BE | Earnings Review — Q2 2026

Verdict: ACCELERATING on every key line — revenue growth, gross margin, EBITDA margin, and non-GAAP EPS — with product mix as the structural driver. Q2 2026 is the strongest quarter in the 12-quarter window and the first billion-dollar+ print.

Print: Revenue $1,065.4M (+165.5% YoY vs $401.2M; +29% vs FMP street $826.1M). Non-GAAP diluted EPS $0.78 (+680% YoY vs $0.10; 2.0× street $0.39). Non-GAAP GM 34.3% (+610 bp YoY). Adj. EBITDA $253.4M (23.8% margin, +1,351 bp YoY). Product revenue $935.4M (+215% YoY, ~88% of total).

Guidance: Third consecutive FY26 raise — rev mid $4.05B ($3.9–4.2B, ~+100% YoY vs FY25 $2.024B); non-GAAP GM held ~34%; non-GAAP OI mid $850M ($800–900M, ~21% OM); non-GAAP EPS mid $2.70 ($2.55–2.85). FMP street already at rev mid ($4.05B); EPS guide mid ~3% above street $2.62. No formal Q3 or FY27 company guide.

Tone: More categorical on “standard / supplier of choice,” more precise on guide architecture (backlog conversion + reserved in-year capacity; no single-project dependence), tighter formal KPI set (supplemental FCF removed from deck; informal CFOA baseline ~$375M+). Open about AI CapEx uncertainty; confident on capacity, services margins (22%, fifth straight double-digit), and time-to-power economics.

Contradictions (4): FY26 opening OI guide transcript garbles ($125–475M vs authentic $425–450M); Q1 YoY “+13.4%” vs true +130% (transcript digit drop); services DD-margin streak reset in Q4’25 transcript; capacity disclosure 2 GW → 5 GW → no-comment without bridge. Backlog vs RPO remains a definitional tension (~$0.44B audited product/install RPO vs multi-billion commercial backlog).

Forward catalysts: H2 conversion (~$2.23B still needed at guide mid after H1 $1.82B); customer deposits $445M combined (+314% YoY); Brookfield $25B shelf / IDF $2.6B draw pace; Project Jupiter gas-permit timing; scandium overhang; Q3 print 2026-10-27 (~$1.05B / $0.66 street).

Consolidated quarterly trend

| Metric | Q3'23 | Q4'23 | Q1'24 | Q2'24 | Q3'24 | Q4'24 | Q1'25 | Q2'25 | Q3'25 | Q4'25 | Q1'26 | Q2'26 | |---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:| | Revenue ($M) | 400.3 | 356.9 | 235.3 | 335.8 | 330.4 | 572.4 | 326.0 | 401.2 | 519.0 | 777.7 | 751.1 | 1,065.4 | | Rev YoY % | — | — | — | — | -17.5 | +60.4 | +38.6 | +19.5 | +57.1 | +35.9 | +130.4 | +165.5 | | Non-GAAP GM % | 31.6 | 27.4 | 17.5 | 21.8 | 25.2 | 39.3 | 28.7 | 28.2 | 30.4 | 31.9 | 31.5 | 34.3 | | GM YoY (bps) | — | — | — | — | -640 | +1,190 | +1,120 | +640 | +520 | -740 | +280 | +610 | | Adj. EBITDA ($M) | 66.4 | 39.8 | -18.2 | 10.2 | 21.3 | 147.3 | 25.2 | 41.2 | 59.0 | 146.1 | 143.0 | 253.4 | | EBITDA margin % | 16.6 | 11.1 | -7.7 | 3.0 | 6.5 | 25.7 | 7.7 | 10.3 | 11.4 | 18.8 | 19.0 | 23.8 | | Non-GAAP EPS ($) | 0.15 | 0.07 | -0.17 | -0.06 | -0.01 | 0.43 | 0.03 | 0.10 | 0.15 | 0.45 | 0.44 | 0.78 | | GAAP diluted EPS | -0.80 | 0.02 | -0.25 | -0.27 | -0.06 | 0.38 | -0.10 | -0.18 | -0.10 | 0.00 | 0.23 | 0.62 |

