| Oct 22, 2026 | -11.24% | Q3 FY2026 | Read transcript briefingQuarter in one view- Third consecutive record quarter and third guidance raise of the year: revenue $1.24B (+138% y/y, +16% q/q), total GM 34.5% (+412 bps y/y), non-GAAP operating income $286M (+500%+ y/y, 23.1% margin), adjusted EBITDA $299M (24% of revenue), non-GAAP EPS $0.94 (GAAP $0.78 vs −$0.10 a year ago).
- Full-year guide raised again: revenue $4.4–4.6B (~122% growth at midpoint, from $3.9–4.2B), op income $960M–1.03B (from $800–900M; 22% margin at midpoint vs 14% at the original February guide), EPS $3.10–3.40 (from $2.55–2.85), GM held at ~34%. Implied Q4 revenue: ~$1.35–1.55B.
- The cash story inverted: CFO was a use of $27M and FCF a use of $181M in the quarter (inventory, contract assets, receivables build), against +$42M FCF YTD; cash fell to $2.5B from $2.7B. Capex $153M in Q3, $231M YTD — 4x all of 2025. This is the first quarter where the pulled FCF guidance looks consequential rather than cosmetic.
- Headline commercial event: AEP revealed 700 MW of carbon-capture-enabled Bloom servers for Google ("near net zero," first-of-its-kind) — resolving the AEP offtake that slipped past its Q2 deadline unaddressed last quarter. Also: US Army conditional award via partner Indelible, first units shipping to Texas, S&P 500 inclusion.
- Brookfield and IDF — last quarter's centerpiece — were not mentioned once. Backlog "expanding faster than revenue" repeated, still with no figures, and a direct analyst request to quantify it was refused.
What management is focused on- TAM broadening beyond hyperscalers is the dominant narrative shift: C&I (survey of 200+ businesses across 15 segments: 25% of US C&I facilities expect full on-site power by 2030, 35% by 2035), utilities/front-of-the-meter (rate-basing Bloom servers as T&D assets via IRPs), military/critical infrastructure (Army/Indelible), and international (local partner discussions, "good news" promised soon). IEA +40% global demand by 2035 and $3T→$6T electricity market frame the pitch.
- CCS elevated from vision to named deal: the 700 MW AEP/Google project is K.R.'s proof that "combustion technologies do not offer an affordable and scalable pathway to carbon capture" and Bloom does; he reiterated a "greater than trillion dollar" CCS opportunity and said hyperscalers are now asking about CCS optionality in conversations.
- Speed-to-power economics, now sourced externally: K.R. cited NVIDIA's CEO (Berlin GTC) that one quarter of delay on a 1 GW data center = $10B of missed revenue ($1B per 100 MW per quarter) — an escalation of last quarter's in-house $12–24B/GW-year math, now attributed to the industry's most important voice.
- Capacity as a "dial," 5 GW language restored: the Fremont real-estate expansion is framed as preparing for "what is after" the 5 GW milestone, which K.R. says "we can clearly see that coming." No timeline, no post-5 GW number (Strouse's >10 GW question unanswered), and K.R. explicitly denied ever giving quarterly capacity-add figures when Kallo referenced 250–300 MW/quarter.
- "Self-funded model" as the financial identity: Edwards repeatedly framed the business as self-funding — operations funding 4x capex — despite the negative-CFO quarter. He directed investors to full-year cash generation, not quarterly extrapolation.
- Token-cost-of-electricity is now the fully articulated pricing doctrine: LCOE → delivered grid power → rack power → token power cost, with Bloom claiming to be "the overwhelming cost leader" on the all-in basis.
Key numbers and quarter mechanics- Revenue $1.24B (+138% y/y, +16% q/q). Product $1.07B (+179% y/y, +15% q/q) — first $1B+ product quarter, 87% of mix. Service $67.6M (+15% y/y, ~flat q/q). Installation $84.9M (+29% y/y, +66% q/q) — Edwards flagged installation as timing/mix-driven and deliberately decoupled from product.
- Margins: product GM 38.4% (+250 bps y/y, +120 bps q/q) achieved "in a quarter when we were both ramping new production lines and expediting deliveries"; service GM 21.4% (+700 bps y/y, second straight quarter above 20%); total 34.5%. GM track record cited: 28.7% (2024) → 30.3% (2025) → 34.5% (Q3 2026); original 2026 guide was ~32%.
- Opex $141M (+26% y/y vs revenue +138%); op margin 23.1% (+60 bps q/q, from 8.9% a year ago). YTD revenue $3.05B is already 50% above all of 2025; Q3 op income alone exceeded full-year 2025 op income.
- Cash mechanics (the quarter's key disclosure): product revenue recognized on control transfer; billing follows contract milestones and "can occur in a different quarter" — the gap sits in contract assets/receivables. Edwards: working capital investment will continue; Rusch's contract-asset question got "they're gonna turn" with turnover as "our primary focus" but no aging, balance, or collection timeline.
- Guidance mechanics unchanged: two layers (backlog assessed customer-by-customer vs site readiness + probability-weighted in-quarter pipeline); where Q4 lands in the range depends on "the timing of a small number of deliveries around year end."
- Not disclosed: backlog dollars/MW, Brookfield/IDF contribution or utilization, inventory balance, customer prepayments, capacity GW figure beyond the 5 GW milestone, capex per GW, tariffs, Oracle warrant status, revenue concentration.
Product and launch scorecard- AEP/Google CCS (700 MW) — the quarter's flagship and the first commercial CCS evidence ever offered: named utility, named hyperscaler, named MW, "near net zero." Caveats: disclosed by AEP, not Bloom; no revenue timing, pricing, or Bloom contract size given; Wyoming referenced by an analyst as a "100 MW class approaching gigawatt class" site (700 MW) without management confirming project geography.
- Nebius (New Jersey, 300 MW) — first hard MW figure attached to this relationship: servers "ready for the 300 MW," planning approvals in 3 months, first power block ready 2 months after approvals; delivery proceeds block-by-block paced to customer construction. Strong time-to-power evidence; revenue recognition timing not stated.
- US Army / Indelible — conditional award of long-term leases for Bloom-powered generation at Army installations. Genuinely new vertical (defense/critical infrastructure) but "conditional" and partner-mediated; no MW, dollars, or timeline.
- Utility front-of-the-meter channel — IRP inclusion cited as already happening; K.R. expects "significant wins" and frames rate-basing as the unlock. No named utility order beyond AEP.
- Texas — first units shipping this quarter; evidence of competitiveness in lowest-cost power states. No size given.
- 800V DC — quantified claim upgraded: $5B saved over first 5 years of a 1 GW data center vs AC architecture (per Bloom's September report); "in conversations with all the hyperscalers"; timing explicitly deferred to customers ("not if, how soon"). Still no named deployment or revenue.
- CHP — repositioned as a C&I vector (steam/cooling needs; ~20% power-demand reduction via cooling). Still no commercial order.
- CCS economics — new specificity: hyperscalers would accept a 10–15% cost premium for zero-carbon power and "Bloom will be able to deliver in that range, all in"; monetization deliberately agnostic (partner bolt-on, end-to-end shared proceeds, or customer-kept credits). "Carbon-capture-ready" expected to become a standard purchase feature.
- International — "engaged with potential local partners in a few markets," wins promised "soon." Nothing signed or named.
- Silent: Brookfield, IDF/Oaktree consortium, Oracle Jupiter warrants, Korea/SK ecoplant, hydrogen, scandium, tariffs.
Sell-side read-through- Strouse (JPMorgan): Fremont expansion rationale and whether footprint now implies >10 GW — K.R. confirmed preparation "past the milestone of the 5-gigawatts" but gave no number or timeline. GM follow-up drew Edwards' "measured outlook" framing: startup costs and expediting now, scale benefits later — a soft yes to 2027 GM upside without commitment.
- Amicucci (Evercore): pressed the negative-CFO quarter for puts/takes or one-offs — got none; the answer was philosophy ("self-funding," judge full-year) rather than decomposition. His utility-TAM question produced the rate-basing/front-of-the-meter thesis, the most substantive new-market answer of the call.
- Arcaro (Morgan Stanley): commercial status of CCS/CHP/800V — produced the CCS uniqueness claim, the $5B 800V savings figure, and CHP-as-C&I-vector; all adoption timing deferred to customers.
- Dendrinos (RBC): is sub-90-day time-to-power holding as backlog grows? Answer was indirect — Nebius block-by-block mechanics and the 300 MW readiness claim, no general lead-time number. Automation follow-up got the 600-employees-at-100MW→1GW→5GW productivity claim.
- Gupta (UBS): LCOE vs real cost — produced the full token-cost cascade. His water question yielded a garbled, internally inconsistent number set (CCGT "25 thousand gallons" vs Bloom "1.4 million," then "not even a drop in the ocean") — directionally clear (Bloom uses vastly less water) but the figures as transcribed cannot be relied on.
- Kallo (Baird): capacity-add cadence (250–300 MW/quarter) — flatly denied as ever given; "dial" only. International/NIMBY question got the fuel-efficiency argument (higher gas prices abroad widen Bloom's advantage) and a tease of near-term news.
- Thakkar (BMO): largest operating US site — "100 megawatt class already," with Wyoming "approaching the gigawatt class because 700 MW." First on-record sizing of the deployment base.
- Mandloi (Mizuho): directly asked how to interpret "backlog growing faster than revenue" against ~140% revenue growth — refused ("I am not going to directionally bracket it"), deferred to year-start 2027 guidance. The most pointed evasion of the call.
- Ocalan (Bernstein): CCS in hyperscaler conversations — "significant interest from all our hyperscale customers," capture-ready systems as near-term default, 10–15% acceptable cost premium.
- Rusch (Oppenheimer): contract-asset conversion — "they're gonna turn," no metrics; CCS monetization — model-agnostic answer.
- Notable: no analyst asked about Brookfield or IDF one quarter after they headlined; no questions on Oracle warrants, tariffs, inventory, or prepayments. The operator's mid-Q&A shift to one-question-only limited follow-up pressure.
Management credibility- The raise pattern is now three-for-three in 2026 (op income: $425–450M original → $600–750M → $800–900M → $960M–1.03B), and Q3 delivered $1.24B against an implied H2 that required ~$2.1–2.4B — the under-promise/over-deliver discipline K.R. explicitly invoked when deflecting Mandloi. Q4 implied at $1.35–1.55B is a modest sequential step (+9–25%) versus what Q3 just did, so the guide looks conservative again.
- The cash-flow warning from last quarter's bear case materialized on schedule: FCF guide pulled in Q2, then Q3 prints −$27M CFO / −$181M FCF with cash down ~$200M. Edwards' "self-funding business" framing is defensible on YTD (+$42M FCF) but the refusal to give Amicucci any puts/takes or one-offs for a negative quarter — while raising everything else — is a disclosure-quality step back.
- 5 GW capacity language restored after last quarter's conspicuous retreat ("the 5-gigawatt milestone... we can clearly see that coming"), and the Fremont expansion gives it physical evidence. But K.R.'s flat denial of ever giving capacity-add numbers (against Kallo's 250–300 MW/quarter reference) continues the pattern of capacity disclosure on management's terms only.
- AEP delivered — the slipped Q2 deadline that went unmentioned last quarter resolved into the largest named project in company history (700 MW, Google, CCS). That converts a credibility question into a credibility deposit, though Bloom let AEP make the announcement and offered no contract economics.
- Backlog assertion now formally unverifiable by choice: repeated in prepared remarks, then a direct quantification request refused. With 2027 guidance promised at year-start, management has set its own accountability deadline.
- Brookfield/IDF silence one quarter after a fivefold shelf expansion and a "meaningful contributor" quarter is a real omission — no utilization update, no revenue recognized, no third partner. Combined with financier-mediated concentration disclosed last quarter, the financing layer is now a black box.
- High-confidence unaudited assertions continue: 600 employees constant from 100 MW to 5 GW (10x productivity), "overwhelming cost leader" on token cost, CCS deliverable within a 10–15% premium, >$1T CCS TAM, water-use figures that were internally contradictory as stated.
What changed versus the prior quarter- Guidance raised a third time: revenue $4.4–4.6B (from $3.9–4.2B), op income $960M–1.03B (from $800–900M), EPS $3.10–3.40 (from $2.55–2.85); GM held at ~34% for the third straight guide.
- Cash flow flipped negative: Q3 CFO −$27M and FCF −$181M vs +$226M/+175M in Q2; cash $2.5B vs $2.7B; capex accelerated to $153M/quarter (4x 2025's full-year spend YTD). Working capital (inventory, contract assets, receivables) is now the stated swing factor, with an explicit commitment to keep investing it.
- AEP resolved — spectacularly: from an unmentioned missed deadline to a revealed 700 MW CCS project for Google; CCS moved from assertion to named commercial deal, with 10–15% cost-premium economics disclosed for the first time.
- Nebius quantified: 300 MW with servers ready, vs last quarter's unnamed-MW "meaningful contributor" framing.
- Brookfield/IDF went from centerpiece to zero mentions; no financing-shelf update of any kind.
- Capacity disclosure partially restored: 5 GW milestone re-embraced plus Fremont expansion announced; last quarter's ambiguity ("2 GW" uncorrected) superseded — but no post-5 GW figure and a denial of quarterly add-rate guidance.
- New verticals opened: defense (Army/Indelible), utility front-of-the-meter/rate-basing, Texas entry, C&I survey data (25% by 2030), international partner tease — a deliberate narrative broadening beyond hyperscalers.
- Service margin held >20% (21.4%, second consecutive) but service revenue growth is only +15% y/y and flat q/q — the annuity is profitable, not yet fast-growing.
- Time-to-power math escalated and outsourced: NVIDIA CEO's $10B/quarter/GW figure replaced Bloom's own $12–24B/year framing as the cited authority.
- Prior watchlist items still open: Oracle Jupiter warrants, backlog dollars, tariffs, inventory/prepayment detail, Brookfield utilization.
Bull case- Three raises in three quarters with Q4 implied at only +9–25% sequential against a Q3 that grew +16% — the guide again looks sandbagged, and management's two-layer mechanics (backlog conversion + probability-weighted in-quarter pipeline) have now been validated at $1.24B scale.
- The AEP/Google 700 MW CCS deal is a category-creating proof point: it resolves the AEP overhang, names the most demanding customer class, opens the "trillion-dollar" CCS TAM with claimed deliverable economics (10–15% premium), and differentiates structurally against turbines/engines that "do not offer an affordable and scalable pathway to carbon capture."
- Margin structure is proving durable under stress: 38.4% product GM while ramping new lines and expediting; opex +26% vs revenue +138%; 23.1% op margin; Edwards signals further scale benefits into 2027 as startup costs roll off.
- Demand breadth is now multi-vector with evidence: hyperscalers (Nebius 300 MW executing block-by-block), utility channel (AEP, IRP inclusion, rate-basing path), defense (Army award), C&I (survey-quantified), Texas (low-cost-state competitiveness), international (partner talks). Single-project dependence keeps falling.
- Capacity conviction is now backed by real estate and 4x capex, funded YTD from operations; the 5 GW milestone is re-committed and management is visibly preparing beyond it.
- Speed-to-power remains the unanswerable pitch, now with NVIDIA's CEO supplying the $10B/quarter/GW urgency math and Nebius (3-month permitting, 2-month first block) as the live demonstration.
- S&P 500 inclusion broadens the holder base and is external validation of sustained profitability.
Bear case- Cash conversion is now a demonstrated problem, not a hypothetical: −$27M CFO and −$181M FCF in the strongest revenue quarter in company history, cash down $200M, and management declined to decompose the quarter for one-offs while repeating "self-funding." Contract assets and receivables are growing with revenue recognized ahead of billing — quality of revenue timing now depends on milestone structures investors cannot see.
- The guide's Q4 dependence on "a small number of deliveries around year end" concentrates the full-year print on a handful of acceptances — the same control-transfer/billing-lag mechanics that just produced a negative-cash quarter.
- Brookfield/IDF silence removes the financing-uptake narrative at exactly the moment working capital absorption rises; if financier-mediated demand slowed, the call gave no window into it.
- Backlog opacity is now explicit policy: "growing faster than revenue" asserted while refusing to bracket it even directionally, with same-year book-and-ship mechanics making backlog duration structurally unassessable from outside.
