Bloom Energy

Phase A — Understand the business

Lens 1 · Company Overview

Bloom Energy designs, manufactures, sells and (sometimes) installs solid-oxide fuel cell (SOFC) systems — the Bloom Energy Server — that convert natural gas, biogas or hydrogen into electricity through a non-combustion electrochemical reaction. The pitch is firm, always-on, on-site power with ~99.9%+ availability, near-zero criteria pollutants (NOx/SOx/PM), no water use in steady state, and — the line that matters in 2026 — "time to power" measured in ~90 days versus multi-year utility interconnection queues.

Four revenue lines:

• Product (the box itself) — FY2025 $1,531.3M, 76% of revenue, the growth engine.
• Installation — FY2025 $204.1M, lumpy/low-margin (gross loss in FY2025).
• Service (maintenance contracts on the installed fleet) — FY2025 $228.3M, recurring.
• Electricity (PPA / managed-services power sales) — FY2025 $60.4M, shrinking as legacy PPAs are repowered/redeployed.

Customers / channel: mostly direct US salesforce segmented by vertical; expanding utility relationships. The buyer base is now anchored by AI data centers (Oracle's Project Jupiter in New Mexico, 100% Bloom-powered), plus semiconductor fabs, utilities (American Electric Power 1 GW supply agreement), and commercial/industrial. A second product line — the Bloom Electrolyzer (same SOFC platform, makes hydrogen) — is nascent and effectively pre-revenue; it took a $21.9M inventory/impairment write-down in FY2025.

Contract structure / key terms — the load-bearing detail: Bloom does NOT primarily sell power; it sells equipment, and the capital to buy that equipment increasingly comes through structured finance vehicles — the SK ecoplant Korean JV and, decisively, the Brookfield "Financing Structure" of Fund JVs formed in August 2025. Brookfield is the principal owner of those JVs and consolidates them; Bloom uses the equity method and books "Equity in loss of unconsolidated affiliates" (−$40.4M FY2025). Revenue is recognized at customer acceptance (product) — point-in-time, which makes quarterly product revenue inherently lumpy and acceptance-timing-sensitive. This finance-partner-funded, acceptance-recognized, related-party-heavy model is the entire analytical crux of the name (see Lenses 4, 10, 13).

Lens 2 · Supply Chain

Upstream → Bloom → end customer, named:

• Inputs / materials: rare-earth and ceramic powders, zirconia/yttria electrolyte materials, nickel, balance-of-plant components. Bloom does its own cell printing in Fremont, CA and stack/column assembly in Delaware. Specific upstream suppliers are not named in the filing — chokepoint risk on specialty ceramics and on the fuel-cell-grade nickel/rare-earth supply is real but undisclosed (n/a — not named in filing).
• Manufacturing (the company): Fremont (cells) + Delaware (assembly). Capacity plan: ~1 GW today → ~2 GW annual by end of 2026. This ramp is the single biggest execution chokepoint — the order book (Oracle 2.8 GW) is now far larger than nameplate capacity.
• Fuel: natural gas via existing pipeline infrastructure (the systems run 24×7 and need constant fuel). Pipeline access / gas interconnect at each site is a project-level dependency.
• Capital partners (functionally part of the "supply chain" for this model): Brookfield Asset Management (up to $5B deployment financing) and SK ecoplant (Korea JV, manufacturing + equity holder until it ceased being a related party 2025-07-10).
• Distribution / OEM-to-buyer: direct salesforce + utility partners → hyperscaler/industrial sites. Oracle is the marquee channel (Project Jupiter); Equinix is the long-standing reference customer (100MW+ across 19 data centers); AEP the utility offtaker.

Single-source / chokepoint summary: (1) own manufacturing capacity is the binding constraint; (2) finance-partner dependency (Brookfield/SK) is a structural chokepoint — no project financing, no deployment; (3) specialty-materials supply is undisclosed and worth monitoring.