Absolute revenue — 12 quarters ($M)

Segment revenue drivers ($M)

| Driver | Q3'23 | Q4'23 | Q1'24 | Q2'24 | Q3'24 | Q4'24 | Q1'25 | Q2'25 | Q3'25 | Q4'25 | Q1'26 | Q2'26 | |---|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:|---:| | Product | 305.0 | 261.8 | 153.4 | 226.3 | 233.8 | 471.7 | 211.9 | 296.6 | 384.3 | 638.5 | 653.3 | 935.4 | | Product YoY % | — | — | — | — | -23.3 | +80.2 | +38.1 | +31.1 | +64.4 | +35.4 | +208.3 | +215.4 | | Install | 21.9 | 26.0 | 11.4 | 42.7 | 32.1 | 36.1 | 33.7 | 37.4 | 65.8 | 67.3 | 25.9 | 51.0 | | Service | 47.5 | 52.6 | 56.5 | 52.5 | 50.8 | 53.8 | 53.5 | 54.4 | 58.6 | 61.7 | 61.9 | 69.0 | | Electricity | 25.8 | 16.5 | 14.0 | 14.2 | 13.8 | 10.8 | 27.0 | 12.8 | 10.4 | 10.2 | 9.9 | 10.0 | | Product % of total | 76.2% | 73.4% | 65.2% | 67.4% | 70.8% | 82.4% | 65.0% | 73.9% | 74.0% | 82.1% | 87.0% | 87.8% |

Driver read: Product is the entire story — Q2'26 product +215% YoY to $935.4M (88% of total), with product non-GAAP GM expanding +290 bp YoY to 37.2%. Service accelerating modestly (+27% YoY) with margin recovery to 22.0% (+980 bp YoY). Install lumpy; Electricity residual drag. Mix shift to Product is structural and margin-accretive.

Annual trend + forward

| Year | Revenue $M | YoY | Non-GAAP GM % | Adj. EBITDA $M | EBITDA margin | Non-GAAP EPS | GAAP EPS | |---|---:|---:|---:|---:|---:|---:|---:| | 2021 | 972 | — | 21.7 | 14.0 | 1.4% | -0.55 | -0.95 | | 2022 | 1,199 | +23.3% | 23.0 | 30.1 | 2.5% | -0.41 | -1.62 | | 2023 | 1,333 | +11.2% | 25.8 | 81.8 | 6.1% | -0.10 | -1.42 | | 2024 | 1,474 | +10.5% | 28.7 | 160.7 | 10.9% | 0.28 | -0.13 | | 2025 | 2,024 | +37.3% | 30.3 | 271.6 | 13.4% | 0.76 | -0.37 | | 2026E (co. guide mid) | 4,050 | ~+100% | ~34% | — | — | 2.70 | — | | 2026E (FMP street) | 4,050 | +100% | — | — | — | 2.62 | — | | 2027E (FMP street) | 6,680 | +65% | — | — | — | 4.80 | — |

Three trends jump off the table:

1. Revenue growth is accelerating, not just high. YoY path last five quarters: +19.5% → +57.1% → +35.9% → +130.4% → +165.5%. Step-change starts Q4'25 / Q1'26 as AI data-center product deliveries scale (product mix 74% → 88%).
2. Gross margin expanding with volume, not compressing. Non-GAAP GM +610 bp YoY to 34.3%; last three sequential prints (31.9% → 31.5% → 34.3%) put the company at / through FY26 ~34% guide.
3. Profitability inflection complete and still accelerating. Adj. EBITDA margin 10.3% → 23.8% (+1,351 bp YoY). Non-GAAP EPS 0.10 → 0.78 (+680%). GAAP op income flipped from −$3.5M loss in Q2'25 to $182.2M profit.

