, are year-end 2025 figures. "Growing faster than revenue" has no new number.
• Tariffs, scandium, and cash. The July hold at 34% did not add a tariff or a metal-price caveat. A guide cut, or a gross-margin miss they do not attribute to expediting, would be the tell. Formal cash-flow guidance was withdrawn. The verbal "$375 million plus" of operating cash flow is not a promise. Adjusted EBITDA of $650 million–$800 million was not reaffirmed. Do not score FMP's $986.9 million FY EBITDA against that abandoned band.

TSM on October 15 is the only clean demand print before Bloom. EQT on October 20 is a gas read, not a fuel-cell comp. GEV, Vertiv, Enphase, and the hyperscalers report the same day or after. A weak TSM print is the one external excuse available before October 27. It would not answer what share of the $3.9–$4.2 billion is already in the Q3 shipment plan.

Bloom does not guide the quarter. On the Q3 2025 call KR refused a Q4 pinpoint because a project can fall on either side of December 31. Nothing on the Q1 or Q2 2026 calls reversed that. The only live outlook, set July 28 and unchanged through September 24, is full-year 2026.

Classification: in line, not a sandbag, and conservative only at the low end. The process is the conservative one. FY2025 finished $2,024.0 million, +15.7% versus the $1.75 billion midpoint and above the high. FY2024 revenue finished inside the range, −1.7% versus the midpoint, so they do not beat every midpoint. The only recent quarterly comment — Q2 revenue "at least as good as Q1" — was a floor of $751.1 million. Q2 printed $1,065.4 million, +41.8% above that floor. That does not mean the current range is still 10% light. FMP is sitting on it. First-half revenue is 44.8% of the $4.05 billion midpoint. The high end explicitly includes in-year bookings that are not yet converted. Street second-half revenue of $1,058.3 million plus $1,223.8 million is $2,282.1 million, +2.2% versus the $2,233.6 million residual midpoint and inside the residual range.

Items said and then removed from the formal page: February's capex range of $150–$200 million, the Q1 cash-from-operations figure of $200 million, and the Q1 adjusted-EBITDA band. Edwards said the supplemental deck was aligned back to the metrics they actually guide, and that an oral baseline of about $375 million-plus of operating cash flow is not formal guidance. First-half adjusted EBITDA is already $143.0 million plus $253.4 million, or $396.4 million, so the old $650–$800 million band is stale. H1 non-GAAP EPS is $0.44 plus $0.78, or $1.22. The $2.70 midpoint leaves about $1.48 for the second half. Street Q3 plus Q4 EPS ($0.679 + $0.834) is $1.513, the same slightly-above-midpoint shape as revenue.

Completed years, versus the last numeric guide: 2024 revenue in range and profit above the high; 2025 revenue, margin, and operating income all above the high. They have not yet had to defend a cut. History says they would rather raise again than walk a high number back. That makes another raise possible. It does not make the current midpoint a lowball versus a consensus that has already moved inside it.

Q3 2026 has not been reported. Segment consensus does not exist in FMP. Internal is n/a on every row. Percent difference is (consensus − guide mid) / guide mid, and it is n/a wherever the company did not guide.

Current quarter — Q3 2026

Drivers sit above consolidated revenue because this is an on-site solid-oxide business that recognizes product on delivery and acceptance. Service is the ratable line. None of the four is guided.

If Q2's non-product run-rate of about $130 million held, a $1,058 million quarter would put product near $930 million, about +142% versus $384.3 million. That is an implication, not a street number. It would still be a clear deceleration from +215% product growth, and it would be roughly flat versus Q2 product of $935.4 million.

Eight quarters — the trajectory into this print

Columns are calendar quarters. Growth is year over year, same quarter prior year. Q3 2026 is not a column.

Revenue dollars in that table are millions. The growth rate decelerated into Q2 2025 (+19.5%), then reaccelerated for four straight prints, with the step from Q4 2025's +35.9% to Q1 2026's +130.4% the break. Margin expansion reaccelerated with it, from +280 bps in Q1 to +610 bps in Q2, after the Q4 mix air pocket. Service margin expansion is narrowing as the base improves: +1,320 bps in Q1 to +980 bps in Q2, with the level at 22.0%. That is the "20% plus" Edwards said they expect to sustain, not a new guide. Domestic mix moved from 59% in Q2 2025 to 90% in Q2 2026, +3,100 bps. That is the data-center mix shift. Current plus non-current deferred revenue was $327.1 million plus $117.9 million at Q2 2026 ($445.0 million), versus $60.0 million plus $47.6 million a year earlier ($107.6 million), +313.6% year over year. It is a timing buffer, not a hidden backlog, next to a $1 billion quarter. Megawatts accepted were last indexed in 2024. Do not invent a run-rate.