Lens 3 · Competitive Advantages (moats)

What's genuinely defensible:

• Solid-oxide efficiency + the platform. Higher electrical efficiency than reciprocating engines / small turbines, non-combustion (permits faster in air-quality non-attainment zones), modular kW→hundreds-of-MW, serviceable while online. Vs other fuel-cell chemistries (PEM/MCFC/PAFC) Bloom's SOFC has the durability/efficiency edge and no platinum catalyst.
• Time-to-power + DC-native architecture. ~90-day deployment and native DC output compatible with emerging 800 VDC data-center standards — a genuine fit with high-density AI racks that grid power and turbines (2028-2030 delivery slots) cannot match on schedule.
• Installed-base / switching cost. A growing fleet under multi-year maintenance contracts (service revenue $228M, now gross-profitable) creates recurring, sticky economics and reference density (Equinix, now Oracle).
• The capital relationship as a moat. Locking Brookfield into a $5B financing structure is itself a competitive advantage — it removes the customer's capex objection. Rivals without a captive multi-billion financing partner can't underwrite gigawatt deals the same way.

Bargaining power: Currently strong over customers (everyone is power-starved; turbine OEMs are sold out to 2028-2030) — this is a demand-shock-conferred pricing window, not a permanent structural moat. Over suppliers, weaker (specialty materials, and the company is capacity-constrained itself). Honest read: the durable moat is "best non-combustion SOFC platform + installed base + financing rails"; the huge current advantage ("we can power your AI campus in 90 days when no one else can") is cyclical and time-boxed — it compresses the moment turbine/engine capacity catches up (2027-2029) or the grid queue clears.

Lens 4 · Segments

Bloom reports by product line (above) and discloses geography only loosely. The revenue mix by line, FY2023→FY2025:

Then Q1 FY2026 detonates the trend: total $751.1M (+130.4% YoY), product $653.3M (+208.4%), gross margin 30.0%. Product went from 76% of FY25 revenue to ~87% of Q1-26 revenue — the company is now overwhelmingly a box-shipping business to AI data centers.

The geographic/customer cut that matters more than product mix is concentration (Lens 13): related-party revenue was $892.0M in FY2025 — 44% of total — and $373.3M in Q1-26 alone — ~50% of the quarter. The growth is real but it is narrow: a handful of JV-financed hyperscaler projects.

---

Phase B — Measure performance

Lens 5 · Earnings Result (latest print: Q1 FY2026, ended 2026-03-31)

The defining print in the company's history:

• Revenue $751.1M vs $326.0M PY → +130.4%. Product $653.3M (+208.4%), driven explicitly by "multiple projects executed through the joint venture with Brookfield, including a major hyperscaler project".
• Gross profit $225.5M → 30.0% margin (vs $88.7M / 27.2% PY).
• Income from operations +$72.2M (vs −$19.1M PY) — Bloom earned in ONE quarter what it earned in all of FY2025 at the operating line.
• GAAP net income to common +$70.7M; EPS $0.25 basic / $0.23 diluted (vs −$0.10 PY) — first GAAP-profitable quarter.
• Non-GAAP EPS $0.44 vs consensus ~$0.09–0.14 — a ~4× beat.
• Guidance RAISED hard: FY2026 revenue $3.4–3.8B (from $3.1–3.3B), ~80% growth at midpoint; non-GAAP gross margin ~34%; non-GAAP operating income $600–750M; non-GAAP EPS $1.85–2.25.
• Balance-sheet / working-capital flags: customer deposits +$72.9M in the quarter (forward-demand signal, bullish); contract assets rose to $305.9M (revenue recognized-not-billed — watch the AR↔contract-asset conversion); SBC $57.0M in the quarter (8% of revenue).
• Market reaction: shares surged ~16% post-market on the print and the stock is up 1,300%+ over the trailing year. The tape is pricing acceleration as the new normal.

FY2025 full-year for the record: revenue $2,024.0M (+37.3%), gross margin 29%, operating income +$72.8M (first full-year GAAP operating profit; FY24 +$22.9M, FY23 −$208.9M), but net loss to common −$88.4M (EPS −$0.37) — the net loss was entirely below the line: debt-conversion inducement −$66.2M + loss on extinguishment −$32.3M + equity-in-loss of JV affiliates −$40.4M, from the November 2025 0%-notes refinancing and the Brookfield JV formation. Operationally the company turned the corner in FY2025; financially the capital restructuring masked it.