Q2 2026 was a large double beat. Revenue +$239M / +29%; non-GAAP EPS +$0.39 / 2.0× street. L4Q beat rate 100% on both lines; L12Q 70% rev / 67% EPS (mixed because 2023–mid-2024 contained misses). Dollar beat magnitude still improving: rev surprise +$92M → +$129M → +$211M → +$239M; EPS +$0.05 → +$0.14 → +$0.32 → +$0.39.

Heatmap (last 8 quarters)

Full history (L12Q)

| Quarter | Rev actual | Rev cons. | Rev surprise | Rev | EPS actual (NG) | EPS cons. | EPS surprise | EPS | |---|---:|---:|---:|:--:|---:|---:|---:|:--:| | 2023Q3 | $400.3M | n/a | — | · | $0.15 | $0.03† | +$0.12 | B | | 2023Q4 | $356.9M | n/a | — | · | $0.07 | $0.08† | −$0.01 | M | | 2024Q1 | $235.3M | $251.4M† | −6.4% | M | −$0.17 | −$0.10† | −$0.07 | M | | 2024Q2 | $335.8M | $307.8M† | +9.1% | B | −$0.06 | −$0.06† | $0.00 | I | | 2024Q3 | $330.4M | $383.7M† | −13.9% | M | −$0.01 | $0.08† | −$0.09 | M | | 2024Q4 | $572.4M | $507.4M† | +12.8% | B | $0.43 | $0.29† | +$0.14 | B | | 2025Q1 | $326.0M | $384.6M‡ | −15.2% | M | $0.03 | −$0.07‡ | +$0.10 | B | | 2025Q2 | $401.2M | $376.0M‡ | +6.7% | B | $0.10 | $0.01‡ | +$0.09 | B | | 2025Q3 | $519.0M | $427.1M‡ | +21.5% | B | $0.15 | $0.10‡ | +$0.05 | B | | 2025Q4 | $777.7M | $648.5M‡ | +19.9% | B | $0.45 | $0.31‡ | +$0.14 | B | | 2026Q1 | $751.1M | $539.9M‡ | +39.1% | B | $0.44 | $0.12‡ | +$0.32 | B | | ▶ 2026Q2 | $1,065.4M | $826.1M‡ | +29.0% | B | $0.78 | $0.39‡ | +$0.39 | B |

† Prior-run public aggregate. ‡ FMP. I = in-line.

| Window | Revenue | EPS | Pattern | |---|---|---|---| | L12Q | 7/10 = 70% | 8/12 = 67% | Mixed | | L4Q | 4/4 = 100% | 4/4 = 100% | Consistent beater | | Post-inflection (2024Q4–2026Q2) | 6/7 = 86% | 7/7 = 100% | Strong beater |

Management variance explanation (not framed as “beat vs street”): (1) first $1B+ quarter from accelerating data-center deliveries; (2) IDF financier model (Nebius offtake) a meaningful Q2 contributor; (3) lumpy campus deliveries, not single-customer dependency; (4) structural opex leverage — rev +166% YoY vs opex +48%; (5) margin mix (product GM 37.2%, services 22%). No quantitative MW/ship vs plan bridge provided.

Headline: Third consecutive FY2026 raise — revenue mid +$450M to $4.05B (~+100% YoY), non-GAAP OI mid +$175M to $850M (~21% OM), non-GAAP EPS mid +$0.65 to $2.70. Gross margin held at ~34%. No formal Q3 or FY27 company guide.

Waterfall — FY26 guidance changes (prior → new → consensus)