- The AEP/Google deal, for all its scale, came with zero Bloom-side economics — no revenue timing, no margin, no contract structure (financier? direct? PPA?) — and CCS at a 10–15% all-in premium is an assertion with no delivered project behind it.
- Capacity spend is escalating (4x capex, Fremont expansion) against deliberately vague capacity numbers and a denied add-rate; if demand pauses, the "dial" has never been tested downward.
- Service revenue — the supposed annuity — grew only 15% y/y and was flat sequentially; the mix is 87% product, so the model remains a hardware ramp with hardware-ramp working-capital physics.
- High-confidence claims (600 employees to 5 GW, token-cost leadership, water figures as garbled) remain unverifiable, and the water answer's internal inconsistency suggests loose quantitative discipline on non-financial metrics.
Next-quarter watchlist- 2027 guidance (promised at year-start): the first full forward-year framework — revenue growth rate, GM trajectory after Edwards' "measured outlook" hint, and whether backlog finally gets a dollar/MW figure. K.R. has staked credibility on this print.
- Cash flow: whether Q4 collections convert the contract-asset/receivable build (Edwards' "they're gonna turn") and full-year CFO lands anywhere near the old informal ~$375M+ baseline; inventory and prepayment balances in the 10-Q; whether FCF guidance is restored for 2027.
- AEP/Google 700 MW: Bloom-side contract disclosure — revenue recognition timing, counterparty structure (direct vs financier), margin, CCS scope and schedule; whether Wyoming is confirmed as the site.
- Brookfield/IDF: any utilization, revenue-through-financier disclosure, or third consortium partner; two consecutive silent quarters would mirror the Oracle Jupiter pattern.
- Nebius 300 MW: pace of block deliveries and revenue recognition; whether the "we will not be the bottleneck" claim holds as blocks scale.
- Oracle Jupiter: warrants, committed MW, status — now multiple quarters unaddressed.
- Capacity: Fremont expansion timeline/capex, any post-5 GW target, and whether the 5 GW milestone date is ever specified.
- Utility/defense/international: first named front-of-the-meter rate-based order, Army/Indelible conversion from conditional to firm, and the teased international wins.
- CCS: whether "carbon-capture-ready" becomes a disclosed order attribute and whether the 10–15% premium claim gets project-level support.
- Service: third consecutive >20% margin quarter and any reacceleration in service revenue growth beyond +15%.
|
| Jul 28, 2026 | -1.85% | Q2 FY2026 | Read transcript briefingQuarter in one view- First-ever $1B+ quarter: revenue $1.065B (+166% y/y, +42% q/q), GM 34.3% (+604 bps y/y), non-GAAP operating income $240M (22.5% margin, +1,536 bps), adjusted EBITDA $253M (~24%), non-GAAP EPS $0.78 (GAAP $0.62), OCF $226M, FCF $175M, cash $2.7B.
- Full-year guidance raised again, one quarter after the last raise: revenue $3.9–4.2B (from $3.4–3.8B; ~100% growth at midpoint), GM held at ~34%, non-GAAP operating income $800–900M (from $600–750M; ~21% margin at midpoint), EPS $2.55–2.85 (from $1.85–2.25). Edwards also revealed the original start-of-year op-income guide was $425–450M — the third upward revision of that line this year.
- Brookfield expanded its financing framework fivefold, from $5B to $25B (announced June), after ~9 months of observed execution. A second financing consortium — Industrial Development Funding with Oaktree, MUFG, and Morgan Stanley — reached $2.6B cumulative commitment and was "a meaningful contributor to the quarter" via the Nebius deal.
- Nebius named as a customer that cancelled combustion orders and switched to Bloom; IDF is purchasing the servers on cash terms against identified sites and delivery schedules.
- K.R. claims all major U.S. hyperscalers plus 12+ neoclouds/AI labs/colos have "validated and approved" Bloom; backlog "growing at a faster pace than revenue" (stated twice, no figures).
- Free cash flow guidance was pulled from the supplemental presentation — Edwards framed it as aligning disclosure to "where we truly guide," then gave an informal ~$375M+ CFOA baseline.
What management is focused on- The "four frictions" framework — capital, community, permitting, speed — is the new organizing narrative; Brookfield/IDF financing shelves are presented as the capital-friction solution arranged "in advance" of gigawatt demand.
- "Bloom will not be your bottleneck": capacity added continuously in "Copy Exact increments" ahead of committed orders; capacity planning driven by a "sophisticated algorithm" against 30–40 GW of 2027 AI data center turn-ons. Management refused to give capacity numbers ("that's all we are going to comment").
- Book-ship-revenue inside the same fiscal year is now a stated, repeated mechanic — new customers not in year-end backlog will ship this year; guidance is built in two layers (backlog conversion + reserved capacity for in-year bookings).
- Fungibility as risk mitigation: servers redeployable site-to-site; 2026 guidance "not dependent on any single project" — a direct response to reported project-development challenges at large AI sites (Arcaro's question).
- Time-to-power economics quantified for the first time: a 1 GW data center generates $12–24B/year revenue for a full-stack AI provider; one month of earlier power = $1–2B of customer revenue. This is the value-based pricing justification.
- Service business elevated: 22% margin (from −21% at IPO), 80% of 2025 orders were repeat orders — K.R. explicitly framed service as "a big driver to enterprise value."
- Long-term vision restated: DC power as primary, CHP (heating + cooling), carbon capture, 90%+ combined fuel efficiency, appliance-like deployment across data centers, EV fleet charging, residential microgrids.
Key numbers and quarter mechanics- Revenue $1.065B (+166% y/y, +42% q/q); product revenue $935M (+215% y/y, +43% q/q), ~90% of total — mix is heavily product-weighted.
- GM 34.3% (+604 bps y/y); product GM 37.2% (+193 bps q/q, +291 bps y/y); service GM 22% (+977 bps y/y, fifth straight double-digit quarter). Management committed to sustaining 20%+ service margins "over the long term" while flagging stack-replacement timing as a quarterly swing factor.
- Operating income $240M (22.5% margin); opex grew 48% vs revenue +166% — Edwards called the leverage "structural, not a 1 quarter effect" (fixed R&D/G&A base, automation in support functions).
- OCF $226M (+$439.5M y/y); FCF $175M; cash $2.7B (up from $2.52B).
- Guidance mechanics: H1 delivered ~$1.82B; the $3.9–4.2B guide implies H2 of ~$2.1–2.4B — continued sequential ramp but far less back-loaded than last quarter's guide implied.
- Edwards on cash conversion: ~100% of the $175M op-income raise drops to CFOA; "~$375M plus" as the new informal CFOA baseline. Formal FCF guidance withdrawn.
- Margin trade-off disclosed: management will sacrifice "a point of margin in a given quarter" to expedite delivery for strategic customers — a new, deliberate caveat on quarterly GM variability.
- Not disclosed: backlog dollars/MW (only directional "growing faster than revenue"), capacity GW figure, inventory, capex, customer prepayment balances, Brookfield revenue contribution, tariffs, Oracle/Jupiter status.
Product and launch scorecard- Brookfield expansion ($5B → $25B): the quarter's flagship commercial event. Evidence offered: Brookfield watched 9 months of execution and interviewed customers (oldest 15+ years) before quintupling. Caveats: it is a financing "shelf," not orders; deployment pace "depends on uptake"; no revenue timing attached.
- IDF/Oaktree/MUFG/Morgan Stanley consortium ($2.6B cumulative): first quantified contribution — "meaningful contributor to the quarter" via Nebius offtake, cash terms, identified sites and schedules. This is the most concrete third-party-funded deal mechanics disclosure to date.
- Nebius: named as a turbine/engine canceller that switched to Bloom — the first named "abandoned alternative" customer, though K.R. claimed "several" such customers this year.
- Hyperscaler validation: "all major U.S. hyperscalers" + 12+ neoclouds/labs/colos validated — but Strouse's attempt to split using/shipped/definitive-agreement was declined; no names, no MW.
- 800V DC: referenced as an existing capability within the value proposition ("Bloom's ability to provide 800-volt DC power") and as the future primary architecture — still no named DC-native deployment or revenue.
- CHP/absorption chilling: back in the vision statement ("use the heat to do both heating and cooling," 90%+ combined efficiency) — still no commercial order disclosed.
- Carbon capture: "ready for carbon capture" cited as a pricing/value attribute; "Bloom is able to do carbon capture better than anybody else" — assertion without a named project.
- Scandium: addressed via blog/8-K — claims sufficient economic supply "to power the planet," visibility for 25 GW of deployments, no China dependence; everything else "proprietary." First time supply-chain risk on a critical input got a quantified (if unauditable) answer.
- Silent: Oracle/Project Jupiter status and warrant execution (not mentioned once), AEP offtake (the Q2 finalization deadline passed unaddressed), Korea/SK ecoplant, hydrogen, international, tariffs.
Sell-side read-through- Strouse (JPMorgan): tried to decompose "validated" into using/shipped/backlog — refused. Capacity follow-up also deflected ("that's all we are going to comment"), though K.R. volunteered the 30–40 GW 2027 industry turn-on frame.
- Dendrinos (RBC): hyperscaler diligence process (NDA-level capacity reviews as the "validation" gate); Brookfield $20B execution timing — answer: it's a shelf, pace depends on uptake, no window given.
- Arcaro (Morgan Stanley): the sharpest risk question — exposure to headline project delays. Edwards: MSAs + Copy Exact redeployability + "the financier is on the hook to take delivery of the equipment from Bloom" — a significant contractual-protection disclosure. K.R. added guidance has no single-project dependence. Scandium follow-up got the 25 GW/no-China answer.
- Amicucci (Evercore): caught the pulled FCF guidance — Edwards' "alignment" explanation plus the ~$375M CFOA baseline was the most newsworthy financial disclosure in Q&A. Inference question drew the $12–24B/GW-year revenue math and the distribution-grid ("surface streets") inference argument.
- Kallo (Baird): competitive supply-demand — K.R. conceded all fast power technologies will have a place near-term, then argued "total cost to tokens" (not LCOE) as the winning metric. Chinese/open-source models: Jevons Paradox answer — cheaper tokens mean more power demand.
- Gupta (UBS): 3–4 year product vision — DC-native, CHP, carbon capture, appliance form factor. Qualitative; no milestones or dates.
- Mandloi (Mizuho): capacity capex inflation — K.R. claimed factory ROI of "a few months," distinguishing Bloom from industrial-age capex; no dollar figures. Notably, Mandloi's framing referenced expansion "from 2 gigawatts" — the 5 GW claim from last quarter was not repeated by management on this call.
- Ocalan (Wolfe, per transcript): fuel-cell competition (molten carbonate) — K.R. claimed "very high 90s" percent share of data-center fuel-cell market; deflected sizing competitors.
- Rusch (Oppenheimer): pricing evolution and displacement count — K.R. restated value-based pricing, declined displacement quantification, and pivoted to the service-margin celebration (closing the call).
- Notable: no analyst asked about Oracle/Jupiter, AEP, Brookfield revenue recognition, inventory, or tariffs — the accountability gaps from prior quarters persist despite Brookfield being the headline.
Management credibility- Delivered: Q2 came in at $1.065B vs "at least as good as Q1" ($751M) — massively above the floor; the raised H2 guide from last quarter is already being superseded upward. The pattern of under-guiding and raising continues, now three op-income revisions in a year ($425–450M → $600–750M → $800–900M).
- Edwards' second call was materially more substantive than his debut: deal-model walkthrough (CapEx sale vs PPA/capacity/lease; financier as the revenue-counterparty of record), concentration-is-timing explanation, CFOA conversion math. This is a disclosure-quality upgrade — but note it also clarifies that reported "customer" concentration can be financiers, not end users.
- New inconsistency to check: last quarter's "5 GW footprint" capacity claim was not repeated; Mandloi's question referenced "2 gigawatts" and management did not correct it. Capacity disclosure has gone from specific (5 GW) to deliberately non-specific in one quarter.
- FCF guidance withdrawal is a credibility negative dressed as simplification — pulling a forward metric while raising everything else invites the question of whether working capital (prepayments, inventory build "ahead of the ramp") is less favorable than implied. The ~$375M CFOA baseline partially offsets.
- "Backlog growing faster than revenue" (repeated for emphasis) with no backlog figure is an unverifiable assertion — especially notable since book-and-ship-same-year mechanics should mechanically shrink backlog duration.
- Brookfield fivefold expansion is a genuine third-party validation deposit — but it is financing capacity, not purchase orders, and management conflated the two rhetorically ("capital of that quality does not follow press releases... it follows firm bankable orders" — while disclosing no order detail).
- "Very high 90s" data-center fuel-cell share, "ROI of a few months" on factories, "25 GW of scandium visibility" — all unaudited assertions delivered with high confidence and no supporting data.
- Oracle Jupiter — last quarter's defining win — went entirely unmentioned. Either it is progressing silently or something changed; the omission itself is the signal to track.
What changed versus the prior quarter- Scale step-change: first $1B+ quarter; revenue +42% sequentially; product revenue now ~90% of mix (vs $653M/$751M last quarter).
- Guidance raised for the second consecutive quarter: revenue to $3.9–4.2B (from $3.4–3.8B), op income to $800–900M (from $600–750M), EPS to $2.55–2.85 (from $1.85–2.25); GM held at ~34% and is now being delivered in-quarter (34.3%) rather than promised for H2.
- Brookfield moved from three quarters of silence to the centerpiece: $5B → $25B expansion plus a second consortium (IDF/Oaktree/MUFG/MS, $2.6B) with named end customer (Nebius) and cash-terms mechanics.
- Revenue-model transparency upgraded: Edwards explained financier-as-customer accounting, lumpiness/rotation of quarterly concentration, and contractual protections (financier obligated to take delivery).
- FCF guidance withdrawn; informal ~$375M+ CFOA baseline substituted.
- Capacity disclosure regressed from "5 GW footprint" to no number, with a new "sophisticated algorithm" framing and fungibility/redeployment as the risk answer.
- Service margin stepped up to 22% (from 18%) with a 20%+ long-term sustainability commitment; 80% repeat-order statistic disclosed for 2025.
- Scandium supply risk addressed publicly for the first time (25 GW visibility, no China dependence).
- Oracle/Jupiter, AEP offtake (deadline was this quarter), tariffs, inventory, and backlog dollars all went unaddressed — several prior watchlist items remain open.
Bull case- The raise-raise pattern now has two data points in two quarters, with H2 implied at ~$2.1–2.4B against visible mechanics: backlog conversion plus reserved capacity for in-year book-and-ship — and Q2 already proved the model at $1.065B.
- Operating leverage is compounding, not one-time: opex +48% vs revenue +166%, op margin 22.5%, and management committed to opex growth staying "well below" revenue growth — the $800–900M op-income guide (~21% margin) looks internally consistent.
- Financing is no longer a constraint: $25B Brookfield shelf + $2.6B IDF consortium + "more in the wings" removes the customer-capex friction for pay-over-time deals, and Brookfield quintupled after auditing execution — the strongest external validation since Oracle Jupiter.
- Demand breadth claims widened: all major U.S. hyperscalers validated, 12+ neoclouds/colos, named turbine-cancellation win (Nebius), "several" alternative-solution abandonments, backlog growing faster than revenue.
- Service annuity inflecting: 22% margin, 20%+ long-term commitment, 80% repeat orders, 10–15 year data-center durations (per last quarter) — a compounding base under product cyclicality.
- Contract structure de-risks delays: financiers obligated to take delivery, equipment redeployable across sites, guidance explicitly built to absorb project push-outs.
- Time-to-power value math ($1–2B per month per GW for customers) supports pricing power and the 37%+ product margin without needing LCOE competitiveness.
Bear case- The FCF guide was pulled the same quarter everything else was raised — with inventory being built "ahead of the ramp" and prior quarters' prepayment tailwinds, cash conversion deserves scrutiny; ~$375M CFOA baseline is informal and low relative to $850M op income.
- Backlog opacity deepened: "growing faster than revenue" asserted twice with no dollars, MW, or duration — and same-year book-and-ship mechanics mean backlog quality/duration is structurally harder to assess.