Lens 6 · Earnings Calls (sentiment trend)

No transcripts on disk (transcripts/ empty). From web coverage: management tone has shifted from "path to profitability / margin discipline" (2024-2025) to "Bloom is becoming the standard for on-site power," "100% adoption in major projects like Project Jupiter," and an explicit demand-pull narrative around AI. The recurring phrases now are "time to power," "standard for on-site power," and capacity ("1 GW → 2 GW"). What they've stopped emphasizing: cash-burn/runway anxiety and the electrolyzer/hydrogen TAM (de-emphasized after the FY25 write-down). The tone is the most confident in the company's public history — which, at 27× sales, is itself a contrarian flag. `` — verify against the primary transcript before quoting management directly.

Lens 7 · Comps

Peer table — on-site / data-center power. Multiples are `` with source/date; market caps as of mid-June 2026.

Dividend yield: BE pays none; ROE not meaningful (accumulated deficit $3.99B, just-turned-profitable) — 5-yr avg ROE n/a — negative equity history.

The comp read is the whole bear thesis in one row: Bloom trades at ~4× the EV/Sales and ~4× the forward P/E of GE Vernova, a vastly larger, profitable, backlog-rich beneficiary of the same AI-power tailwind. Either Bloom's growth/margin trajectory is structurally superior enough to justify a 4× premium, or the multiple is the trade. GuruFocus pegs GF Value at $26.32 vs a ~$291–329 price — >1,000% above fair value.

Lens 8 · Stock-Price Catalysts (what moves >5%)

Pattern over the cycle:

• 2019-09 (Hindenburg short report): −21% to ~$3.31 — accounting/liabilities fear (see Lens 10).
• 2024-2025: grind off lows; the catalyst was AI-power demand recognition + each hyperscaler/utility signing.
• 2025-10-28 (Oracle partnership announced): +25.7%.
• 2026-04-13 (Oracle expansion to 2.8 GW / Project Jupiter): further leg up.
• 2026-04-29 (Q1 print + guidance raise): +16% post-market; whole fuel-cell complex ripped (FCEL +32%, PLUG +9%).
• Analyst PT hikes (Barclays → $254, Bernstein init $276) moved it on the margin.

What the market actually reacts to: (1) named hyperscaler/utility deal scope (GW contracted) and (2) the earnings beat-and-raise cadence. It is a momentum/AI-derivative name — it trades on deal headlines and the AI-capex meta-trade, not on DCF. That cuts both ways: the same sensitivity that drove a 13× year will drive a violent de-rate on any AI-capex wobble or a single delayed/cancelled GW.

---

Phase C — Judge people & books

Lens 9 · Management

K.R. Sridhar — Founder, Chairman & CEO since 2001 (~25 years). Ex-University of Arizona aerospace/mechanical-engineering professor; the SOFC stack is literally his invention (originated from a Mars-life-support oxygen-generator concept). NAE inductee (2016); Time "Tech Pioneer" (2009).

• Track record: the quintessential mission-founder — took a moonshot solid-oxide technology from lab to a $2B-revenue, GAAP-operating-profitable company over two decades. But also: ~20 years of cumulative losses (accumulated deficit $3.99B ) and a 2019 restatement on his watch. The profitability inflection is recent and partner-financed, not a long track record of capital discipline.
• Skin in the game: owns 1.36% ($1.17B at current price) — large in dollars, modest in percentage for a founder-CEO. Total comp ~$3.5M (73% equity/bonus).
• Capital allocation: the honest grade is mixed-trending-better. Heavy serial dilution (shares 205.7M FY22 → 280.0M end-FY25, ~36% growth in three years, much via convertible-note inducement ). SBC is high and rising ($145M FY25, +75% YoY). BUT the November 2025 refi to 0% convertible notes due 2030 that added ~$1.4B to cash was genuinely smart — zero coupon, pushed maturities out, funded the capacity ramp. ROE/ROIC turning positive only now.
• Red flags: (1) insiders net-sold ~$83M over the trailing 12 months into the run — directionally bearish signal; (2) large new CEO equity award granted 2024-12-18 inflated FY25 G&A/SBC; (3) the related-party/JV web (SK ecoplant, Brookfield) concentrates revenue and complicates the accounting (Lens 10).
• Archetype: visionary founder-engineer, not a capital-allocation operator. Bull: the conviction and technical credibility that built the platform. Bear: founders in love with the mission tend to dilute relentlessly and under-police accounting complexity — both are present here.