Guidance ladder — prior vs new

| Metric | Initial (Feb'26) | Prior (post-Q1) | New (post-Q2) | Δ vs Prior mid | Δ vs Initial mid | |---|---:|---:|---:|---:|---:| | Revenue Low | $3.1B | $3.4B | $3.9B | | | | Revenue High | $3.3B | $3.8B | $4.2B | | | | Revenue Mid | $3.20B | $3.60B | $4.05B | +$450M / +12.5% | +$850M / +26.6% | | Non-GAAP GM | ~32% | ~34% | ~34% | 0 bp | +200 bp | | Non-GAAP OI Low | $425M* | $600M | $800M | | | | Non-GAAP OI High | ~$450M* | $750M | $900M | | | | Non-GAAP OI Mid | ~$450M | $675M | $850M | +$175M / +26% | ~+$400M | | Implied OM @ rev mid | ~14% | ~18.8% | ~21.0% | ~+220 bp | ~+700 bp | | Non-GAAP EPS Low | $1.33 | $1.85 | $2.55 | | | | Non-GAAP EPS High | $1.48 | $2.25 | $2.85 | | | | Non-GAAP EPS Mid | $1.41 | $2.05 | $2.70 | +$0.65 / +32% | +$1.29 / +92% |

*Treat $425–450M as authentic year-start OI guide (see Contradictions — Q4 transcript $125–475M is likely garble).

H1 actual vs full-year guide

| | Amount | |---|---:| | 2026Q1 revenue | $751.1M | | 2026Q2 revenue | $1,065.4M | | H1 2026 revenue | $1,816.4M | | FY26 rev mid | $4.05B | | Implied H2 revenue | ~$2,234M (~$1,117M / qtr) | | H1 share of FY mid | ~45% | | H1 non-GAAP EPS | $0.44 + $0.78 = $1.22 | | Implied H2 non-GAAP EPS | ~$1.48 |

H2 still needs to out-earn H1 — consistent with delivery timing / site readiness, not classical seasonality. Remaining guide is not de-risked solely by H1 beats.

Guide vs consensus (FMP)

| Metric | New Guide Mid | Consensus | vs Consensus | |---|---:|---:|---:| | FY26 Revenue | $4.05B | $4.05B (17 analysts) | In-line (0%) | | FY26 Non-GAAP EPS | $2.70 | $2.62 (15 analysts) | +$0.08 / +3.1% | | Q3 2026 (FQ+1) | No company range | ~$1.045B / ~$0.66 | Street only | | FY27 (FY+1) | No company range | $6.68B / $4.80 | Unendorsed street curve |

Credibility note: Three consecutive raises with large beats is bullish for near-term earnings power and a yellow flag that early guides were conservative or demand was non-linear. Watch whether post-Q3 the company finally holds rather than re-raises.

| Metric | Q3'24 | Q4'24 | Q1'25 | Q2'25 | Q3'25 | Q4'25 | Q1'26 | Q2'26 | |---|---:|---:|---:|---:|---:|---:|---:|---:| | Revenue YoY % | −17.5% | +60.4% | +38.6% | +19.5% | +57.1% | +35.9% | +130.4% | +165.5% | | Rev Accel (bps QoQ) | −2,897 | +7,783 | −2,181 | −1,906 | +3,760 | −2,123 | +9,450 | +3,515 | | Non-GAAP EPS YoY % | −106.7% | +514.3% | NM | NM | NM | +4.7% | +1,366.7% | +680.0% | | Revenue ($M) | 330.4 | 572.4 | 326.0 | 401.2 | 519.0 | 777.7 | 751.1 | 1,065.4 | | Non-GAAP EPS ($) | −0.01 | 0.43 | 0.03 | 0.10 | 0.15 | 0.45 | 0.44 | 0.78 |

Accel = QoQ change in the YoY growth rate (bps). NM = prior-year EPS ≤0. Q2'26 EPS “decel” vs Q1 is a base-effect artifact (lapping $0.03→$0.44 vs $0.10→$0.78); absolute EPS still stepped +$0.34 QoQ and +$0.68 YoY to an 8-quarter high.