FQ+1 guide (expected) — Q4 2026

A Q4 range is not expected. The annual range is the disclosure. Italic below means "they will not put out a Q4 band," not a forecast of the print.

Street Q4 at +57.4% is another step down from the Q3 street's +103.9%. That is what a flat-to-up second half looks like after a $1,065 million Q2, not a demand break. What the call is more likely to touch is the FY2026 range: reaffirm $3.9–$4.2 billion, or raise it if Q3 clears the second-half run-rate.

FY+1 guide (old) — FY2027 is unguided

Bloom has not issued FY2027 guidance. On the Q4 2025 call KR said they have not given long-term guidance and would not let a supplier's growth comment be attributed to the company. The Q2 2026 call did not add a 2027 framework. "Old" means that status is unchanged. The standing FY2026 tape, already in the table above, is the only guide this call can update.

Five years

The FY2026 street at $4,089.8 million is about +102% versus FY2025. That is a further acceleration of the annual growth rate, delivered by a second half that grows much more slowly than the first half already did. Service margin is the clean multi-year fundamental, from −18.0% in 2023 to 22.0% in Q2 2026. It is not what this print will be graded on. First-half 2026 product revenue of about $1,588.8 million already exceeds FY2025 product of $1,531.3 million.

The baseline is the July 28 call, Simon Edwards's second as CFO. Revenue and non-GAAP operating income were raised, with the new revenue floor above the old ceiling of $3.8 billion. Gross margin was held. The formal cash figure was taken out. Nothing between July 28 and September 24 walks those numbers back. Daloopa's 2026Q3 document index returned no filings. The IR events calendar, current as of September 23, lists no upcoming event.

Tone across four calls. Q3 2025 was confident and deliberately vague: 2025 would be "better" than the stated guide, and they would not pinpoint Q4. February 2026 was the first formal year guide, with visibility capped at about six months and an explicit refusal of a supplier's long-term growth rate. April was Edwards's first call and the most numeric raise, "barring any global shock" on the 34% margin, plus a one-off Q2 revenue floor. July was faster demand language and less willingness to be pinned down: not one project, will trade a margin point for speed, explicitly agnostic on whether AI capex keeps accelerating, and no longer giving capacity figures. CEO Sridhar is the same speaker across all four. The CFO seat changed before the April raise, not into this quarter. That is stability of the raise, not a new-CFO reset.

Assumptions still under the range. Two layers, both already inside $3.9–$4.2 billion: signed backlog converting against site-readiness dates, plus capacity held back for in-year book-and-ship. A quarter that is merely flat with Q2 is not, by itself, a guide cut. Servers are described as fungible. Edwards said the financier, not the end user, is on the hook to take the equipment, and that a quarter can look concentrated because one or two campuses hit their delivery window. Brookfield's framework, from $5 billion to $25 billion in June, is a financing shelf. KR would not put a window on drawing the extra $20 billion. It is not a 2026 revenue line. Operating expense is supposed to grow much slower than revenue. Q2 revenue was up about 166% and operating expenses were up 48%, on the call. The opex ratio cut from 15% to 13% is the full-year version of that claim.

What got more specific, and what left the page. More specific: the dollar range, twice; operating leverage, now a stated ~21% margin at the midpoint; service profitability as a lasting level; the scandium answer, limited to three sentences (enough recoverable supply, visibility on 25 gigawatts, not dependent on China). Less specific: the quarter, which they will not guide; the gross-margin trajectory, which stopped rising; cash, which left the printed guide; capacity, which left the gigawatt vocabulary used in April. Guidance accuracy on completed years supports the management test on profit — 2024 and 2025 non-GAAP operating income both finished above the high — and is mixed on revenue. They have not had to defend a cut.

Since the guide. MiTAC (August 6) and Power Connect (August 19) add sites and a faster install, not a disclosed gigawatt order. The September 16 800-volt paper is a company model, not a result. S&P 500 inclusion effective September 21 changed the holder base and not the backlog. Project Jupiter is the project the market will try to put inside a guide management says does not depend on one project. The April relationship was described at up to about 2.45 gigawatts on that campus, inside a master agreement of up to 2.8 gigawatts. The gas lateral's targeted in-service date moved to February 1, 2027. On September 24 Oracle sent a force majeure notice so it can delay payments if the campus misses a 2028 online date. Oracle said the project remains on its planned schedule. Bloom has not updated the range. A 2028 campus is not a mechanical hole in Q3 dollars. It does test the portfolio claim if more than one large 2026 delivery window moves, because the midpoint still requires a second half above a flat Q2 run-rate.