Lens 10 · Forensic Red Flags

Auditor: Deloitte & Touche LLP (since 2020), unqualified opinion on FY2025 financials AND on internal control (ICFR effective) as of 2025-12-31, dated 2026-02-09. Clean — important, given the history.

Critical Audit Matters flagged by Deloitte (the auditor telling you where the bodies could be):

1. Product revenue recognition timing (customer acceptance) — high judgment; auditor confirmed acceptance dates with customers.
2. Consolidation assessment of the Brookfield Fund JVs (ASC 810) — whether Bloom should consolidate. This is the single most important accounting judgment in the file: Bloom keeps the JVs off its balance sheet (equity method) while booking the product revenue from selling boxes into them. Aggressive-but-audited.

Self-identified risk areas, every figure labeled:

• Revenue / related-party concentration: $892.0M of FY2025 revenue (44%) and ~50% of Q1-26 was related-party. Revenue recognized but not yet billed (contract assets) climbed to $305.9M at Q1-26 — watch the conversion to cash.
• Cash flow vs earnings: FY2025 operating cash flow +$113.9M and FCF ≈ +$57.1M (capex $56.8M) — cash generation is real and positive, which materially de-risks the old Hindenburg "burns cash, hides liabilities" thesis. But inventory built +$119M (to $643M) ahead of 2026 demand — a bet on the ramp; if deals slip, that's stranded working capital.
• SBC flattering non-GAAP: $145M FY25 / $57M in Q1-26 (8% of revenue) is added back to the headline non-GAAP EPS ($0.44 vs $0.25 GAAP). The 4× "beat" is a non-GAAP number — the GAAP profit ($0.25) is the honest figure.
• Off-balance-sheet financing: financing obligations $244M + non-recourse/managed-services structures + the equity-method JVs. This is the lineal descendant of the 2019 MSA controversy — the structure that hid service liabilities then is the structure funding growth now. Audited, but inherently opaque.
• Deferred-tax / going-concern: full $872.6M valuation allowance on domestic DTAs; $2.3B federal + $1.5B CA NOLs — management does not yet believe future US taxable income is "more likely than not." A telling internal signal that durable GAAP profitability is not yet assumed even by the company.
• Recurring impairments: $21.9M electrolyzer inventory/asset write-down + $12.7M CIP impairment in FY25 — the hydrogen story is being quietly de-funded.

Regulatory findings:

• SEC EDGAR EFTS (LR + AAER), 2021-2026: zero Litigation Releases and zero AAERs naming Bloom Energy.
• Historical (material, web): the September 2019 Hindenburg Research report alleged ~$2.2B in undisclosed MSA service liabilities, overstated fuel-cell life, and pre-IPO write-down timing. Bloom subsequently filed an 8-K stating prior financials "should not be relied upon" and restated its MSA revenue accounting. A Delaware Court of Chancery books-and-records demand (Jacob v. Bloom Energy) was partially granted in 2021. This is resolved and old, but it is the reason the JV/off-balance-sheet structure deserves permanent scrutiny.
• Non-SEC enforcement (FTC/DOJ/etc.): no material current hits surfaced.
• 10-K Item 3 (Legal Proceedings): no material undisclosed litigation flagged beyond ordinary course in the FY2025 filing.
• Net: No current regulatory/legal finding — verified via SEC EDGAR EFTS (LR, AAER), web search, and 10-K Item 3 as of 2026-06-21. The forensic risk is structural (accounting complexity + related-party concentration + non-GAAP reliance), not an open enforcement action.

---

Phase D — Project & stress-test

Lens 11 · Forward Projection (FY2026 / FY2027 / FY2028 non-GAAP EPS)

Built bottom-up from the Q1-26 actual + management's own FY2026 guide. No forecast.ts create in watchlist mode — base case logged below for the record only.