Revenue YoY % with inflection annotations

Inflection map

| # | Quarter | Signal | Evidence | |---|---|---|---| | 1 | Q3'24 | Revenue trough | YoY −17.5% (330.4 / 400.3); non-GAAP EPS −$0.01 | | 2 | Q4'24 | First growth inflection | Revenue YoY +60.4% (+7,783 bps accel); non-GAAP EPS 0.43 (+514% YoY) | | 3 | Q2'25 | Growth local min (still +ve) | Revenue YoY decelerated to +19.5% (−1,906 bps) — still expanding | | 4 | Q3'25 | Re-acceleration | Revenue YoY +57.1% (+3,760 bps); capacity roadmap to 2 GW by Dec 2026 framed | | 5 | Q1'26 | Hypergrowth step-change | Revenue YoY +130.4% (+9,450 bps — largest accel); non-GAAP EPS 0.44 | | 6 | Q2'26 | Still accelerating; first $1B+ | Revenue YoY +165.5% (+3,515 bps further) to $1,065.4M; non-GAAP EPS $0.78 (+680% YoY) |

Trajectory assessment: Unambiguously accelerating into hypergrowth. Two consecutive large positive revenue accels in 2026H1 (+9,450 then +3,515 bps). This is not a one-print spike followed by fade — the YoY rate is still rising. Structural read: shift from lumpy C&I fuel-cell growth into AI data-center on-site power at scale, with product mix (~88%) driving both top-line and margin.

| # | Catalyst | Status after Q2 | Timing | Signal | |---|---|---|---|---| | 1 | H2 backlog conversion + in-year bookings | Guide mid ~$4.05B; H1 $1.82B → ~$2.1–2.4B still needed in H2 | Continuous; Q3 print 2026-10-27 | Product rev $935M (+215% YoY) shows conversion is real; FMP Q3 ~$1.045B / $0.66 | | 2 | Deferred rev & customer deposits | Current $327.1M (+69% QoQ, +446% YoY); combined $445M (+314% YoY) | Each quarter BS | Cleanest hard cash-demand KPI; cannot be redefined like “backlog” | | 3 | Audited RPO vs commercial backlog | Product/install RPO $442.4M (2026-06-30); YE25 product backlog $6B + service $14B | Every 10-Q | Definitional gap not closed after a $1B quarter — short thesis wedge remains | | 4 | Brookfield shelf $5B → $25B | Expanded 2026-06-30; pace = offtake, not fixed spend | Multi-year | Bankability validation; related-party rev rotates by delivery window | | 5 | IDF / Oaktree / MUFG / MS $2.6B; Nebius | IDF a meaningful Q2 revenue contributor | Near-term site readiness | Third-party capital converting into product revenue same quarter | | 6 | Hyperscaler validation → named deploy mix | All major U.S. hyperscalers + 12+ neoclouds “validated”; will not split active vs backlog | Continuous logo news | Pure “validated” language largely priced; named CODs move models | | 7 | Oracle Project Jupiter — gas/permit | NM State Land Office rejected Energy Transfer pipeline ROWs twice | Live multi-year | Timing risk for public MW path; FY26 guide not single-project dependent | | 8 | Scandium / Hunterbrook overhang | 8-K + call: not China-dependent; 25 GW visibility; no supplier names | Until third-party proof | Stalemate; reputational/multiple risk more than near-term rev guide | | 9 | Capacity vs 2027 30–40 GW AI build | “Capacity not a constraint”; continuous Copy Exact adds | Quarterly language | Any new GW/yr disclosure is a catalyst | | 10 | Margin sustainability ~34% GM / ~21% OM | Q2 GM 34.3% at/above guide; OI guide $800–900M | Each quarter | Risk: expedite trade-offs and project mix in H2 | | 11 | Q3 2026 earnings | First H2 delivery test | 2026-10-27 | Compare to 2025Q3 $519.0M (~+100% YoY at $1.045B street) |

Hierarchy for the PM: Print catalysts cleared. Highest-signal ongoing positives = deposits + product mix + multi-financier conversion. Highest-signal unresolved risks = RPO definitional gap, Jupiter gas path, scandium, quarterly lumpiness. Street FY26 sits on guide mid — further upside needs another raise or multi-year demand proof, not just another beat of a lagging bar.