The lead-plaintiff deadline on the scandium complaint is September 28. The alleged class period ends July 8, twenty days before the raise. Unproven. Treat it as a disclosure overhang, not a change to 34% or to the revenue range, unless they say otherwise in October.

The catalyst that pays this quarter is still product revenue at a triple-digit year-over-year rate with non-GAAP gross margin held near 34%. A print near $1.06 billion and an unchanged full-year guide would confirm the July raise. It would not be new information. The upside left in the quarter is narrow: product growth decelerates by less than the Q3 base effect implies, or they raise $3.9–$4.2 billion again. The downside that would matter is a cut, a margin slip they do not attribute to expediting, or any sentence that ties 2026 revenue to Jupiter, the Wyoming conditions, or the Nebius "this year" date.

Scope is non-earnings news since the July 28 report, through September 24. The commercial tape is more sites and a faster install, not a new disclosed gigawatt order. Nothing in the window updates the July 28 guide.

Industry, not a Bloom release: on August 7 the U.S. Department of War announced a conditional loan for an Australian scandium project, with first production targeted for 2028 and a named offtake that is not Bloom. It does not de-risk Q3 supply. Plaintiff-firm reminders on Nevins v. Bloom Energy are one case, class period February 27, 2025 through July 8, 2026, lead-plaintiff deadline September 28. Unproven, and not a new operating disclosure. Pre-window deals that later articles recycle — Brookfield to $25 billion on June 30, the Oracle expansion on April 13 — are context, not news in this window.

On the FMP series Bloom is mixed over 12 reported quarters and a consistent beater over the last 4. Non-GAAP diluted EPS was a beat in 8 of 12, a miss in 3, and exact in-line once (2024 Q2, −$0.06 on both sides). Revenue was a beat in 8 of 12 and a miss in 4. The last four reported quarters, 2025 Q3 through 2026 Q2, are 4 of 4 on both lines. Dollar surprise is improving, not deteriorating. Average EPS surprise went from about +$0.01 (oldest four) to +$0.05 to +$0.22 (last four). The last two prints were +$0.32 and +$0.39. Average revenue surprise went from −$19 million to −$6 million to +$168 million. Percentage EPS surprise is not the trajectory. A +$0.09 beat against a penny estimate in 2025 Q2 prints as +726% and looks larger than the +$0.39 / +100% beat in 2026 Q2. It was not.

FMP epsActual matches non-GAAP diluted EPS in all 12 quarters. It does not match GAAP. GAAP diluted was still a loss through 2025 Q3, $0.00 in 2025 Q4, $0.23 in 2026 Q1, and $0.62 in 2026 Q2, against non-GAAP of $0.45, $0.44, and $0.78. The street comparison is the non-GAAP line, which is also the basis of the annual EPS guide.

2025 Q1 revenue is the row that flips. FMP retains a $384.6 million estimate, a −15.2% miss on $326.0 million. The print-day bar was near $293 million, and on that bar it was a beat. Scored on FMP it stays a miss. If that one row uses the print-day bar, the 12-quarter revenue rate is 9/12. Direction on the other revenue rows, and on every EPS row except a one-cent argument in 2024 Q2, agrees with the print-day checks.

The misses that were real were acceptance timing, not a lost year. 2023 Q4 was a Korea delivery reset. 2024 Q1 was the warning that acceptances would slip, and the print landed inside management's own range and outside the street's. 2024 Q3 was the last clean double miss, and management said Q4 would put the second half back. It did: 2024 Q4 beat, and that started a six-quarter EPS beat streak. From late 2025 the explanation changed from catch-up to data-center deliveries running ahead of an outlook they had just published. 2026 Q2 is the proof the beat was not a stale model. The estimate had reset to $0.39 and $826.1 million. Actual non-GAAP EPS of $0.78 still doubled it, and revenue of $1,065.4 million beat by about $239 million, a wider dollar gap than Q1. The same call raised the year again.

The bar into October 27 is no longer the stale one that produced +255% and +39% surprises. FMP is at $0.679 and $1,058.3 million. What the record supports is a beater whose misses were timing, whose last four prints all cleared the street, and whose dollar beat was still getting larger in July after the street had moved. What it does not support is extrapolating the Q1 percentage. A timing slip can still miss a single quarter. Those misses were followed by a catch-up quarter, and the last two catch-up quarters were raises.

Cells are the FMP surprise. Color follows dollars for EPS and percent for revenue, using #1E8449 and #C0392B. Q3 2026 is not on the map.