Anchors: FY2026 guide — revenue $3.4–3.8B, non-GAAP GM ~34%, non-GAAP op income $600–750M, non-GAAP EPS $1.85–2.25. Q1-26 already did $751M rev / $0.44 non-GAAP EPS, so the guide implies acceleration through the year (back-half loaded on capacity).

`` — every line derived from the FY26 guide grown at scenario rates; non-GAAP→bear EPS falls despite flat-ish revenue because installation-mix and SBC drag the margin. GAAP EPS runs ~$0.80–1.20 below non-GAAP on SBC + below-the-line financing items.

Base case logged (not committed to forecast.ts in this loop): BE FY2026 non-GAAP EPS >= $2.00, p≈0.60, resolves 2027-02-28. The valuation problem is independent of which scenario hits: even the bull $5.20 FY28 non-GAAP EPS against a ~$329 price is a ~63× FY28 P/E — three years out, on the optimistic case, on a non-GAAP number. The earnings can triple and the stock can still be expensive.

Lens 12 · Bull vs Bear

Bull case. Bloom is the only on-site power technology that ships in 90 days, runs non-combustion (permits in air-quality-constrained AI corridors), outputs native DC for 800 VDC AI racks, and comes pre-wrapped with a $5B Brookfield financing rail that removes the customer's capex objection — at the exact moment US data-center power demand nearly triples to ~134 GW by 2030 and turbine OEMs are sold out to 2028-2030. The Oracle relationship (2.8 GW, Project Jupiter 100% Bloom) is a template other hyperscalers will copy; AEP (1 GW) proves utility-scale offtake. The company just printed its first GAAP-profitable quarter with positive FCF, 30% gross margins inflecting toward 34%, and an 80%-growth guide. This is a genuine secular winner in the defining infrastructure bottleneck of the decade.

Bear case (2-3 ways it permanently impairs — or just de-rates 60%):

1. Valuation is the position. ~27× 2026 sales / ~135× forward earnings / ~1,062× forward FCF vs GE Vernova at ~6× sales / 31× earnings for the same tailwind. GF Value $26 vs ~$300+ price. You can be right on the business and lose 60% on multiple compression alone.
2. Concentration + cyclical moat. ~50% related-party revenue funneled through Brookfield/SK JVs; the "90-day advantage" evaporates as turbine/engine capacity (CAT +$725M engine capex; GE Vernova; Cummins) and grid queues catch up by 2027-2029. The demand-shock pricing window is not a permanent moat.
3. AI-capex sensitivity. This is a high-beta derivative of the AI-capex meta-trade. Any hyperscaler capex pause, an Oracle/Project-Jupiter delay, or a single cancelled GW re-rates the whole non-GAAP-EPS-times-135 stack violently. Inventory is already built ($643M) against deals that must close.

Pre-mortem (18 months out, thesis broke): It's December 2027. A hyperscaler trimmed 2028 capex; Oracle pushed two Project-Jupiter phases right; a competitor (GE Vernova co-location plant, or a turbine-supply easing) won the next two gigawatt RFPs on price. Bloom still grew revenue ~15% and is more GAAP-profitable than ever — and the stock is down 65%, because it was priced for 50%+ growth in perpetuity and "merely good" is a catastrophe at 135× earnings. The business was fine; the multiple was the trade.

Are multiples too high? Yes, unambiguously — on every framework (P/E, EV/S, P/FCF, GF Value, vs peers). Contrarian view of what the market refuses to see: the market is treating the Oracle 2.8 GW deal as a floor (the first of many) when the prudent base case is that it's closer to the ceiling of near-term visibility — capacity-constrained at 2 GW, concentration-heavy, and lapping +130% comps that mathematically must decelerate. The consensus PTs ($203–283 median) sit below the ~$329 price — even the sell-side, mid-euphoria, can't reach the tape.