- Oracle Jupiter vanished from the narrative one quarter after being the defining win — no status, no committed MW, no warrant execution; combined with the unmentioned AEP Q2 deadline, two prior catalysts are unaccounted for.
- Capacity disclosure went backward (5 GW claim not repeated; "2 GW" framing uncorrected) while management insists capacity won't constrain — the capex cost of "Copy Exact increments" remains undisclosed against a "few months ROI" assertion.
- Revenue concentration is now explicitly financier-mediated: IDF was a "meaningful contributor" in Q2, and quarterly leaders rotate — reported customer concentration may mask true end-demand concentration, and Brookfield/IDF-related revenue is not broken out.
- Product mix at ~90% of revenue with GM trade-offs ("we will sacrifice a point of margin to expedite") makes quarterly margins more volatile even if the ~34% full-year rate holds.
- "Very high 90s" share claims and validation counts are unverifiable; molten carbonate and other fuel-cell entrants were dismissed without sizing.
- The guide still requires H2 revenue roughly 15–30% above H1 — execution risk remains back-loaded, and management's own framing bakes in assumed project slippage offset by unnamed "out of the blue" orders.
Next-quarter watchlist- Oracle Jupiter: any status update, committed vs "up to" MW, revenue timing, warrant execution — two consecutive silent quarters would be a red flag.
- AEP offtake: the promised Q2 2026 finalization passed without mention — confirm met, slipped, or dropped.
- Cash flow: whether H1 CFOA (~$300M) tracks the ~$375M+ full-year baseline; inventory build "ahead of the ramp," prepayment balances, and capex in the 10-Q; whether FCF guidance is restored or stays withdrawn.
- Backlog: any dollar/MW refresh to the ~$6B product / ~$14B service figures to substantiate "growing faster than revenue"; disclosure of non-Oracle hyperscaler names.
- Brookfield/IDF: pace of shelf utilization, revenue recognized through financier counterparties, related-party/concentration disclosures in the 10-Q, and any third financing partner.
- H2 ramp: Q3 revenue trajectory vs the implied ~$2.1–2.4B H2; GM holding ~34% while management trades margin for speed; product margin vs 37.2%.
- Capacity: whether the 5 GW figure is re-asserted, revised, or stays retired; capex per GW of additions; any new factory announcement.
- Service: sixth straight double-digit margin quarter and whether 20%+ holds through stack-replacement timing; any backlog-duration disclosure supporting 10–15 year contracts.
- Scandium: whether the 25 GW supply visibility claim gets third-party or filing support; any cost impact.
- Competitive: first evidence of molten carbonate or other fuel-cell wins in data centers against the "high 90s" share claim; further named turbine/engine cancellations.
|
| Apr 28, 2026 | +27.21% | Q1 FY2026 | Read transcript briefingQuarter in one view- Record Q1: revenue $751.1M, GM 31.5% (+~280 bps y/y), operating income $129.7M (vs $13.2M), adjusted EBITDA $143M (vs $25.2M), EPS $0.44 (vs $0.03), OCF +$73.6M — first-ever positive Q1 operating cash flow. Management claims "greater than 100% y/y growth," but the transcript also states "up 13.4% year-over-year" — one figure is a transcription error (100%+ is consistent with the "first quarter of >100% growth" claim); verify against the press release.
- Full-year 2026 guidance raised materially: revenue $3.4–3.8B (from $3.1–3.3B; ~80% growth at midpoint), GM ~34% (from ~32%), non-GAAP operating income $600–750M, EPS $1.85–2.25. Note: the op-income guide resolves last quarter's incoherent "$125M to $475M" transcript artifact — the real range is far above 2025's $221M.
- The quarter's defining event: Oracle's Project Jupiter — an up-to-2.45 GW AI factory in New Mexico — will be powered 100% by Bloom, replacing previously planned gas turbines and diesel backup. Announced by Oracle "last night." Described as one of the largest islanded microgrids in the world.
- New permanent CFO: Simon Edwards (ex-Grok, per transcript), joined two weeks ago — ends the ~9-month acting-CFO arrangement.
- Capacity posture escalated again: current footprint supports 5 GW/year; capacity now added "continuously, hundreds of megawatts a quarter"; new factories beyond 5 GW framed as normal course.
What management is focused on- "Becoming the standard for on-site power" now anchored to the Oracle Jupiter sole-source win — K.R. frames it as the lighthouse playbook executing: "where Oracle is going is where the broader market is headed."
- Demand diversification claim: "well more than half of current data center backlog comes from other hyperscalers, neo clouds and colocation providers" — i.e., Oracle is not the whole story, though no names or sizes given.
- Community acceptance elevated to a core selling pillar: air quality, minimal/no water use, no noise, no resident rate increases — explicitly credited as one of two reasons Oracle pivoted away from turbines/diesel. The Rhode Island analogy (2.5 GW CCGT = state's capacity, ~1M showers/day of water, car-equivalent pollution) is the new rhetorical vehicle.
- "Not order constrained and not capacity constrained" — revenue pace is set by customers' greenfield construction speed, not Bloom's supply. This is a deliberate inversion of the turbine industry's 2029+ delivery backlog narrative.
- Installation innovation disclosed for the first time: skid-mounted solution yielding "an order of magnitude reduction in field time" — claimed 100 MW projects deployable faster than any competing technology.
- Inference as the next leg: K.R. agreed inference will be "much bigger than training" in total GW, distributed at the edge near dense populations — where clean/permitting attributes matter most.
- "Bridge power" declared dead: hyperscaler conversations about bridging to grid "nonexisting today."
Key numbers and quarter mechanics- Q1: revenue $751.1M; product revenue $653.3M (all-time high, up y/y and sequentially); service revenue $61.9M (+15.6% y/y); GM 31.5% (+~280 bps); product margin 35.3% (+22 bps); service margin 18% (+13 pts y/y; fourth straight double-digit quarter, ninth straight profitable); operating income $129.7M (17.3% margin, +1,300 bps); adjusted EBITDA $143M (~19% margin); EPS $0.44.
- OCF +$73.6M — first positive Q1 ever, driven by profitability, collections, and customer prepayments to reserve capacity (prepayments are a new, notable cash mechanic).
- Cash $2.52B — flat vs Q4's $2.5B; no new financing mentioned this quarter.
- Guidance mechanics: Q2 revenue "at least as good as Q1"; raised guide implies H2 acceleration (Q1 $751M + Q2 ≥$751M leaves $1.9–2.3B for H2).
- Service contracts: 100% attach reaffirmed; data-center service durations now stated at 10–15 years on average (vs ~6–7 years cited previously) — a material lengthening of the annuity stream if accurate.
- Capacity numbers are commercial product only; service capacity baked in separately (per K.R.'s answer to Mandloi).
- Not disclosed: backlog dollar update, MW figures, inventory level, capex, book-and-ship share, Brookfield/related-party revenue, tariffs. The 5 GW capacity claim and "hundreds of MW per quarter" additions are unaudited management assertions.
Product and launch scorecard- Oracle Project Jupiter: the largest single commercial event in company history — up to 2.45 GW, 100% Bloom, replacing turbines AND diesel backup, islanded microgrid. Sole-source status validates the "no grid, no diesel, no batteries, no turbines" architecture. Caveats: "up to" 2.45 GW (committed portion undisclosed), no revenue timing, no pricing, and the previously disclosed Oracle warrant agreement was not mentioned — execution status still unknown.
- 800V DC: K.R. asserts adoption is "inevitable" (copper and transformer scarcity forcing the shift) and claims current deployments already run with "no grid, no diesel, no batteries" using fuel cells plus ultracaps. Still zero disclosed DC-native revenue or named customer commitments — Gupta's channel-check framing went unconverted into a commercial data point.
- Battery-free load following: now implicitly deployed — "100% Bloom one-stop solution" with ultracaps replacing batteries is presented as shipping reality in current projects, an upgrade from last quarter's "lab demonstrations."
- Absorption chilling/CHP: not mentioned this quarter — back into silence after Q4's revival.
- Utilities channel: new color — favorable regulation/rate-basing driving interest from both gas and electric utilities; AEP not named but "largest utility deal" referenced by Kallo.
- International: still delayed ("a pause before it takes off"), now attributed to Russia/Europe and Qatar gas dynamics; US remains ~the whole story.
- Silent items: Brookfield (third quarter), AEP offtake finalization (was promised for Q2 2026 — not addressed), hydrogen, Korea/SK ecoplant, tariffs, inventory.
Sell-side read-through- Strouse (JPMorgan): service contract duration — extracted the 10–15 year data-center duration disclosure, the most concrete new annuity-economics fact of the call. Edwards declined long-term margin guidance.
- Arcaro (Morgan Stanley): pricing power — K.R. refused to benchmark vs turbines ("apples and oranges," value-based only); supply chain — claimed flat shop-floor headcount from 200 MW to ~10x that via automation/upskilling, supplier speed bumps possible but "absolutely not" worried.
- Dendrinos (RBC): pressed the capacity-doctrine change (customer-called → continuous) — answer: secular AI demand conviction, "analog dial not digital step function"; beyond 5 GW requires new factories, treated as normal course.
- Amicucci (Evercore): teed up the inference TAM argument (accepted and amplified); bridge-power question produced the "bridge to a bridge" dismissal — grid-tied need declared moot.
- Gupta (UBS): again the most technically loaded questions (ultracaps vs batteries; 800V DC conversion-gear shortages) — K.R. jokingly offered him a job; the mutual-credibility dynamic from prior quarters continues, but answers remained qualitative.
- Kallo (Baird): channel evolution (utilities, reshoring) and international timing; cost/stack-life follow-up got the standard "genius of and," double-digit cost-downs reaffirmed.
- Rusch (Oppenheimer): installation cadence — surfaced the skid-mount/order-of-magnitude field-time reduction, which management admitted it had never disclosed; project-level economics participation question went unanswered (second half of his question was ignored).
- Mandloi (Mizuho): operating leverage at 5 GW and service capacity mix — answered only with the product-vs-service capacity clarification; no leverage quantification.
- Bagri (Citi): price elasticity vs CCGT premium (CCGT pricing +10–20% YTD) — K.R. rejected the framing entirely ("we don't think about price, cost, elasticity") and converted it into closing remarks.
- Notable: no analyst asked about Brookfield, related-party revenue, inventory, tariffs, AEP offtake timing, or the Oracle warrants — the accountability threads remain dropped for a second consecutive quarter.
Management credibility- Delivered: the Q4 guide ($3.1–3.3B) was raised within one quarter with the low end above the prior high end — consistent with the pattern of conservative guidance, though it also means the prior guide was stale on arrival.
- Resolved: the incoherent 2026 operating income range from last quarter's transcript is superseded by a clean $600–750M; the permanent CFO hire closes a ~9-month governance gap.
- New inconsistency to check: Q1 revenue growth stated as both ">100%" (twice, including "first quarter of greater than 100% growth in company history") and "up 13.4%" — one is a transcript error; verify against the 8-K. Also "gross margin from 30% in 2025" vs the 30.3% FY figure — minor.
- Service duration claim (10–15 years for data centers) contradicts prior ~6–7 year characterizations — could reflect mix shift or loose language; checkable against service backlog duration disclosures.
- The 5 GW capacity claim is new and large (vs the 2 GW signal two quarters ago) with no capex figure attached this quarter — the "capital-light" assertion needs the capex line to verify.
- Oracle Jupiter is a genuine credibility deposit: named customer, named site, named capacity, announced by the customer itself — the strongest third-party validation to date. But "up to 2.45 GW" leaves the firm committed portion undisclosed, and the unexecuted warrant agreement was not revisited.
- Edwards' debut was scripted and safe; he deferred long-term guidance and took no risky positions — appropriate but uninformative.
- Brookfield silence is now a three-quarter pattern; the missed European year-end project remains unaccounted for.
What changed versus the prior quarter- Guidance raised materially one quarter after being set: revenue +$300–500M at the ends, GM +2 pts to ~34%, op income to $600–750M (from the garbled prior range), EPS initiated at $1.85–2.25.
- Oracle moved from unexecuted warrants and unspecified "projects" to a named, sole-source, up-to-2.45 GW deployment — the largest disclosed win in company history.
- Capacity doctrine: from "2 GW signal" and "continuous decision" to a stated 5 GW footprint capability with "hundreds of MW per quarter" additions and new factories as normal course.
- CFO: permanent hire (Simon Edwards) after ~9 months acting.
- Service economics: data-center contract durations now quoted at 10–15 years (vs 6–7 previously); service margin 18% and fourth straight double-digit quarter.
- Cash flow: first positive Q1 OCF ever, with customer capacity-reservation prepayments appearing as a new working-capital source.
- Installation: skid-mount/order-of-magnitude field-time reduction disclosed for the first time.
- GM trajectory: Q1 31.5% vs Q4 31.9% — roughly flat sequentially, but the full-year guide implies meaningful H2 margin ramp to ~34%.
- Not addressed vs prior watchlist: AEP Q2 offtake finalization, Oracle warrant execution, inventory drawdown, Brookfield, tariffs, backlog dollar/MW update.
Bull case- Oracle Jupiter is a category-defining proof point: a Tier-1 hyperscaler publicly replacing turbines and diesel with a 100%-Bloom islanded microgrid at up to 2.45 GW — and management says >half of data-center backlog sits with other hyperscalers/neoclouds/colos.
- Guidance raise one quarter in (midpoint growth 60% → 80%, GM +2 pts, op income ~3x 2025 at midpoint) with Q2 ≥ Q1 implies visible backlog conversion, not hope.
- Operating leverage finally printed: op margin +1,300 bps y/y, EBITDA margin ~19%, EPS $0.44 vs $0.03 — the flat-profitability critique from Q4 was answered.
- Service annuity strengthening: 100% attach, 10–15 year data-center durations, 18% margins, nine straight profitable quarters — compounding underneath product growth.
- First positive Q1 OCF with customer prepayments funding capacity — working-capital mechanics improving even while scaling.
- Capacity no longer the constraint: 5 GW footprint, continuous additions, order-of-magnitude faster field installation — the growth limiter shifts to customers' construction speed.
- Inference TAM framed as larger than training and structurally suited to Bloom's clean/permitting attributes; utility channel opening via rate-basing.
Bear case- "Up to 2.45 GW" is not a firm order — committed MW, pricing, revenue timing, and cancellation terms for Jupiter are all undisclosed; the unexecuted Oracle warrant agreement was not mentioned.
- The raised guide requires sharp H2 acceleration (H1 ~$1.5B vs FY $3.4–3.8B) and a GM ramp from 31.5% to ~34% — execution risk is back-loaded.
- Revenue growth figure internally inconsistent in the transcript (13.4% vs >100%) — must be verified before modeling.
- Backlog quality still opaque: no updated dollar figure, no MW, no duration, no concentration detail behind the ">half from others" claim.
- Brookfield/related-party revenue unaddressed for a third straight quarter; the promised European inference project remains missing.
- Inventory, capex, and tariffs all undisclosed on the call — the working-capital question from prior quarters is unanswered despite the positive OCF print (which was aided by customer prepayments, a potentially non-recurring lever).
- Pricing discipline is asserted, not evidenced: refusal to discuss price vs CCGT (+10–20% YTD) leaves the margin-ramp story dependent on cost-downs alone.
- Service duration claims (10–15 yrs) conflict with prior disclosures — if mix-driven, fine; if loose language, the $14B service backlog quality is less certain.
Next-quarter watchlist- Verify against the 8-K/press release: Q1 revenue growth rate (13.4% vs >100% transcript conflict) and FY2025 GM reference (30% vs 30.3%).
- Oracle Jupiter: firm vs "up to" MW, revenue recognition timing, delivery schedule, and whether the warrant/strategic partnership agreement is now executed; any Oracle appearance in related-party or backlog disclosures.
- AEP: the Q2 2026 offtake finalization deadline is this coming quarter — met, slipped, or silently dropped.
- H2 ramp credibility: Q2 revenue ≥ $751M as guided; GM trajectory toward the ~34% full-year target; product margin holding ~35%+.
- 10-Q disclosures: inventory level (did the $643M balance finally draw down?), capex vs the $150–200M guide, customer prepayment balances, Brookfield/related-party revenue, tariff exposure.