Scorecard: 8 Well Answered · 6 Deflected/Avoided (14 Q&A threads). Management confident on demand, time-to-power, services margins, and opex leverage — systematically refused to quantify hyperscaler mix, capacity MW/GW, Brookfield draw timing, scandium usage, and displacement volumes.

| # | Analyst | Topic | Badge | |---:|---|---|---| | 1 | Strouse | Hyperscaler active vs backlog vs pipeline split | Deflected — will not break counts | | 2 | Strouse | Capacity magnitude / acceleration vs last call | Deflected — “not a constraint”; no MW figures | | 3 | Dendrinos | Supply-chain diligence with hyperscalers | Well answered — process description | | 4 | Dendrinos | Brookfield $20B utilization timing | Deflected — shelf framing, no $bn/year guide | | 5 | Arcaro | Project-delay contractual exposure | Well answered — MSA fungibility; financier takes delivery | | 6 | Arcaro | Scandium supply / usage / stockpile | Deflected — 25 GW visibility only; no usage rates | | 7 | Amicucci | FCF guide pull / capital allocation | Well answered — informal CFOA ~$375M+ bridge | | 8 | Amicucci | Training vs inference demand | Well answered — both; $1–2B/mo delay math at 1 GW | | 9 | Kallo | Turbine/engine competitive set | Well answered — total cost to token revenue | | 10 | Kallo | Open-source / Chinese models → power | Well answered — Jevons paradox bull case | | 11 | Gupta | 3–4 year product vision | Well answered — DC + heat + carbon capture | | 12 | Mandloi | Factory CapEx / ROI on expansion | Well answered — ROI “a few months” | | 13 | Ocalan | Fuel-cell data-center market share | Well answered — “very high 90s” (company claim) | | 14 | Rusch | Pricing, target margins, displacement count | Deflected — pivoted to services 22% GM story |

Highest-signal deflections

| Topic | Street wanted | Management gave | Why it matters | |---|---|---|---| | Hyperscaler mix | Active vs backlog vs pipeline counts | “We don’t break it up” | Concentration / conversion risk opaque | | Capacity plan | MW/GW capacity, ramp vs last call | “Not a constraint”; 30–40 GW industry 2027 | Hard to underwrite multi-year volume | | Brookfield timing | Utilization window | Shelf only; offtake-driven | Financing framework ≠ recognized revenue | | Scandium | Usage intensity, inventory months | 25 GW visibility; not China-dependent | Short thesis unresolved without third-party proof | | Displacement | % of book that cancels turbines/engines | Nebius named as example only | Share-shift magnitude unquantified |

Highest-signal well-answered themes

• Time-to-power economics: 1 GW full-stack AI facility ~$12–24B revenue/year; one month delay ≈ $1–2B foregone revenue.
• Project delays: Guide not single-project dependent; Copy Exact fungibility; financier remains on the hook.
• Cash conversion: Latest ~$175M OI raise seen as ~100% drop-through to CFOA → ~$375M+ informal baseline.
• Services quality: Fifth consecutive double-digit services GM at 22% (vs −21% at IPO).

Four genuine contradictions / document-level inconsistencies, ranked by materiality. Branding: red-border alert cards.

| # | Topic | Severity | Likely cause | |---|---|---|---| | 1 | Opening OI guide $125–475M vs $425–450M | High | Transcript garble | | 2 | Q1 YoY +13.4% vs +130% | High | Digit drop in transcript | | 3 | Services DD-margin streak “first” vs chain | Medium | Transcript / redefinition | | 4 | Capacity 2 GW → 5 GW → no comment | Medium–High | Disclosure policy shift | | T | Backlog vs RPO | High for modeling | Definitional (commercial ≠ ASC 606) |

Macro frame — AI power scarcity, not classic macro

Management did not discuss Fed funds, CPI, or consumer sentiment. The entire macro frame is power infrastructure vs AI compute CapEx.