Lens 13 · Devil's Advocate (short-seller)

Dismantling the bull:

• Revenue concentration is the kill-switch. ~50% of Q1-26 revenue ran through Brookfield/SK related-party JVs. This is not diversified hyperscaler demand — it's a few financed mega-projects. If Brookfield's appetite cools, or one anchor project slips, revenue and the multiple collapse together.
• The moat is rented, not owned. Bloom's decisive advantage today is availability — it can deliver when turbines (2028-2030 slots) can't. That's a supply-chain accident of timing, not a durable structural moat. The most dangerous competitor bulls underestimate is GE Vernova (and the recip-engine duo CAT/Cummins): vastly larger, profitable, building capacity now, and able to undercut on $/kW once their queues clear. Bloom's SOFC is more efficient, but the buyer in a power panic optimizes for time and certainty, and that window closes.
• The accounting rhymes with 2019. The same off-balance-sheet/managed-services structural DNA that hid ~$2.2B of MSA service liabilities pre-restatement is the JV/financing-obligation machinery powering growth now. Deloitte signs it (twice flagging it as a Critical Audit Matter), but ASC 810 consolidation of the Brookfield JVs is a judgment — and judgments at the edge of aggressiveness are where this management has erred before.
• Worst capital-allocation tells: relentless dilution (205.7M→280.0M shares in 3 yrs), $145M SBC, $83M net insider selling into the run, a full $872.6M DTA valuation allowance (management itself doesn't underwrite durable US profit).
• What must hold for ~$329: ~50%+ revenue growth sustained for years, GM to 34%+, Oracle 2.8 GW deploying on schedule, and no AI-capex wobble — simultaneously.
• If growth disappoints 20-30%: at ~135× forward earnings, a deceleration from 80% to ~50% growth is a multiple-halving event; a deceleration to 15-20% (the bear path) is a 60-70% drawdown even with rising GAAP profit.
• Single permanent-impairment scenario: an AI-capex retrenchment in 2027-2028 coincident with turbine-supply normalization — Bloom loses the "only one who can deliver" status, the JV financing partner pulls back, the $643M inventory strands, and the stock round-trips most of the 13× move. Plausibility: moderate (~25-30%), and the payoff is enormous — which is the entire short setup.

Lens 14 · Management Questions (ordered by information value)

1. Of FY2026 revenue, what % is contracted-and-financed (Brookfield/SK JV-backed) vs. on Bloom's own balance sheet, and how concentrated is the order book among the top three hyperscaler end-customers?
2. What is committed Brookfield deployment capital remaining of the $5B, and what triggers a further draw vs. a pullback?
3. On the Oracle 2.8 GW: how much is firm PO vs. intent, what is the contracted deployment schedule by half-year through 2027, and what are the cancellation/reschedule terms?
4. What is the realistic exit-2026 and exit-2027 manufacturing capacity, and what is the gating constraint (specialty materials? labor? capital?) on going past 2 GW?
5. Walk through the ASC 810 judgment on the Brookfield Fund JVs — under what change in facts would you have to consolidate them, and what would that do to the balance sheet and reported revenue?
6. What is the all-in delivered $/kW of an Energy Server today vs. a gas reciprocating engine and a small gas turbine, and how does that gap evolve as turbine supply normalizes in 2028-2029?
7. When do you expect to release the $872.6M domestic DTA valuation allowance — i.e., when does management underwrite durable US GAAP profitability?
8. Bridge GAAP EPS ($0.25 Q1) to non-GAAP ($0.44): how should investors think about the SBC run-rate as a true economic cost as the share count grows?
9. What is the expected share-count trajectory through 2028 given convertible-note conversion (0% notes due 2030), the Oracle warrant (3.53M @ $113.28), and ongoing equity comp?
10. What is contracted service-attach and the long-run service gross margin as the fleet ages — and how confident are you in the fuel-cell life assumptions underpinning service liability accruals (the 2019 issue)?
11. How exposed is project economics to IRA Section 48 ITC / 45V / 45X — and what is the downside if those credits are curtailed?
12. What is the natural-gas / fuel-supply and interconnect dependency per site, and how do you de-risk gas availability for a 100%-Bloom-powered AI campus?
13. Where is the electrolyzer/hydrogen business after the FY25 write-down — funded growth or harvest/wind-down?
14. What is the realistic gross-margin ceiling for a product-heavy mix at scale, and how much is manufacturing automation vs. pricing power conferred by the current supply shortage?
15. What single competitive development would most threaten the 2027-2028 plan, and how are you positioned for it?

---