- Backlog refresh: any update to the ~$6B product / ~$14B service figures, MW disclosure, or identification of the non-Oracle hyperscaler/neocloud customers.
- Capacity: evidence of the "hundreds of MW per quarter" additions; any announcement of factory #2 beyond the 5 GW footprint and its capital cost.
- Service: whether 10–15 year durations show up in backlog disclosures; margin holding double-digit for a fifth quarter.
- 800V DC: first named DC-native deployment or revenue; absorption chilling — first commercial order or another silent quarter.
- Edwards' first full quarter: whether the new CFO changes disclosure quality (backlog detail, segment economics, long-term framework) or continues the K.R.-centric script.
|
| Feb 5, 2026 | +4.71% | Q4 FY2025 | Read transcript briefingQuarter in one view- Fifth consecutive record-revenue quarter: Q4 revenue $777.7M (+35.9% y/y); FY2025 revenue a record ~$2.0B (+37.3%), landing well above the prior $1.65–1.75B guidance range — the unquantified "better than stated" language from Q3 is now confirmed in the print.
- First-ever backlog disclosure: product backlog ~$6B, +140% y/y, with half a dozen hyperscale/neocloud end customers (vs one a year ago); ~$14B service backlog with 100% service attach on new product orders. This answers multiple quarters of deferred backlog questions.
- Q4 non-GAAP GM 31.9% — down sharply from 39.3% in Q4 2024 (mix-driven per management); FY GM 30.3% vs 28.7%. Product margin 37%; service margin ~20% in Q4, eighth consecutive profitable service quarter.
- Q4 non-GAAP operating income $133M (vs $133.4M), adjusted EBITDA $146.1M (vs $147.3M), EPS $0.45 (vs $0.43) — profit roughly flat y/y despite +36% revenue; FY adjusted EBITDA $271.6M, FY operating profit $221M (+$113.4M on +$550.1M revenue, ~20.6% drop-through).
- Cash $2.5B (boosted by convertible bonds); Q4 OCF +$113.9M, capex $57M; FCF positive for second consecutive year. Inventory ended at $643M, "slightly higher than expected" — the promised Q4 drawdown did not happen.
- First quantified 2026 guidance: revenue $3.1–3.3B (+55–65%), GM ~32%, non-GAAP operating income "$125M to $475M" (as stated on the call — an implausibly wide range, likely a transcript artifact; verify against the press release), capex $150–200M, OCF ~$200M.
- AEP exercised its option under the 1 GW agreement — sale is "unconditional," offtake finalization expected Q2 2026. Oracle warrant transaction disclosed but not yet executed.
What management is focused on- "Becoming the standard for on-site power" remains the master narrative, now backed by backlog numbers rather than adjectives. K.R. explicitly predicted the analyst question would shift from "why expand capacity" to "when will you expand more."
- Backlog diversification as the headline proof: six hyperscale/neocloud end customers vs one a year ago; C&I backlog +135% y/y across telecom, manufacturing, logistics, retail, healthcare, education; no oversized single-customer concentration claimed.
- Geographic mix shift: >80% of US backlog now outside California/Northeast (was >80% inside two years ago) — framed as evidence Bloom is cost-competitive even in low-power-cost, gas-rich states, not just a high-price-market niche.
- 800V DC elevated from Q&A topic to prepared-remarks pillar: every server shipped from now is 800V DC-ready with a removable AC adapter; retrofits offered for the installed base. "Bloom, and only Bloom, natively produces 800V DC today."
- New "apps on a platform" framing: 800V DC, battery-free AI load-following, and absorption chilling (CHP, revived after a silent Q3) presented as smartphone-style apps. Absorption chilling claimed to cut data-center electricity use "at least 20%," with a live demo cooling Bloom's own factory; no HFCs as an added selling point.
- Capacity philosophy restated: expansion is "everyday business," capital-light, ROIC in "a few months," standing orders with suppliers; the 1→2 GW signal was a one-time confidence demonstration, and future expansions won't necessarily be pre-announced.
- AI capex supercycle as demand backdrop: Amazon (~$200B 2026 capex) and Google ($175–185B) cited from the prior two days; K.R. argued visibility beyond ~six months doesn't exist and refused 2030 framing.
Key numbers and quarter mechanics- Q4: revenue $777.7M (+35.9%); GM 31.9% (vs 39.3% Q4 2024); product margin 37%; service margin ~20%; operating income $133M; adjusted EBITDA $146.1M; EPS $0.45; OCF +$113.9M; capex $57M.
- FY2025: revenue ~$2.0B (+37.3%); GM 30.3% (vs 28.7%); operating profit $221M; adjusted EBITDA $271.6M; service gross profit $29.7M; FCF positive second straight year.
- Backlog: product ~$6B (+140% y/y); service ~$14B; 100% service attach on new orders. No MW, duration, or cancellation-terms detail given.
- Cash $2.5B, explicitly attributed to convertible bond issuance — the balance-sheet jump is financing, not operations.
- Inventory $643M, "slightly higher than what we expected," justified as preparation for a strong 2026 — the Q3-promised Q4 drawdown was quietly abandoned without acknowledgment.
- 2026 guidance: revenue $3.1–3.3B; GM ~32%; operating income "$125M to $475M" (as transcribed — flag for verification); capex $150–200M; OCF ~$200M. OCF guide of ~$200M on $3.1B+ revenue implies continued heavy working-capital absorption.
- Book-and-ship: "significant double-digit percentage" of 2025 revenue; expected to remain double-digit in 2026 — K.R. framed it as a competitive weapon (customers arriving after other vendors miss commitments).
- Q4 GM decline of ~740 bps y/y attributed to project mix; management reiterated cost-downs and OpEx efficiency as the offset, pointing to EBITDA rather than margin.
- Not disclosed: related-party/Brookfield revenue share (the $288M Q3 question went entirely unaddressed), tariffs (third quarter of silence, though K.R. referenced "a tariff regime" generically in the cost answer), MW figures, CFO search.
Product and launch scorecard- 800V DC: now a shipping configuration, not a roadmap item — all new units DC-ready with AC adapter; backward-compatible retrofits offered. Still zero disclosed revenue or customer commitments; timing of data-center DC adoption deflected to customers ("ask them"). Gupta's research note was publicly praised — management is actively cultivating this narrative with the sell side.
- AEP: the quarter's biggest concrete commercial event — option exercised under the 1 GW agreement, sale unconditional, delivery will be taken regardless of offtake timing (offtake expected Q2 2026). K.R. added that Bloom and AEP are working on "several projects together." This partially answers three quarters of silence on the 900 MW remainder, though total MW conversion under the agreement remains undisclosed.
- Oracle: warrant transaction revealed (market-priced, "not penny warrants," framed as enterprise-value alignment) but the strategic partnership agreement is not yet executed — details withheld. Repeat-order traction claimed ("working with them on many projects") but still no sizing.
- Absorption chilling (CHP): revived with specifics — ≥20% data-center electricity reduction claim, ~5°C chilled water, operating demo at Bloom's factory, "customers super interested." Interest stage only; no orders or revenue.
- Battery-free load following: positioned as a key differentiator vs turbines/engines and as eliminating the battery supply-chain constraint; lab demonstrations to prospective customers cited, no deployments disclosed.
- Brookfield: not mentioned once in the entire call — remarkable given it was Q3's headline and drove $288M of related-party revenue. The year-end European AI inference data center promised last quarter was not addressed.
- Gas-provider/hyperscaler deal, CoreWeave, Equinix: none mentioned by name this quarter.
- International: explicitly deprioritized — LNG infrastructure constraints mean Europe/Asia lag "a few more years"; US is the focus "for the foreseeable future." This walks back the Q3 international-expansion framing.
- Silent items: hydrogen, SK ecoplant, Korea, SVP, repowerings, tariffs, Brookfield, CFO search.
Sell-side read-through- Arcaro (Morgan Stanley): follow-on demand — got the "two-thirds of C&I business is repeat customers" stat and Oracle repeat-project color. Capacity triggers: answer was "continuous decision, no milestones" — the go/no-go framework question from prior quarters is now permanently deflected.
- Dendrinos (RBC): 800V DC deployment timing — deflected to customers; R&D roadmap — produced the battery-free load-following reveal.
- Gupta (UBS): absorption chillers vs vapor compression — the most substantive technical answer of the call (≥20% electricity reduction, HFC-free, supply-chain diversification). K.R. opened by praising Gupta's 800V DC note — the mutual-credibility dynamic continues.
- Sandler (Baird): wins vs CCGT — K.R. refused win/loss specifics ("ask the end user") but cited "hundreds of megawatts going into gigawatts" single-site stamp sizes as implicit evidence. Stack life/service risk — answered with eight profitable quarters, 20% Q4 service margin, trillions of cell-hours, 6B+ daily data points, digital twins; ended with a challenge: "are you placing enough enterprise value on this service business?"
- Blum (Wells Fargo): MTAR's 30%-CAGR-to-2030 comment — firmly disowned ("we have not provided any long-term guidance... ask them"), replaced with the six-month-visibility doctrine. Backlog US/international mix — declined; international qualitatively pushed out years.
- Rusch (Oppenheimer): pricing leverage — "we don't have to choose between growth and profitability," value-based pricing, no specifics. M&A — selective-only, "lighting up the planet is a good day job."
- Strouse (JPMorgan): non-AI geographic mix — flat refusal ("we just don't do that... give me another question"). Book-and-ship — significant double-digit share of 2025, expected to persist.
- Elmaghrabi (BTIG): the two best housekeeping answers — AEP option exercised with unconditional sale (offtake Q2 2026), and the Oracle warrant structure (market-priced, unexecuted, case-by-case for other hyperscalers).
- Parks (Tuohy): input-cost visibility/long-term contracting — generic cost-down answer; follow-up cut off for time.
- Notable: no analyst asked about Brookfield, related-party revenue, the missed inventory drawdown, tariffs, or the CFO search — the Q3 accountability threads were dropped by the sell side, not just management.
Management credibility- Delivered on the big one: FY2025 revenue ~$2.0B vs the $1.65–1.75B range — the vague Q3 "better than stated" proved conservative, and first-time quantified 2026 guidance plus first-time backlog disclosure partially repair the information-deficit criticism.
- Failed on inventory: the drawdown promised for Q4 (itself slipped from "H2") did not occur — inventory rose to $643M, "slightly higher than expected," with no acknowledgment of the prior commitment. Second consecutive quiet slip on this line item.
- Service margin language inconsistency: Kurzymski called Q4 "the first straight quarter of double-digit margins in the service business" — Q3's summary shows service was already at 14.4% (second straight double-digit quarter then). Either a misstatement or a restatement; verify against the press release.
- The 2026 operating income range as transcribed ("$125M to $475M") is not credible as stated — the low end is below 2025's $221M while revenue grows ~60%. Must be checked against the 8-K before use.
- Brookfield silence is a credibility negative: after $288M (~55%) of Q3 revenue was related-party and a year-end European project was promised, zero update was volunteered and none was demanded. The disclosure-avoidance pattern from Q3 continues.
- MTAR disavowal was handled cleanly and quickly — good discipline on not letting a supplier set guidance.
- AEP answer was specific and checkable (unconditional sale, Q2 offtake) — a credibility deposit after three quarters of silence on the 900 MW.
- Oracle warrants: disclosed only under direct questioning, agreement unexecuted — the pattern of material structure details surfacing via analysts rather than prepared remarks persists.
- CFO: acting-PFO arrangement continues with zero update; now roughly nine months without a permanent CFO and the topic has vanished from the script.
- K.R.'s demand calls continue to age well (Amazon/Google capex prints landed two days before the call), and the 55-day Oracle delivery was referenced again as the execution proof point.
What changed versus the prior quarter- Guidance: from unquantified "better than stated" to full 2026 numbers ($3.1–3.3B revenue, ~32% GM, capex $150–200M, OCF ~$200M) — the quantification drought broke.
- Backlog disclosed for the first time: ~$6B product (+140%), ~$14B service, six hyperscale/neocloud end customers — a structural improvement in disclosure, though still no MW or duration.
- AEP moved from three quarters of silence to option exercised with an unconditional sale; offtake timing (Q2 2026) is now a dated milestone.
- Oracle relationship deepened structurally: warrant transaction (unexecuted) revealed — a new form of customer alignment not previously disclosed.
- 800V DC graduated from Q&A narrative to shipping product configuration with retrofit offer.
- CHP/absorption chilling returned after a silent quarter with a quantified claim (≥20% electricity reduction) and a live factory demo.
- Brookfield went from headline to zero mentions; the promised year-end European inference project is unaccounted for.
- International posture reversed: from "European project by year-end" and LNG-unlock optimism to "international lags a few more years; US is the focus."
- Inventory story broke: drawdown abandoned, balance rose to $643M.
- Cash jumped from $627M to $2.5B via converts — leverage/financing now part of the story.
- Q4 GM declined y/y (31.9% vs 39.3%) — the first y/y GM decline in the record-revenue streak, attributed to mix.
- CFO search, tariffs, and related-party revenue all went unaddressed — the silent-items list now includes the prior quarter's biggest governance and revenue-quality questions.
Bull case- Demand is now quantified: $6B product backlog (+140%), $14B service backlog, six hyperscale/neocloud customers vs one, C&I backlog +135% — the "lighthouse replication" thesis now has numbers attached, and 2026 guidance of $3.1–3.3B implies ~55–65% growth with backlog coverage.
- AEP conversion de-risks the largest disclosed pipeline item: unconditional sale means revenue regardless of offtake timing, and "several projects together" suggests expansion beyond the original 1 GW.
- Service is becoming a second business: eight straight profitable quarters, 20% Q4 margin, $14B backlog with 100% attach — a growing annuity-like profit stream with AI-driven fleet economics (trillions of cell-hours, digital twins).
- 800V DC is now a shipping differentiator ahead of the 2027 NVIDIA transition, with backward compatibility removing the adoption-timing risk for customers; absorption chilling adds a ≥20% electricity-savings claim that compounds the on-site value proposition.
- Geographic diversification (>80% of backlog outside CA/Northeast) demonstrates competitiveness in cheap-power states — undercuts the "only works where power is expensive" bear argument.
- Balance sheet: $2.5B cash, two straight FCF-positive years, capacity expansion self-funded at claimed months-long ROIC.
- Hyperscaler capex acceleration (Amazon ~$200B, Google $175–185B) validates the demand backdrop days before the print.
Bear case- Q4 GM fell ~740 bps y/y (31.9% vs 39.3%) and Q4 operating income/EBITDA were flat y/y despite +36% revenue — operating leverage did not show up in the quarter; the 2026 ~32% GM guide implies only modest recovery.
- The inventory drawdown promise failed silently; inventory rose to $643M and 2026 OCF guidance of ~$200M on $3.1B+ revenue implies working capital keeps consuming cash as growth accelerates.
- Brookfield/related-party revenue went completely unaddressed — the ~55%-of-Q3-revenue question is now two quarters old with no update, and the promised year-end European project appears to have missed or been dropped without comment.
- Oracle warrants are unexecuted and were disclosed only under questioning; Oracle sizing remains unknown; the gas-provider hyperscaler deal from Q3 was never announced or mentioned again.
- International is now explicitly years away — the TAM narrative narrowed to the US, and the European inference project is unaccounted for.
- The 2026 operating income range as stated is incoherent ($125M low end below 2025 actuals) — either a transcript error or genuinely sloppy guidance communication; the service-margin "first double-digit quarter" claim contradicts the prior two quarters.
- No permanent CFO ~nine months in, with the topic now entirely absent from calls.
- Backlog quality unknown: no MW, duration, cancellation terms, or customer-concentration detail behind the "no oversized concentration" assertion.
- Competitive win/loss evidence vs CCGT still refused; stamp-size claims ("hundreds of MW going into gigawatts") are unverified assertions.
Next-quarter watchlist- Verify against the 8-K/press release: the 2026 operating income range ("$125M to $475M" as transcribed) and the service-margin "first double-digit quarter" claim — both look wrong as stated.
- 10-K related-party disclosure: Brookfield revenue share for Q4/FY2025, JV equity balances, arm's-length pricing discussion — and any explanation for the missing year-end European inference project.