| Theme | Management statement | Read-through | |---|---|---| | Surplus grid capacity gone | Century of ratepayer-socialized surplus is over; new DC load needs new infra | Worse for pure grid-tied schedules; better for islanded/on-site (BE) | | Ratepayers will not subsidize corporate load | Faster/cheaper path is islanded on-site power | Structural tailwind for on-site power OEMs | | 2027 AI capacity build | 30–40 GW new AI data-center capacity turning on in 2027 | Macro sizing of next-year power need vs multi-year interconnect lag | | Time-to-power economics | 1 GW campus → $12–24B rev/yr; 1 month delay ≈ $1–2B; “chips without power are inventory” | Explains pricing power and turbine cancelations | | Jevons / open-source models | Token efficiency up → usage up → more power, not less | Dismisses “cheaper models kill power demand” bear case | | Inference / urban siting | Cannot put a gas turbine in Manhattan; distribution = “surface streets” | Longer-duration TAM beyond multi-GW training campuses | | Legacy backlog as weakness | “A 4-year backlog is not a trophy — it’s a concession of constrained supply” | Negative framing of long turbine/engine lead times |

Named counterparties

| Entity | Role | Q2 / recent fact | Implication | |---|---|---|---| | Oracle (ORCL) | First direct U.S. hyperscaler | Project Jupiter up to 2.45 GW Bloom-only (Q1); standard after Oracle + 55-day power delivery | Better BE; worse turbines on that project; Jupiter gas-permit is timing risk | | Nebius | Neocloud customer | Canceled turbines/engines for Bloom this quarter; IDF offtake meaningful Q2 rev | Concrete displacement + revenue conversion | | Brookfield | Financier | Framework $5B → $25B (5×, Jun 2026) | Capital shelf removes PPA friction; uptake offtake-driven | | IDF + Oaktree + MUFG + MS | Project finance | Cumulative $2.6B; IDF live model | Multi-bank capital still open for SOFC AI power | | U.S. hyperscalers (set) | Pipeline | All majors + 12+ neoclouds/labs/colo “validated” | Broad diligence claim; concentration can still dominate any single quarter | | AEP / AWS (prior) | Utility + hyperscaler channel | Ohio deployments; unconditional FC orders | Dual-track channel proof | | Equinix / CoreWeave (prior) | Colo / neocloud | 100+ MW Equinix; CoreWeave Illinois | Channel diversity beyond hyperscalers | | Turbine / recip OEMs | Competitors | Customers cancel for Bloom; near-term gap large enough for all tech | Near-term co-existence; AI islanded awards favor modular non-combustion when permitting matters | | Other fuel cells | Niche peers | “Very high 90s” share of DC fuel-cell deployments (company claim) | High bar for MCFC entrants | | Hunterbrook | Short research | Scandium / backlog thesis (Jul 8); company 8-K + 25 GW rebuttal | Overhang until third-party verification |

Cross-company synthesis

1. Power is the bottleneck, chips are inventory — elevates BE (and all fast power OEMs) in the AI value chain for 1–3 years.
2. 30–40 GW of 2027 AI capacity is the call’s cleanest industry sizing number.
3. Financing is scaling faster than classic industrial cycles — Brookfield 5× in ~9 months.
4. Displacement is real — Nebius (Q2) and Oracle Jupiter (Q1) are concrete cancel-turbine case studies.
5. Traditional macro is absent — this name’s “macro” is AI CapEx + grid failure + community opposition to combustion and rate-base subsidies.
6. Peer industrials with long DC backlogs get a double-edged message: volume demand is huge, but time-to-power awards increasingly go modular/non-combustion when permitting/community matter.

Bloom printed a hypergrowth double beat with margins expanding and the third FY26 raise in three prints. Product mix at 88% is the structural driver; customer deposits quadrupled YoY are the cleanest demand signal the balance sheet offers. Street has caught the FY26 revenue mid, so the next leg of the thesis is H2 conversion credibility, RPO/backlog definitional clarity, financier shelf draw, and 2027 multi-GW durability — not another lagging-estimate beat. Four transcript-level contradictions (OI guide baseline, Q1 YoY digit, services streak, capacity disclosure) and the backlog-vs-RPO tension remain diligence items; they do not reverse the operating trajectory.

Data sourced from Daloopa