- AEP: offtake finalization in Q2 2026 as stated; MW delivered and revenue recognized under the exercised option; evidence of the "several projects together."
- Oracle: execution of the warrant/strategic partnership agreement, any sizing, and whether Oracle appears in backlog or related-party disclosures.
- Q1 2026 execution vs the $3.1–3.3B guide: implied Q1 revenue run-rate, GM trajectory toward ~32%, and whether product margins hold ~37%.
- Inventory and OCF: does inventory finally fall in 2026, and is the ~$200M OCF guide tracking — or does growth keep absorbing cash?
- Backlog quality: any disclosure of MW, duration, conversion timing, or the identity of the six hyperscale/neocloud end customers; book-and-ship share staying double-digit.
- 800V DC: first customer commitment or deployment of the DC-native configuration; retrofit uptake.
- Absorption chilling: first commercial order vs continued "customer interest."
- Capacity: whether the next expansion beyond 2 GW is announced or simply executed (per the new "continuous decision" doctrine), and capex tracking vs the $150–200M guide.
- CFO search: permanent hire or a full year with an acting PFO.
- Tariffs: whether the 10-K quantifies exposure after three quarters of call silence.
|
| Oct 28, 2025 | +18.03% | Q3 FY2025 | Read transcript briefingQuarter in one view- Fourth consecutive record-revenue quarter: Q3 revenue $519M (+57% y/y), non-GAAP GM 30.4% (vs 25.2% Q3 2024, +510 bps), non-GAAP operating income $46.2M (vs $8.1M), adjusted EBITDA $59M (vs $21M), non-GAAP EPS +$0.15 (vs -$0.01). Product margin 35.9%; service margin 14.4% — second straight double-digit service quarter, seventh consecutive profitable service quarter.
- OCF positive at +$20M ("primarily working capital improvements"); total cash $627M. Inventory drawdown deferred again — now expected in Q4 as shipments accelerate (previously "H2").
- Guidance language upgraded but not quantified: "we expect fiscal 2025 to be better than our previously stated annual guidance on our financial metrics" — no revised numbers given, and Mandloi's question on why no Q4 guidance was answered with the customer-readiness/timing framework, not numbers.
- Oracle milestone delivered: 90-day power commitment fulfilled in 55 days — the prior quarter's binary credibility test passed ahead of schedule.
- New headline: Brookfield partnership — $5B "inaugural" investment, Bloom named preferred on-site provider across Brookfield's ~$1T infrastructure portfolio (140 data centers, ~1 GW critical load), Brookfield will finance Bloom-sourced PPA deals, and a Bloom-powered European AI inference data center to be announced by year-end. $288M of Q3 revenue was related-party, driven by Brookfield JV equity investments.
- New channel disclosures: first deal with a major gas provider selling power to a third unnamed hyperscaler (announcement deferred to the hyperscaler); CoreWeave named as a neocloud customer (Illinois site); Equinix cited at 100+ MW deployed.
What management is focused on- "Becoming the global standard for on-site power" is the new framing — a trillion-dollar market claim, with the telecom playbook (AT&T lighthouse 2011 → Verizon/T-Mobile → 100+ MW) as the explicit template for AI.
- Seven AI ecosystem channels, each with a lighthouse customer: hyperscalers (Oracle), electricity providers (AEP/AWS), gas providers (unnamed deal), co-location (Equinix), neoclouds (CoreWeave), data center developers (permits being filed publicly), infrastructure owners (Brookfield).
- 800V DC architecture as the next differentiation wedge: K.R. gave an extended technical answer (prompted by Gupta) arguing NVIDIA's move to 800V DC by 2027 is a "must-have" and that Bloom has shipped DC-native units for 15 years, currently converting DC→AC — positioning Bloom as already built for the Rubin-era rack densities. This was volunteered as a script "miss" and is clearly a new narrative pillar.
- FERC/DOE 60-day interconnection proposal spun as a tailwind: even with faster interconnection, "BYOP" curtailment language means on-site power remains necessary; Bloom pitched to utilities for ancillary/reactive-power services (wide power-factor range vs narrow for turbines/engines, hot standby vs 5–6 minute turbine starts).
- Competitive framing sharpened vs turbines/engines: no air pollution, no batteries needed for load-following (solid-state), faster capacity expansion, more power per unit of gas ("watts to tokens"), future-proofing for DC/carbon capture/green molecules.
- Capacity: 2 GW by December 2026 "all systems go," now explicitly tied to "~4x our 2025 revenue"; OpEx being spent on talent/capability to plan beyond 2 GW. "We will not be the bottleneck" repeated twice under questioning.
Key numbers and quarter mechanics- Q3: revenue $519M (+57% y/y); GM 30.4%; product margin 35.9%; service margin 14.4%; operating income $46.2M (~8.9% operating margin); adjusted EBITDA $59M; EPS $0.15; OCF +$20M; cash $627M.
- 9M revenue = $1,246M ($326M + $401M + $519M) = 75.5% of the $1.65B low end and 71.2% of the $1.75B midpoint — Q4 needs ~$404M at the low end, which Q3 alone already exceeds; the "better than guidance" language is arithmetically conservative.
- $288M of related-party revenue in Q3 (per 10-Q, surfaced by Thakkar) — attributable to Brookfield JV structures where Bloom made "fairly small" equity investments with a cap. That is ~55% of Q3 revenue from a related party — a material mix/disclosure development not addressed in prepared remarks.
- Mix shift disclosure: majority of transactions are now PPA structures; some CapEx deals each quarter; managed services (sale-leaseback) discontinued and not expected to return.
- Guidance: "better than previously stated" across financial metrics — no revised range, no Q4 number. K.R. explicitly defended the vagueness: projects can fall on either side of December 31, "65 days left," customer readiness (not supply) is the swing factor.
- Inventory: drawdown now a Q4 event (was "H2" last quarter) — the level-load-to-shipment conversion story slipped another quarter within the year.
- No backlog, MW bookings, tariff, or capex figures disclosed. Tariffs — last quarter's unreconciled "4% on materials" — went entirely unmentioned this quarter.
- Brookfield capital commitments: none beyond the small capped JV equity investments (Elmaghrabi's question answered directly by Kurzymski).
Product and launch scorecard- Oracle: delivered in 55 days vs the 90-day promise — the key execution proof point landed. Still no MW, revenue, or margin sizing; Blum's sizing question was deflected ("ask them"), with only Oracle's "first of many" press-release language cited. Expansion evidence remains Oracle's words, not Bloom's disclosure.
- Brookfield: the quarter's biggest announcement — $5B initial investment, preferred-provider status across a ~$1T portfolio, PPA financing backstop, European AI inference project promised by year-end (a new dated, checkable milestone). But $288M related-party revenue means a large share of current revenue is now tied to a partner in which Bloom holds equity — revenue quality and arm's-length pricing are now legitimate questions.
- Gas provider channel: first deal signed, converting gas to power for a third hyperscaler; details withheld pending the hyperscaler's announcement. Unverifiable this quarter.
- CoreWeave: named as operating at an Illinois high-performance data center — new named neocloud proof point, no size given.
- Equinix: 100+ MW across multiple states — cumulative, not new volume.
- AEP: mentioned only in passing (AWS context); no update on the remaining 900 MW of the 1 GW agreement — now a third quarter without conversion news.
- 800V DC: positioned as a built-in 15-year-old capability awaiting the 2027 NVIDIA transition — zero revenue today, but a credible-sounding moat claim; watch for customer validation.
- International: European inference project (Brookfield) is the first concrete international AI data point; Tokyo policy sentiment on LNG cited; still no new named international orders. CCS interest in Europe mentioned ("tremendous interest") — first CCS mention after two silent quarters, but interest only, no deals.
- Not mentioned: hydrogen, SK ecoplant (Kallo referenced the 80 MW SK project; K.R. did not engage with it specifically), Korea, SVP, repowerings, CHP (silent for the first time in several quarters).
Sell-side read-through- Arcaro (Morgan Stanley): pace of commercial activity — K.R. escalated from "robust" (three prior calls) to "clearly accelerating and palpable," including non-AI C&I. Competitive question drew the turbines/engines "band-aided mechanical-age" framing.
- Dendrinos (RBC): Brookfield structure/timeline/financial benefit — got the $50B invested/tripling, 140 data centers/1 GW, PPA financing role, and the year-end European announcement. International: power shortages in Frankfurt, Munich, Dublin, Taipei, Delhi, Mumbai; LNG policy unlock as the catalyst.
- Gupta (UBS): FERC 60-day proposal (BYOP/curtailment spin) and the 800V DC question that produced the call's most substantive technical answer. Gupta again opened with praise for the 55-day delivery — management's credibility bank with him is growing.
- Amicucci (Evercore): utilization of 2 GW into 2027 and the 4x revenue implication — K.R. declined utilization specifics, deferred to "backlog and other things next year," emphasized ROIC discipline. Inference vs training: same architecture, fewer LEGO blocks; inference siting near populations favors Bloom's no-pollution/no-noise profile.
- Kallo (Baird): project size scaling (SK 80 MW reference; permits with "huge numbers") — "nothing in our scaling has scaling risk," larger systems more reliable via hot-swappable modularity; "much larger sizes" in negotiation. Go/no-go beyond 2 GW: refused criteria — "we will not be the bottleneck."
- Strouse (JPMorgan): long-term margin targets — explicitly deferred to annual guidance in ~90 days; the only framework offered: double-digit annual cost-downs, electricity pricing pressure set by shortage macros, operating discipline, non-capital-intensive factories.
- Blum (Wells Fargo): Oracle opportunity sizing — deflected to Oracle; only "first of many" cited. A soft evasion.
- Thakkar (BMO): the only analyst to surface the $288M related-party revenue — Kurzymski confirmed Brookfield JV equity investments triggered the disclosure, "fairly small" and capped. The most important housekeeping answer of the call.
- Rusch (Oppenheimer): mix shift to financed structures (majority PPA; managed services dead) and CFO search — "sense of urgency, but no sense of rush," no timeline. Now ~six months without a permanent CFO.
- Mandloi (Mizuho): why no Q4 guidance — answered with the timing/customer-readiness framework; K.R. effectively conceded the raised-but-unquantified guidance is a 65-day visibility artifact.
- Elmaghrabi (BTIG): Brookfield capital commitments — none beyond small capped equity investments. Clean answer.
- Notable absences: no analyst asked about tariffs (the Q2 4% figure remains unreconciled and now unmentioned), the AEP 900 MW, or backlog.
Management credibility- The Oracle 55-day delivery is a major credibility deposit: a dated public commitment beaten by 35 days. K.R. earned the right to the "we ship ahead of time" claim this quarter.
- Guidance handling is a mixed signal: "better than previously stated" without numbers is directionally positive but preserves maximum optionality; Mandloi's question exposed that the raise is not being formalized because of December 31 timing risk — prudent, but it also means the market is underwriting Q4 on trust again.
- The $288M related-party revenue is the quarter's credibility test in the other direction: more than half of Q3 revenue came through Brookfield JV structures with Bloom equity in them, and this surfaced only via an analyst reading the 10-Q, not in prepared remarks. The amounts are said to be small/capped on the equity side, but the revenue concentration was not proactively disclosed on the call.
- Tariff silence: after Q1's "100 bps GM" and Q2's "4% on materials," Q3 offered nothing — with GM at 30.4%, the offset story is empirically holding, but the original reconciliation question is now two quarters stale and management appears to have simply dropped the topic.
- Capacity claims remain assertion-based: "all systems go," "we will not be the bottleneck," no timeline/cost detail beyond the prior ~$100M, no utilization framework, backlog deferred to "next year."
- CFO search: "urgency but no rush" after ~six months is a thin answer; Kurzymski is now running calls comfortably but the open seat is becoming a governance data point.
- Consistency: level-loading/inventory-drawdown language repeated but the drawdown window narrowed from "H2" to "Q4" without acknowledgment — a quiet slip within the script.
- K.R.'s demand calls continue to age well (DeepSeek vindication last quarter, Oracle delivery this quarter); the pattern of declining quantification (no guidance numbers, no backlog, no deal sizing) continues in parallel.
What changed versus the prior quarter- Oracle moved from "PO with 90-day promise" to delivered in 55 days — execution risk on the flagship deal substantially retired; sizing still undisclosed.
- Brookfield is an entirely new pillar: $5B investment, preferred-provider status, PPA financing role, year-end European inference project — and with it, $288M (~55%) of Q3 revenue became related-party, a structural change in revenue composition.
- Guidance stance shifted from "reiterated range" to "better than stated" — directionally raised, numerically withheld.
- Revenue model disclosure: majority PPA now stated explicitly; managed services formally retired.
- New named customers/channels: CoreWeave (neocloud), unnamed gas-provider-to-hyperscaler deal; Equinix quantified at 100+ MW cumulative.
- 800V DC emerged as a new core narrative — absent from all prior calls.
- FERC 60-day interconnection proposal addressed for the first time and spun as net-positive (BYOP/curtailment).
- Inventory drawdown timing slipped from "H2" to "Q4."
- Tariffs disappeared from the script entirely after two quarters of escalating figures.
- AEP 900 MW, CHP, hydrogen, SK ecoplant, Korea all went unmentioned — the silent-items list grew.
- Commercial-momentum language escalated from "robust" to "accelerating and palpable," extended beyond AI to traditional C&I.
Bull case- Execution proof is compounding: Oracle delivered in 55 days, four straight record-revenue quarters, GM at 30.4% with product margins at 35.9%, service margins double-digit for two consecutive quarters — the margin-expansion story is now demonstrated, not projected.
- Guidance is de facto raised: 9M revenue of $1,246M plus "better than stated" language implies the year lands above $1.65B with room; Q4 needs only ~$404M for the low end.
- Brookfield transforms the demand structure: a $5B committed partner with a ~$1T portfolio, a financing backstop for PPAs, and a dated European project — this is a demand and funding channel simultaneously, and the "inaugural" framing implies more capital available.
- Channel breadth is now concrete: seven channels each with a named or described lighthouse (Oracle, AEP/AWS, gas provider, Equinix, CoreWeave, developers filing permits, Brookfield) — the lighthouse playbook claim is supported by visible replication.
- 800V DC positions Bloom ahead of a forced industry architecture shift (NVIDIA 2027); if validated by customers, this is a multi-year moat that turbines/engines structurally cannot match.
- Policy tailwinds stacking: ITC secured through 2032, FERC interconnection reform framed as BYOP-accelerant, LNG diplomacy unlocking Europe/Asia gas acceptance.
- Balance sheet: $627M cash, positive OCF, capacity expansion self-funded, no Brookfield capital obligations beyond small capped equity.
Bear case- ~55% of Q3 revenue ($288M) is related-party via Brookfield JVs in which Bloom holds equity — revenue quality, pricing arm's-length, and concentration questions are now material, and management did not volunteer this on the call.
- The guidance "raise" is unquantified; the company is asking investors to underwrite Q4 and 2026 without numbers, backlog, or MW disclosure — the information deficit is widening as the story gets bigger.
- Oracle remains unsized and its expansion rests on Oracle's "first of many" language; the gas-provider hyperscaler deal is unannounced and unverifiable; CoreWeave is unsized. Much of the channel narrative is still name-checks without economics.
- AEP's remaining 900 MW has now gone three quarters without conversion — the largest single disclosed pipeline item is static.
- Inventory drawdown slipped from H2 to Q4; if Q4 shipments disappoint, the level-loading defense of H1 cash consumption weakens and OCF quality becomes a question.
- PPA-majority mix means more revenue depends on financing structures and partner balance sheets (Brookfield) — different risk profile than direct CapEx sales, with margin and cash-timing implications not yet quantified.
- Tariff exposure unresolved and now undiscussed; the 4%-materials figure was never reconciled and the topic has been dropped rather than answered.
- No permanent CFO ~six months in; "urgency but no rush" is not a plan.
- Competitive dismissal ("band-aided mechanical-age") is marketing, not evidence — no win/loss data vs turbines/engines was offered despite direct questioning.
Next-quarter watchlist- Q4 print vs "better than stated": does FY2025 land above $1.75B midpoint, and does management finally give 2026 guidance with numbers (Strouse was told to wait ~90 days)? Watch GM trajectory vs the ~29% guide and whether product margins hold ~36%.
- Related-party revenue: Q4 10-K/10-Q disclosure on Brookfield JV revenue share, equity investment balances vs the cap, and any arm's-length pricing discussion. Does the $288M/Q3 level persist, grow, or diversify?
- Brookfield European AI inference data center: announced by year-end as promised? Size, structure, and whether Bloom revenue is recognized through the JV.
- Oracle: any sizing, second project, or confirmation of the "many" projects; whether Oracle revenue is identifiable in segment/related-party disclosures.
- Gas-provider/hyperscaler deal: the hyperscaler's own announcement — identity, MW, timeline.
- Inventory and OCF: promised Q4 drawdown and shipment acceleration — check inventory balance and full-year OCF vs the "~FY2024 level" guide.
- AEP 900 MW: any conversion, or a fourth quarter of silence.
- 2 GW expansion: cost/timeline detail, whether the ~$100M sits inside capex guidance, and any utilization or backlog framework for 2027 (Amicucci's deferred question).
- 800V DC: any customer or NVIDIA-ecosystem validation of Bloom's DC-native claim.
- CFO search: permanent hire or further delay; any 10-K disclosure on the transition.
- Tariffs: whether the topic resurfaces in the 10-K or 2026 guidance after a quarter of silence.
- Silent items: CHP (newly silent), hydrogen, SK ecoplant, Korea, SVP, repowerings — and whether international produces a second concrete project beyond the Brookfield European announcement.
|
| Jul 31, 2025 | -1.79% | Q2 FY2025 | Read transcript briefingQuarter in one view- Record Q2: revenue $401M (+19.5% y/y), non-GAAP GM 28.2% (vs 21.8% Q2 2024, +650 bps, again attributed to mix and level-loaded manufacturing), non-GAAP operating income $28.6M (vs -$3.2M), adjusted EBITDA $41.2M (vs $10.2M), non-GAAP EPS +$0.10 (vs -$0.06). Third consecutive quarter of record profits/operating margin; sixth consecutive quarter of service profitability, with service margins hitting double digits for the first time.
- Full-year guidance reiterated: revenue $1.65–1.85B, non-GAAP GM ~29%, non-GAAP operating income now given as a range of $135–165M (prior quarters cited ~$150M — consistent midpoint), OCF and capex "around the same level" as FY2024, 40/60 H1/H2 split maintained.
- Two headline commercial proof points: Oracle partnership (direct hyperscaler purchase order, islanded load-following, power committed in 90 days) and AEP's announcement that AWS and Coralogix are deploying Bloom systems in Ohio — confirmed as part of the existing 100 MW PO, with the remaining 900 MW still pipeline.
- Capacity doubling announced: 1 GW/yr to 2 GW/yr by end of 2026, ~$100M cost spread over quarters, "well funded."
- Balance-sheet actions: $113M of August 2025 convertible notes refinanced into existing 2029 converts; inventory deliberately built in H1 for level-loading, to be worked down as H2 shipments accelerate.
- Policy tailwind: BBB restored ITC for fuel cells from January 2026 (flat 30%); management says safe-harbored volume covers 2025, so customers face "zero gap" through 2032.
What management is focused on- Hyperscaler validation as the quarter's narrative core: Oracle is the first direct hyperscaler customer (previously utility-intermediated); K.R. frames it as proof Bloom can load-follow at large scale, islanded, at "AI speed" (90-day power commitment).
- Time-to-power as the wedge: AEP CEO Bill Fehrman quoted — demand growing at a pace unseen in a 45-year career, interconnection taking 5–7 years; Bloom positioned as the bypass.
- Capacity expansion justified by secular math: hyperscalers spending >$1B/day on capex, >$500B in calendar 2025, implying ~$50B of power capital equipment and "more than one sizable nuclear power plant's worth of baseload needed every month." 2 GW framed as "a start to multi-gigawatts."
- Product architecture differentiation vs turbines/engines: no AC-DC converters or harmonic suppression "band-aids"; CapEx at parity or better vs turbines at equivalent availability, 15–20 percentage points lower fuel consumption, no air-permit friction.
- Continuous product improvement replacing generational step-changes: digital twins, 4.5 trillion field data points, 1M+ stacks, 22,000+ servers, 40+ TWh generated — AI-driven cost reduction credited with holding margin guidance despite a "4% tariff hit on materials" (new, larger tariff figure — see below).
- CHP as a retrofittable "app": high pipeline interest, but customers prioritize speed-to-power first, retrofit later; framed as equivalent to eliminating ~20% of a data center's power need (cooling).
- Skid-mounted "grid to go" mobility: repositionable, fragmentable/aggregatable systems pitched to hyperscalers; K.R. dismisses temporary-power risk — "it's a bridge to a bridge, not a bridge to a solution."
Key numbers and quarter mechanics- Q2: revenue $401M (+19.5% y/y); non-GAAP GM 28.2%; non-GAAP operating income $28.6M (7.1% operating margin per Gupta); adjusted EBITDA $41.2M; non-GAAP EPS $0.10.
- H1 revenue = $727M ($326M + $401M) = 44% of the $1.65B low end and 41.6% of the $1.75B midpoint — tracking slightly ahead of the 40% H1 script; Benjamin (Mizuho) noted the company appears on track toward the top end. K.R. declined to narrow the range, attributing it entirely to customer readiness (greenfield completion, gas connection, permits) — "none of these projects are in jeopardy," but timing can shift weeks to a month.
- Guidance reiterated in full; operating income now expressed as $135–165M (midpoint $150M, consistent with prior "~$150M").
- Tariff language changed materially: K.R. referenced "a 4% tariff hit on our materials" being absorbed within the maintained margin guide — versus "up to 100 bps of gross margin" last quarter. Whether 4% is a materials-cost figure (not GM bps) or an escalation was not clarified on the call; no analyst asked. This is the quarter's most important unexamined number.
- Inventory: deliberately built in H1 via level-loading (visible in cash/inventory per K.R.); to be worked down as H2 shipments accelerate — implies H2 cash conversion depends on the shipment ramp.
- Convert: $113M of August 2025 notes exchanged into 2029 converts — removes near-term maturity, adds financing optionality.
- Capacity capex: ~$100M for the 2 GW expansion, spread over quarters, incremental to the "capex around FY2024 levels" guide? Not explicitly reconciled — worth checking whether the ~$100M is inside or outside the reiterated capex guide.
- Book-and-ship dependence reaffirmed verbatim: "a portion of our revenue for the year would come from deals that we book, build, ship and recognize revenue in the same year." Senyek's question on whether Oracle is embedded in guidance was answered only with this general framework — not a direct yes/no.
- No backlog, megawatt, cash balance, OCF, or capex figures disclosed on the call (consistent with the narrowed disclosure regime).
Product and launch scorecard- Oracle: purchase order (not LOI/MOU — K.R. explicit and slightly bristled when Gosai asked), single data center first project, islanded, Bloom carries primary and secondary load, 90-day power commitment, "many" further projects in discussion. The single most important proof point in company history if executed; the 90-day delivery is now a dated, checkable milestone (roughly end of October 2025).
- AEP/AWS + Coralogix (Ohio): confirmed within the existing 100 MW PO — i.e., not incremental volume; the remaining 900 MW is still "actively" pipeline. K.R. expressed hope to fulfill the full 1 GW and sign "future agreements like this very soon."
- Quanta Computer: islanded load-following microgrid installed in Q2 at Fremont (repeat order after last year's rapid deployment); management expects orders from other AI hardware ecosystem players "soon."
- CHP: commercially offered, high interest, but no confirmed deployment at Oracle; positioned as a post-energization retrofit. Still pre-quantified.
- Capacity: 2 GW by end-2026, ~$100M, timeline detail refused ("we can increase capacity in months"); funded from existing resources.
- International: mix disclosed for the first time — ~30% international / 70% domestic, expected to hold "at least through the next year or so." Taiwan, Germany, Italy, UK still in policy/regulator-alignment phase; Korea "continuing strength." No new international orders announced.
- ITC/safe harbor: 2025 demand protected via safe-harbored volume; 2026–2032 covered by BBB's flat 30% ITC. Notably, the new 30% flat credit is lower than the prior 40–50% (with energy-community/domestic-content adders) — K.R. conceded "subsidies go down a little bit" but argued high electricity prices and cost reduction preserve margins. This is a subtle demand-economics headwind for 2026+ that management spun as neutral.
- Not mentioned: CCS/Chart, hydrogen, SK ecoplant, Korea auction, SVP, repowerings — second consecutive quarter of silence on most of these.
Sell-side read-through- Arcaro (Morgan Stanley, new to coverage): hyperscaler usage (exclusive/primary source? — yes, islanded primary + secondary load) and whether Oracle accelerates more deals (K.R.: yes, "many" projects in discussion); capacity-doubling confidence (pipeline + secular math, no backlog cited).
- Strouse's associate (JPMorgan): capacity timeline (refused specifics) and cost/funding (~$100M, "well funded") — the only hard capacity numbers of the call.
- Gupta (UBS): operating-margin trajectory toward double digits in 2026–27 (K.R.: expect continued improvement, quarterly mix variance); AEP expansion potential (1 GW fulfillment hope, "robust" pipeline). Gupta explicitly credited K.R.'s DeepSeek-blip call — a credibility data point management will welcome.
- Dendrinos (RBC): CHP deployment status (interest high, retrofits later); turbine comparison (K.R.: CapEx parity at matched availability, 15–20 pt fuel advantage, permitting edge; islanded share of data centers now ~40% in surveys vs ~20% six months ago).
- Benjamin (Mizuho, for Mandloi): what would raise the low end (customer readiness only, no shipment constraint); ITC pushout risk into 2026 (denied — "zero gap," safe harbor covers 2025). The cleanest policy answer of the call.
- Senyek (Wolfe): is Oracle embedded in guidance? — answered with the general book-and-ship framework, not a direct confirmation. A soft evasion worth noting.
- Ailani (Jefferies): international progress (30/70 mix disclosed; four named markets still in regulator-education phase).
- Rusch (Oppenheimer): temporary-power/relocation risk (reframed as a feature — skid-mounted "grid to go"; "bridge to a bridge" dismissal of 2-year demand risk).
- Elmaghrabi (BTIG): does ITC clarity enable price increases? — K.R. disclosed the flat 30% vs prior 40–50% structure and leaned on cost reduction + high power prices; no pricing-power claim made.
- Parks (Tuohy): next-gen server economics — answer: no more generations, continuous improvement; the 4% tariff-on-materials figure surfaced here, unprompted and unexplained.
- Landon (Rothschild Redburn): the only analyst to pin down whether AWS/Coralogix is incremental — confirmed inside the 100 MW PO. Important clarification that tempers the headline.
- Gosai (BofA): Oracle deal status (PO, executing, 90 days — KPI/framework questions declined); capital needs (referred to earlier answer, mildly dismissive).
Management credibility- Delivery vs script: H1 revenue $727M is ahead of the 40% H1 pace at the midpoint; Q2 GM 28.2% and third straight record-profit quarter support the level-loading and cost-reduction claims. The operating-income guide was recast as a range ($135–165M) without comment — midpoint unchanged, but the framing shift went unacknowledged.
- The 4% tariff figure is a credibility test: last quarter's "up to 100 bps GM impact" became "4% tariff hit on our materials" with margin guidance still held. These may be reconcilable (materials share of COGS), but management offered no reconciliation and no analyst pressed — either the offset engine is far stronger than disclosed, or the metric shifted. Needs follow-up.
- Oracle specificity is high and falsifiable: PO (not MOU), 90 days, islanded, primary + secondary load. K.R. has now put a dated public stake in the ground — credibility will be binary on this by Q3.
- Capacity-doubling rationale rests on pipeline confidence, not disclosed backlog — with backlog still annual-only, investors again underwrite the ramp on trust. K.R.'s "we will never get past our headlights" framing is asserted, not evidenced.
- Acting CFO Kurzymski ran his first call cleanly but said nothing beyond the numbers; no update on the permanent CFO search — an open item now three months old.
- Consistent message discipline: book-and-ship dependence, 40/60 split, and guidance reiteration all repeated verbatim from Q1; the DeepSeek-blip vindication (Gupta) genuinely strengthens K.R.'s demand-call track record.
- Evasions: Oracle-in-guidance (Senyek), capacity timeline detail, Oracle KPIs/framework, tariff reconciliation, CFO search — mostly minor individually, but the pattern of declining quantification continues.
What changed versus the prior quarter- Hyperscaler relationship upgraded from utility-intermediated to direct: Oracle is the first direct hyperscaler PO; AEP channel produced named end-users (AWS, Coralogix) — but confirmed as within the existing 100 MW, not incremental.
- Capacity stance shifted from "level-load the existing factory" to doubling to 2 GW by end-2026 (~$100M) — the first major capacity commitment of the AI cycle.
- Tariff quantification moved from "up to 100 bps GM" to "4% on materials" — larger-sounding, unreconciled, and unchallenged.
- ITC risk resolved: BBB restored fuel-cell ITC (flat 30%, 2026–2032) and safe harbor bridges 2025 — last quarter's policy overhang is now a claimed tailwind, though the 2026 credit is structurally lower than the 40–50% safe-harbored rate.
- International mix quantified for the first time: ~30/70 international/domestic, expected to hold ~1 year — a subtle de-emphasis of the "Taiwan/Europe take-off" timeline from Q1.
- Operating-income guide recast from "~$150M" to "$135–165M" (same midpoint, new range framing).
- CFO transition executed: Kurzymski (Acting PFO) presented his first call; Berenbaum gone with still no explanation; search status undisclosed.
- Service margins crossed double digits for the first time; service profitability streak extended to six quarters.
- Islanded data-center share claim escalated: ~40% of data centers in surveys now considering islanded power vs ~20% six months ago (management-cited).
- Product roadmap language changed: discrete "generations" retired in favor of continuous improvement — Parks's next-gen question was answered by dissolving the concept.
Bull case- The demand thesis now has named, dated proof points: Oracle PO with 90-day power, AWS and Coralogix deploying through AEP in Ohio, Quanta repeat order installed in Q2 — the "inflection point" claim is no longer purely aspirational.
- H1 revenue of $727M tracks ahead of the 40/60 script at the midpoint, and management attributes the remaining range solely to customer-site readiness, not demand or shipment capability.
- Margin structure is demonstrably improving: 28.2% GM (+650 bps y/y), 7.1% operating margin, double-digit service margins, six straight profitable service quarters — with guidance held despite an asserted 4% materials tariff hit, implying the cost-reduction engine is absorbing real shocks.
- Capacity doubling to 2 GW at only ~$100M signals both demand confidence and capital-efficient scalability; management frames 2 GW as the start of multi-GW buildout.
- Policy de-risked through 2032: safe harbor covers 2025, BBB's 30% ITC covers 2026+ — removes the pushout risk Benjamin raised and extends the demand runway.
- Competitive position strengthening: CapEx parity vs turbines at matched availability, 15–20 pt fuel advantage, permitting speed, CHP retrofit optionality, skid-mounted mobility — and islanded share of data-center planning reportedly doubling to ~40%.
- Balance sheet managed proactively: 2025 convert overhang removed via 2029 exchange; expansion self-funded.
Bear case- The AWS/Coralogix headline is not incremental volume — it sits inside the already-announced 100 MW AEP PO; the remaining 900 MW is still unconverted pipeline after two quarters of "robust" language.
- Oracle is unquantified: no MW, no revenue, no confirmation it's embedded in guidance (Senyek's direct question got a framework answer, not a yes). The 90-day commitment is a hard execution deadline with reputational downside if missed.
- Tariff math is now murkier, not clearer: 100 bps GM (Q1) vs 4% on materials (Q2) with no reconciliation — and the offset remains asserted, not evidenced.
- 2026 ITC is structurally less generous (flat 30% vs 40–50% safe-harbored) — a quiet demand-economics headwind management acknowledged only obliquely.
- H2 still requires book-build-ship-recognize within the year, reaffirmed verbatim, with backlog disclosure still refused; the capacity-doubling decision is being made on pipeline confidence investors cannot see.
- International "take-off" quietly deferred: 30/70 mix held flat "at least through the next year" implies no near-term international inflection despite four named target markets.
- No permanent CFO three months into the search; the acting CFO gave no strategic or balance-sheet commentary beyond the script.
- The ~$100M capacity capex was not reconciled to the reiterated "capex around FY2024 levels" guide — potential guide inconsistency.
- Prior-quarter narratives (CCS/Chart, hydrogen, SK ecoplant, SVP, repowerings) remain unmentioned for a second straight quarter.
Next-quarter watchlist- Oracle delivery: was power available within 90 days (≈ end of October 2025)? Any MW/revenue sizing, expansion to the "many" projects in discussion, and confirmation of how it maps to guidance.
- Tariff reconciliation: demand a bridge between the Q1 "100 bps GM" and Q2 "4% on materials" figures; check Q2 10-Q for tariff cost detail and whether GM holds ~29% as H2 mix shifts.
- H2 ramp math: Q3 revenue must show the shipment acceleration and inventory drawdown management promised; watch inventory and OCF trajectory versus the "~FY2024 level" OCF guide.
- AEP 900 MW: any conversion of the remaining pipeline, second utility agreement, or "future agreements like this very soon" that K.R. teased.
- Capacity expansion: timeline specificity, whether the ~$100M is inside the capex guide, and any supply-chain commitments behind the 2 GW claim.
- CFO search: permanent hire or explanation for the delay; any further detail on Berenbaum's exit.
- ITC mechanics: evidence that safe-harbored 2025 volume is sufficient; any customer behavior signaling 2026 demand impact from the lower 30% flat credit.
- AI hardware ecosystem orders: whether "new orders soon" from Quanta-like players materialize.
- International: any first order in Taiwan, Germany, Italy, or UK — or confirmation that the two-year take-off clock has slipped.
- Service: whether double-digit service margins sustain, and any Conagra-style long-duration contract additions.
- Silent items: CCS/Chart, SK ecoplant 73 MW resale, SVP, repowering contribution — now two quarters overdue.
|
| Apr 30, 2025 | -8.24% | Q1 FY2025 | Read transcript briefingQuarter in one view- Best Q1 in company history: revenue $326M (+39% y/y, above the +20–30% guide given in February — beat driven by "timing of customer projects"); non-GAAP GM 28.7% (vs 17.5% in Q1 2024, +1,000+ bps, attributed to mix and level-loaded manufacturing); first-ever positive Q1 non-GAAP EPS ($0.03 vs -$0.17); non-GAAP operating income +$13.2M (vs -$30.7M); EBITDA $25.2M (vs -$18.2M); fifth consecutive quarter of service profitability.
- Full-year 2025 guidance reiterated in full: revenue $1.65–1.85B, non-GAAP GM ~29%, non-GAAP operating income ~$150M, OCF and capex "around the same levels" as 2024; revenue split now explicitly framed as roughly 40/60 first-half/second-half.
- Tariffs quantified for the first time: up to 100 bps gross-margin impact if current structure persists all year — but guidance held at ~29% on the claim that accelerated cost reduction will fully offset it; no China supply-chain dependence asserted repeatedly.
- CFO transition announced on the call: Dan Berenbaum exits May 1 (roughly one year in the role); Chief Accounting Officer Maciej Kurzymski becomes Acting Principal Financial Officer while a search runs. Berenbaum gave no reason ("you'll hear more from me, I'm sure").
- Backlog disclosure refused — policy is now explicitly "once a year."
What management is focused on- Demand defense amid macro/tariff anxiety: K.R. insists there is "no slowdown" in AI data centers (citing a recent gathering of large cloud providers and Meta's raised capex, plus a Jensen Huang GTC quote — "your revenues are power limited"); quantifies 2025 committed AI infrastructure spend at $600–700B translating to >15 GW of power need.
- Segment framing with a new nuance: C&I split into two buckets — large-load advanced manufacturing/semis/healthcare (robust; manufacturing construction spend cited at ~$250B/yr in 2023–24 vs ~$85B/yr average over 2000s–2010s) versus consumer-facing retail, where decision cycles "may stretch out" — the first acknowledgment of any demand softness, quickly minimized as "a small part of our business."
- Tariff mitigation as a cultural proof point: no China sourcing, two U.S. factories, custom components with pricing control, ~100 active cost-reduction projects reviewed weekly by K.R.; "we're not going to pass it on to customers... not going to take it on ourselves. We're going to find ways to solve it."
- Islanded, battery-free load-following microgrids positioned as the big C&I shift: "most customers coming and talking to us are asking us for islanded power" — no interconnection, no batteries needed, framed as an advantage given battery supply-chain/tariff issues.
- Utility channel elevated to preferred model for large loads: "working with multiple utilities... both electric utilities and gas utilities," announcements gated on customer permission; direct sales remain for customers who prefer them and for small retail.
- International "rifle approach": Italy, Germany, UK in Europe; Taiwan "in a major way" in Asia (AI supply chain, constrained grid, gas-dependent); markets expected to "take off" within two years — no orders quantified.
- Competitive reframing: the true competitor is clusters of 30–50 MW-class aeroderivative/micro turbines and reciprocating engines, not CCGT; Bloom claims to win on cost, TCO, reliability, time-to-power, noise, water, and emissions, and claims competitors' prices "are actually going up."
Key numbers and quarter mechanics- Q1: revenue $326M (+39% y/y); non-GAAP GM 28.7% (vs 17.5% Q1 2024); non-GAAP operating income $13.2M; EBITDA $25.2M; non-GAAP EPS $0.03 (vs -$0.17).
- GM drivers cited: product mix (repowerings present in the quarter — Berenbaum declined specifics) and deliberate level-loading of the factory, enabled by the balance sheet and demand visibility; management says this should reduce the "extreme peaks and valleys" in quarterly GM seen in 2024.
- Guidance reiterated: $1.65–1.85B revenue, ~29% non-GAAP GM, ~$150M non-GAAP operating income, OCF ~2024 level (~$92M per prior quarter), capex ~2024 level (~$59M); H1/H2 split ~40/60.
- Tariff math: up to 100 bps GM impact if current tariffs persist all year, based on a "detailed portfolio analysis" of suppliers, countries, and re-sourcing flexibility; offset claimed via accelerated cost reduction. Sensitivity to a post-90-day-pause reversion to higher reciprocal tariffs was asked twice (Ailani, Mandloi) and not answered — the 100 bps is explicitly premised on tariffs "remaining the way they are."
- Implied mechanics: Q1 at $326M against a $1.65–1.85B guide and a 40/60 split is consistent with the guide (40% of midpoint ≈ $700M H1); K.R. conceded the company must "book, build, ship and recognize revenue for a portion of our second half revenue" — book-and-ship dependence acknowledged explicitly this quarter.
- No cash, OCF, capex, backlog, or megawatt figures disclosed for Q1 on the call; megawatt disclosure formally retired ("every megawatt is not the same"), backlog now annual-only.
Product and launch scorecard- Islanded/load-following microgrids: now the lead C&I demand driver per K.R.; battery-free load-following is the differentiator. No revenue or order quantification; prior quarter flagged this mix as a margin diluter (third-party gear) — not addressed this quarter.
- AEP/utility channel: K.R. deferred FERC collocation/PUC questions to AEP, relaying that AEP is "very confident" signed projects proceed and has "a very robust pipeline"; "several other utilities" (electric and gas) in discussions, no names, sizes, or timing. No update on the remaining 900 MW.
- Conagra: new 15-year contract announced early April — small size but notable duration; used to explain service model (hotbox/field-replacement units averaging ~5-year life, replaced ~2x over the contract; equipment lifetime certified longer by independent engineers; contracts range 5–20 years, Korea typically 20).
- Repowerings: contributed to Q1 mix/margin strength; cadence described as lumpy ("some quarters there'll be some, some quarters there won't be any") — no sizing.
- International: Taiwan, Italy, Germany, UK named as targets; "next two years" to take off — pre-revenue framing, no milestones.
- Scandium supply: pressed by BMO on prior disclosures of scandium in fuel-cell ink; K.R. stated no China dependence, multiple geographies/continents, Bloom is the world's largest consumer, supply secured "for the foreseeable future" — sources refused as IP.
- CCS/Chart, hydrogen, SK ecoplant, Korea auction, SVP: not mentioned at all this quarter.
Sell-side read-through- Percoco (Morgan Stanley): three-part opener — pipeline timing slippage (K.R.: "maybe they will" shift, but volume is "plenty and enough" for guidance); tariff inclusion in the reiterated guide (clarified: 100 bps impact assumed, offset by cost reduction, 29% held); and the CFO departure (Berenbaum declined any explanation — the call's most conspicuous non-answer).
- Gupta (UBS): utility vs direct channel mix for 2025–26 — answer: both, utility preferred for large loads, "several" utilities in talks, announcement timing customer-gated.
- Ailani (Jefferies): repowering margin strength (no specifics given); tariff sensitivity if the 90-day pause lapses — deflected ("we stand by those numbers... how we do it is internal").
- Rusch (Oppenheimer): critical-materials resilience (no contested/war-zone/China sourcing; COVID-era "battle-tested" claim — never a part shortage or factory slowdown); ex-US/ex-Korea traction (Italy, Germany, UK, Taiwan named).
- Levy/Henry (Truist): C&I demand color (islanded shift, $250B manufacturing construction stat, retail stretch-out admitted); margin trajectory for the rest of the year — no upside beyond guide offered ("we are just reiterating guidance").
- Dendrinos (RBC): tariff cost-saving detail (portfolio of ~100 weekly-reviewed projects, no single lever); FERC/PUC bottleneck risk — deferred to AEP, framed as "temporary blips" where regulators lag markets.
- Unidentified analyst: Q1 backlog size — refused outright ("once a year. That is our policy"); next deal from utility or end user — "hopefully both."
- Thakkar (BMO): scandium sourcing (answered substantively but sources withheld); deployed megawatts level-set — refused, megawatt metrics formally retired.
- Elmaghrabi (BTIG): gas infrastructure for large utility deals — secondary trunk extensions take "a couple months to six to nine months," location-specific, claimed not to be the long pole; Conagra duration mechanics.
- Parks (Tuohy): challenged the CCGT-as-competitor framing — K.R. clarified he meant aeroderivative/micro turbines and recips, a genuine correction; greenfield vs expansion sales cycles — answer: cycles "have to necessarily shrink" as AI capex deploys.
- Mandloi (Mizuho): reciprocal-tariff scenario — answer restated the 100 bps methodology without addressing the higher-tariff case.
Management credibility- Delivered above its own near-term guardrail: Q1 +39% vs the +20–30% y/y guide from February — the first time the back-half-loaded model has been paired with a front-half beat; supports the level-loading/visibility claims.
- Guidance held despite a newly quantified tariff headwind — a confidence signal, but the offset (finding 100 bps of accelerated cost reduction) is asserted, not evidenced, and the higher-reciprocal-tariff scenario was evaded twice.
- CFO departure after ~one year, announced on the call with no reason and a "you'll hear more from me" non-answer, is a credibility negative regardless of cause; interim is the four-year Chief Accounting Officer, which limits continuity risk but the timing (one day after the print) invites scrutiny.
- Transparency is narrowing, not widening: backlog now annual-only (refused when asked), megawatt metrics retired with no replacement, repowering mix undisclosed, tariff sensitivity undisclosed. K.R. explicitly offered guidance reiteration as a "positive signal" substitute for backlog data — investors are being asked to underwrite the H2 book-and-ship ramp on trust.
- Consistency check: the "extreme confidence" language on H2 bookings echoes the pre-Q4-2024 posture that was ultimately delivered — the track record now supports the tone, but the structural reliance on in-year bookings is unchanged and now admitted verbatim.
- Prior-quarter items silently dropped: CCS/Chart (elevated last quarter, absent now), hydrogen, SK 73 MW resale follow-through, SVP conversion, and the Q3 2024 GM discrepancy (23.8% vs 25.2%) — none addressed.
What changed versus the prior quarter- Q1 beat the seasonal guide (+39% vs +20–30%) — the back-half-loaded pattern is intact (40/60 split now explicit) but the front half is starting stronger than scripted.
- Tariffs moved from a qualitative "manageable headwind" to a quantified 100 bps GM impact with an explicit offset claim — and guidance held anyway.
- CFO transition: Berenbaum out May 1; acting CFO named; search underway. Last quarter's sharpest evasion (Berenbaum cutting off BMO) now reads differently in hindsight.
- Disclosure regime tightened: backlog confirmed as annual-only; megawatt metrics formally retired on the call; per-kW metrics already gone — three legacy disclosure channels now closed.
- Demand narrative added a caveat for the first time: consumer-facing retail C&I may stretch decisions; simultaneously, islanded battery-free microgrids became the lead C&I pitch.
- International strategy got specific: Taiwan, Italy, Germany, UK named with a two-year take-off framing (prior quarter was "Asia beyond Korea" generically).
- CCS, hydrogen, SK ecoplant, and the Safe Harbor $12–15B framing all went unmentioned — last quarter's biggest narrative items were simply absent.
- Competition framing refined: aeroderivative turbines/recips named as the real rival (CCGT conceded as not comparable after Parks's challenge), with a new claim that competitor prices are rising.
Bull case- Q1 delivered record revenue, +1,000 bps y/y GM improvement, first positive Q1 EPS, and positive operating income — the seasonally weakest quarter is now profitable, de-risking the 40/60 ramp arithmetic.
- Guidance reiterated with a quantified tariff headwind absorbed — if the 100 bps offset is real, it demonstrates the cost-reduction engine can absorb exogenous shocks; if tariffs ease, it becomes upside.
- Demand evidence is broadening: $600–700B committed 2025 AI capex (>15 GW power need per K.R.), manufacturing construction at 3x historical run-rate, islanded microgrid demand shift, multiple electric and gas utilities in talks, four named international target markets.
- Utility channel optionality intact: AEP confident on regulatory path per management's relay; "several" additional utilities in discussions across both electric and gas.
- Service annuity compounding: fifth straight profitable quarter; Conagra 15-year deal illustrates the replacement-cycle economics (hotbox swaps every ~5 years within long contracts, with technology upgrades embedded).
- Balance sheet enables level-loading, which is visibly smoothing margins — a self-reinforcing execution advantage.
Bear case- The guide still requires booking, building, shipping, and recognizing a portion of H2 revenue within the year — admitted on the call — while backlog disclosure was refused; the market must take the H2 ramp on faith for a second consecutive year.
- Tariff math is fragile: 100 bps assumes the current structure holds; the reciprocal-tariff reversion scenario was asked twice and never answered. The offset is a promise, not a demonstrated number.
- CFO exits one day after the print with no explanation — at minimum an optics and continuity risk during a year that depends on flawless H2 execution and cash conversion.
- Disclosure is contracting on multiple axes (backlog, megawatts, repowering mix, tariff sensitivity) precisely as the revenue model becomes more mix-dependent and harder to model externally.
- First demand crack acknowledged: retail/consumer-facing C&I decision cycles stretching; management minimizes it as small, but it is the first segment-level softness admitted in either quarter.
- Q1 GM strength partly from repowering mix — lumpy by management's own description — so 28.7% may not be a clean run-rate; the ~29% full-year guide implies no margin upside from here.
- Last quarter's flagship narratives (Safe Harbor $12–15B, CCS take-off, SK resolution) received zero update — momentum on those items is unverifiable.
Next-quarter watchlist- CFO search outcome and any explanation for Berenbaum's departure; watch for any 10-Q/8-K detail, restatement, or control-language changes around the transition.
- Tariff trajectory: whether the 100 bps assumption survives the 90-day pause expiry; any quantification of the reciprocal-tariff downside scenario that was evaded this quarter.
- H1/H2 math: Q2 revenue versus the ~40% H1 implication (~$370M+ needed in Q2 at midpoint); any slippage re-opens the Q4-concentration risk.
- Gross margin: whether ~29% holds as repowering mix rolls off and tariff costs land; demand disclosure of the tariff offset's components.
- Utility channel: a second utility agreement (name, size, state, regulatory construct); AEP FERC/PUC outcomes and any update on the remaining 900 MW.
- Bookings evidence: with backlog now annual-only, any announced large orders (utility, data center, or Taiwan/Europe first wins) become the only public proof points for the H2 ramp.
- Retail C&I: whether the admitted decision-cycle stretch widens or infects the large-load bucket.
- Cash flow: H1 OCF trajectory versus the ~$92M full-year guide; inventory build from level-loading; any return to receivables factoring.
- Silent items from prior quarters: CCS/Chart first order, SK ecoplant 73 MW resale completion, Korea auction, SVP conversion, and the Q3 2024 GM discrepancy — each is now overdue for an update.
- Conagra-style long-duration contracts: whether 15–20-year structures become a repeatable template and what they imply for service backlog growth.
|
| Feb 27, 2025 | +4.25% | Q4 FY2024 | Read transcript briefingQuarter in one view- Q4 delivered the promised record: revenue $572M (+60% y/y, +73% q/q), ~40% of full-year revenue in one quarter; non-GAAP GM 39.3% (vs 27.4% Q4 2023); non-GAAP operating profit $133M; non-GAAP EPS $0.43; Q4 operating cash flow $484M.
- Full year: record revenue $1.47B (+10.5%), non-GAAP GM 28.7% (vs 25.8%), non-GAAP operating profit $108M (vs $88M), OCF +$92M and FCF positive for the first time since 2019 ($92M OCF less $59M capex). Service non-GAAP gross profit +$4M for the year vs -$33M in 2023, profitable all four quarters.
- Year-end cash $951M; the large SK related-party receivable was collected in Q4 as promised; zero receivables factoring in Q4.
- 2025 guidance initiated: revenue $1.65–1.85B, non-GAAP GM ~29%, non-GAAP operating income ~$150M, OCF and capex "around the same level" as 2024; Q1 2025 guided up ~20–30% y/y; 2025 revenue pattern expected to mirror 2024's back-half weighting.
- Major new disclosure: customers/financiers used the Treasury Safe Harbor provision to secure full ITC (40% nationwide, 50% in energy communities) for systems placed in service through 12/31/2028 — framed as "$12–15B of gross product revenue" potential and "ITC is no longer a Bloom issue through 2028."
- Backlog disclosed: $2.5B product (would be up ~30% y/y excluding the SK ecoplant 500 MW supply-agreement drawdown) and $9B service (100% attach rate, 5–20 year contracts).
- Note a data discrepancy: Berenbaum cited Q3 2024 non-GAAP GM of 23.8%, versus 25.2% in the prior-quarter summary — worth reconciling against the Q3 press release (possible recast or transcript error).
What management is focused on- "Time to power" is now the entire commercial thesis: majority of 2024 revenue was booked, built, shipped, and recognized in the same year (vs 2–3 year cycles four years ago); K.R. calls it "a velocity business... picking up velocity like crazy."
- Utility-channel replication: AEP-style arrangements (utilities procuring Bloom capacity for data-center load) are the lead growth vector; K.R. says "several utilities" are in talks, with the long pole being state-level constructs to keep data-center costs off ratepayers. No timing given.
- AI demand defense: post-DeepSeek, K.R. argued demand "franticness... is going up and not down," quantifying NVIDIA's quarterly chip shipments as 2–2.5 GW of power need, annualizing to 10–13 GW, >6 GW in the U.S.
- C&I diversification: telecom, retail, manufacturing, education, healthcare orders cited; driven by fear of being deprioritized behind data centers and by price unpredictability. Roughly one-third of the >1 GW deployed base is data centers; C&I growing fast but slower.
- Geographic expansion: Ohio and Illinois wins; Great Lakes/Midwest called "a sleeping giant" (reshoring + gas availability + no T&D cost to ratepayers); Virginia flagged as newly attractive; Northeast dismissed on pipeline constraints. Asia beyond Korea via Bloom's own salesforce and SK-sourced deals (both accepted under the SK contract).
- ITC neutralization via Safe Harbor — the single biggest policy de-risking claim of the call.
- Carbon capture elevated: Chart Industries announcement referenced; K.R. says CCS "is going to take off... in the short-term" faster than hydrogen; hyperscaler interest claimed. Still no orders or quantification.
- Capital efficiency: ~1 GW manufacturing capacity now; tripling for ~$150M when demand visibility warrants; "we will not be the limiting factor."
Key numbers and quarter mechanics- Q4: revenue $572M; non-GAAP GM 39.3%; non-GAAP operating profit $133M (+$106M y/y); EPS $0.43; OCF $484M.
- FY2024: revenue $1.47B (+10.5%); GM 28.7%; operating profit $108M (>60% drop-through on +$140M revenue); OCF $92M; capex $59M; cash $951M.
- Q4 cash mechanics: ~$325M AR benefit (per BMO's read of the cash flow statement) driven by collection of the large SK related-party receivable — promised last quarter, delivered. No factoring in Q4 (vs ~$184M YTD through Q3).
- SK ecoplant: Bloom assisted SK in selling a majority of the 73 MW of energy servers held from a delayed project — this appears to resolve the AWS/Ohio 73 MW overhang, though the transcript does not name AWS or the end buyer.
- Backlog: $2.5B product; ex-SK-agreement, product backlog up ~30% y/y; $9B service backlog. The 2023 SK ecoplant extension (500 MW commitment through 2027) mechanically draws down reported backlog.
- Share count: ~294M fully diluted exiting 2024; Q4 EPS used as-converted method; FY EPS used 227M basic (converts anti-dilutive).
- 2025 guide: $1.65–1.85B revenue, ~29% GM, ~$150M non-GAAP operating income; mid-teens drop-through at midpoint; OCF ~2024 level (~$92M); capex ~2024 level (~$59M); Q1 +20–30% y/y; full-year pattern similar to 2024 (i.e., back-half loaded again).
- Margin guidance only ~flat y/y (28.7% → ~29%) despite 39.3% in Q4: Berenbaum cites mix — islanded microgrids, AI load-following, new products incorporating third-party gear.
Product and launch scorecard- AEP (utility channel): first 100 MW deployed in Ohio (per Blum's question; management did not confirm shipment timing specifics). Remaining 900 MW not in explicit guidance commentary; K.R. deferred to AEP's public statements about data-center discussions in Ohio and other states. Senyek's attempt to confirm all 100 units shipped by year-end was deflected ("we don't talk about specific timing of shipments to specific customers").
- Islanded/load-following microgrids: Quanta expanded into "Bloom's largest islanded load following industrial installation"; capability now a named demand driver but also a margin mix headwind (third-party gear).
- SK ecoplant 73 MW resale: majority of the delayed-project servers resold with Bloom's assistance — clears stranded inventory; buyer, pricing, and margin impact undisclosed.
- Carbon capture (Chart partnership): announcement made, investment committed, hyperscaler interest claimed — no orders, revenue, or timeline quantified.
- Cost reduction: another year of double-digit product cost reduction delivered; K.R. clarified it is per annum (15-year track record ex-COVID), answering BofA's question.
- Capacity: ~1 GW now; tripling for ~$150M; Fremont pull-forward question (Elmaghrabi) answered only with "capacity will not limit us" — no acceptance-level trigger disclosed.
- Korea: steady; wins via annual auction and outside development with SK ecoplant/SK Eternix; SK-sourced deals elsewhere in Asia would count under the SK agreement.
- Hydrogen/electrolyzer: mentioned only as a long-dated option ("when nuclear comes, the hydrogen play becomes really interesting") — effectively deprioritized versus CCS this quarter.
Sell-side read-through- Percoco (Morgan Stanley): got confirmation of more AEP-style utility deals in pipeline but no stage/timing; on funding, management ruled out needing capital for the foreseeable future; backlog-for-2025 vs book-and-ship split refused.
- Dendrinos (RBC): backlog composition by segment refused; extracted the Safe Harbor mechanics and the 40%/50% credit correction (not 50%/60%).
- Rusch (Oppenheimer): revenue-recognition question fizzled (no policy change; product revenue on shipment); tariff answer — no China supply-chain dependence, cost reduction "in the DNA," tariffs a manageable headwind; gas-infrastructure dependency in backlog: "across the board," no single answer.
- Gosai (BofA): ASP strategy post-ITC deflected to value pricing; cost reduction confirmed as double-digit per annum.
- Gupta (UBS): the DeepSeek/demand question — K.R.'s most emphatic moment, demand "going up not down"; midstream/oil-and-gas partnership conversations confirmed happening "at all levels," no handicapping.
- Blum (Wells Fargo): AEP beyond the first 100 MW — deferred to the customer; GM guidance flatness explained by mix shift to solutions with third-party content.
- Elmaghrabi (BTIG): Asia deals can ride the SK agreement; Fremont expansion timing non-answer.
- Senyek (Wolfe): tried to pin AEP shipment recognition in Q4 — refused. Shipment-based recognition reconfirmed.
- Harrison (Piper Sandler): Safe Harbor cutoff = placed in service by 12/31/2028; competitive landscape — turbines sold out 3–4 years, "everybody has a play," competition not the constraint.
- Parks (Tuohy): SMR/nuclear dismissed as irrelevant for 6–8 years (K.R. citing his nuclear engineering background); CCS positioned as the nearer decarbonization play.
- Thakkar (BMO): confirmed the ~$325M AR inflow was the SK related-party receivable; asked for an operating-income-to-OCF rule of thumb — Berenbaum cut him off ("We probably won't. Next question, please"), the call's sharpest evasion.
- Levy/Henry (Truist): CCS attach-rate question — enthusiasm but no attach-rate numbers given.
- Ailani (Jefferies): book-and-ship percentage of 2025 guidance refused; regional ease-of-deployment map given (Midwest/Great Lakes best, Virginia improving, Northeast blocked).
Management credibility- The Q4 deliver-or-break moment was met: revenue, full-year GM (28.7% vs ~28% target), operating income ($108M vs $75–100M guide — beat the top end), and H2/full-year OCF positivity all achieved after three reaffirmations. This is a material credibility deposit.
- Dated promises kept: SK receivable collected in Q4 (explicitly confirmed); service profitable all four quarters; double-digit cost reduction again; no Q4 factoring.
- Guidance posture improved: 2025 guidance given with Q1 guardrails (+20–30% y/y) — more near-term specificity than last year's pattern.
- Remaining strains: the 39.3% Q4 GM vs ~29% 2025 guide implies significant margin giveback — explained by mix but unquantified; the $12–15B Safe Harbor figure is an option-value framing, not backlog or orders, and depends on customers exercising it; backlog-for-2025 coverage and book-and-ship dependence both refused, so the 2025 guide's underpinning is opaque; Berenbaum's curt cutoff of BMO's cash-flow-conversion question was the call's worst moment.
- Language shift to note: last quarter's "extremely confident" bridge rested on projects "in various stages of contracting"; this quarter the tone is delivered-results-based, but 2025 is again guided to be back-half weighted — the same execution risk profile repeats.
- The Q3 GM discrepancy (23.8% cited on this call vs 25.2% previously reported) needs reconciliation — if a recast, it should have been explained.
What changed versus the prior quarter- Guidance risk resolved favorably: the extreme implied Q4 ramp (revenue +73% q/q at 39.3% GM) was actually delivered — the bear-case "binary credibility event" broke positive.
- Cash transformed: $549M → $951M, driven by the SK receivable collection and $484M Q4 OCF; factoring went to zero from ~$184M YTD.
- ITC overhang reframed entirely: from "speed bump" framing and ~40% ITC-independence to Safe Harbor secured through 2028 with a $12–15B gross-revenue option — the biggest single narrative change.
- The AWS/Ohio 73 MW watch item appears resolved via the SK resale (majority sold), though never named as such; SVP contract conversion still unmentioned.
- The promised large data-center announcement from last quarter was never explicitly delivered as a direct-to-data-center mega-deal; instead the AEP utility structure (100 MW deployed, 900 MW pipeline) is now the flagship — a channel substitution worth noting.
- Backlog disclosed with new structure: $2.5B product / $9B service, with the SK 500 MW agreement explicitly identified as a backlog drawdown distorting y/y comparison.
- Per-kW metrics formally retired: revenue now recognized on shipment, solutions mix makes ASP/cost-per-kW "less relevant" — the comparability risk flagged last quarter is now permanent; no replacement metrics offered.
- 2025 guidance initiated ($1.65–1.85B, ~29% GM, ~$150M op income) — implies ~12–26% growth but only ~$42M incremental operating income and flat margin.
- New demand channels surfaced: midstream/oil-and-gas companies in active conversations; CCS (Chart) elevated to a near-term investment priority; hydrogen implicitly pushed out.
Bull case- Delivered the highest-stakes quarter in company history: record revenue, 39.3% GM, $133M operating profit, $484M OCF, full-year FCF positive for the first time since 2019 — the back-half leverage thesis is now proven, not promised.
- 2025 guide implies continued profitable growth with demonstrated drop-through (>60% incremental in 2024); Q1 +20–30% y/y gives near-term visibility.
- ITC risk neutralized through 2028 via Safe Harbor at 40–50% credit levels — removes the biggest policy overhang and potentially improves customer economics versus pre-expiry.
- Demand breadth is real: utility channel (AEP replicable, "several" in talks), data centers (one-third of >1 GW deployed, funnel "strong, diverse and robust"), C&I uptick across five named verticals, midstream interest, international expansion paths in Asia and Europe.
- Balance sheet self-sufficient: $951M cash, no factoring, no capital raise needed per management, capacity tripling available for ~$150M with demand-led timing.
- $9B service backlog with 100% attach and now-profitable service economics creates a growing annuity; management targets further service margin gains.
- Competitive window: turbines sold out 3–4 years, nuclear irrelevant for 6–8 years — Bloom's months-to-power positioning faces no near-term supply-side response.
Bear case- 2025 guidance embeds margin compression: ~29% GM vs 39.3% in Q4 and 28.7% for 2024 — mix shift toward third-party-gear-heavy solutions (microgrids, AI load-following) dilutes margin even as revenue grows; operating income guide of ~$150M is only ~39% growth on ~12–26% revenue growth at the low end.
- Back-half-loaded again: 2025 revenue pattern guided "similar to 2024," meaning another year of escalating Q4 execution risk; Q1 +20–30% y/y is off a small base.
- Book-and-ship dependence undisclosed: majority of 2024 revenue was same-year book-and-recognize; the percentage of 2025 guidance covered by existing backlog was refused twice — the guide rests substantially on orders not yet signed.
- The $12–15B Safe Harbor figure is optionality, not demand; customer exercise, financing, and gas infrastructure all sit outside Bloom's control, and gas-readiness timing in the backlog was admitted to range from months to over a year.
- Direct data-center mega-deal still not announced; the flagship win is a utility intermediary (AEP), and the remaining 900 MW has no disclosed schedule — customer concentration and timing risk persist.
- Transparency still degrading: backlog composition refused, per-kW metrics retired without replacement, cash-flow conversion rule-of-thumb refused, AEP shipment timing refused, Q3 GM figure inconsistent with prior reporting.
- Tariff exposure acknowledged as a potential headwind with only qualitative mitigation claims.
Next-quarter watchlist- Q1 2025 versus the +20–30% y/y guide (Q1 2024 revenue base implied by FY figures); any slip would re-open the back-half-loading credibility question.
- Gross margin trajectory: whether ~29% holds or the Q4 mix benefit persists; quantify third-party-gear dilution in solutions revenue.
- AEP: deployment schedule and revenue recognition for the remaining 900 MW; any second utility agreement announced (stage, size, state regulatory construct).
- Safe Harbor conversion: evidence the $12–15B option is turning into actual orders — volume, pricing, and any customer financing structures disclosed.
- SK ecoplant: details of the 73 MW resale (buyer, margin, remaining units); Korea auction results; any SK-sourced Asia deals outside Korea.
- Direct data-center orders: whether the long-promised hyperscale/greenfield deal ever lands, or whether the utility channel permanently substitutes.
- Cash flow: whether OCF tracks the ~$92M 2024 level given back-half weighting (likely negative H1); any return to factoring; working-capital build for the 2025 ramp.
- Capacity: capex run-rate versus the ~$59M 2024 level; any trigger for the $150M tripling investment.
- CCS/Chart: first order, pilot, or quantified attach economics; midstream partnership announcements.
- Reconciliation of the Q3 2024 GM figure (23.8% vs 25.2%) and any restatement disclosure in the 10-K.
- SVP contract conversion — now two quarters without an update.
|
Corbanu