·Dashboard·Research·Work·Archive
/
← wiki
ticker stockpower-semiconductorsfuel-cellsofcai-infrastructuredata-centerhyperscalerOracleAEPBrookfieldSK-ecoplant updated 2026-05-31

BE — Bloom Energy Corporation

Thesis

Verdict: WATCH — the thesis is real, the entry is bad. Bloom Energy is the only scaled solid-oxide fuel-cell (SOFC) platform monetizing the on-site / behind-the-meter AI data-center power gap that grid interconnect queues (four to seven years in the US) cannot serve in 2026-2028. The product ships in 6-12 months, produces firm power on natural gas immediately, and — per the Dong thesis — natively outputs 800V DC, which uniquely matches NVDA's next-gen rack architecture. After fifteen years as a niche on-site power company with negative gross margins, Bloom crossed positive full-year operating income in FY2024 and the Oracle 2.8 GW frame plus Brookfield $5B partnership have reset the narrative from "niche fuel cell stock" to "AI infrastructure pick-and-shovel."

What has to be true for the bull case: the hyperscaler order book has to keep widening — specifically follow-on orders from existing customers (Oracle, AEP, Equinix) plus at least one new named hyperscaler-scale customer — and Oracle/AEP have to draw against their frames at the cadence implied by FY2027 consensus ($5.2B revenue). If draws slip by 12 months, the consensus path slips and the multiple looks indefensible. The reported backlog grew from ~$6B at end of FY2025 to ~$20B post-Oracle expansion, providing 4-5 years of revenue visibility — but $20B includes the Oracle frame at uncommitted-volume pricing.

Why the entry is bad (the consistent house position across every fragment): BE trades at multiples that price in flawless execution for 3+ years against the mid-to-high analyst growth case. The DCF base-case fair value is ~$136 (roughly half the ~$288 post-Q1 price); probability-weighted fair value ~$160 (~45% downside). The current price requires bull-case revenue AND bull-case margins AND a sub-9% discount rate simultaneously — each independently defensible, together demanding essentially flawless decade-long execution. Naming it accurately matters: this is a momentum + earnings-catalyst trade dressed as a growth thesis, not a value entry.

The asymmetry: Dong's bull 5-year target is $600+ (+107% from $290) vs his downside $55-80 (-72%) — bimodal and tilted to upside, but the price already reflects the bullish 1-year scenario. Buying at $290 means betting that Dong's 5-year $600+ comes through ON TIME; anything slower compresses the multiple. The preferred path (Option 3) is a full thesis position (2-3%) only on a pullback to $190-220 on a Q2 print or any execution wobble. Do not chase at $290. The Apr 27 entry-timing call ("wait for a $164 retest") was wrong on direction — the print blew out and the stock went the opposite way — but the valuation-based WATCH framing still holds.

Snapshot

Bloom Energy Corporation — NYSE: BE (Class A common). Bloom makes the box that lets a hyperscaler put a 100 MW data center on land that does not yet have a grid connection: the Bloom Energy Server, a solid oxide fuel cell that runs on natural gas, biogas, or hydrogen and converts it directly to electricity via a non-combustion electrochemical reaction.

  • HQ: San Jose, California
  • Founded: January 18, 2001 (originally Ion America; renamed Bloom Energy 2006)
  • IPO: July 25, 2018 at $15.00/share ($270M raised)
  • GICS: Industrials → Electrical Equipment & Parts
  • Employees: 2,214 (FY2025)
  • Website: bloomenergy.com

Price snapshot — note the moving target across fragments (all genuine, dated):

  • Apr 27, 2026 close: $234.68 — mkt cap $66.7B, EV $66.4B (profile + checklist base)
  • Apr 29, 2026 close: $287.97 — mkt cap $81.9B post-Q1-2026 beat (mgmt-dd + DCF base)
  • May 1, 2026: $290.52 (~$290) — thesis/SOFC file and checklist final update

Valuation (as of Apr 27, 2026 close, $234.68):

Metric Value
Market cap $66.7B
Enterprise value $66.4B
P/E (TTM) N/A (negative EPS)
Forward P/E 75.2x (74.6x per checklist)
Price / Book 85.5x
Price / Sales (TTM) 33.0x (32.5x per checklist)
EV / Revenue (TTM) 32.8x
EV / EBITDA (TTM) 477.6x
FCF yield ~0.09% (FCF $57M on $66.7B mkt cap); checklist cites 0.3% TTM on $188M TTM FCF
Dividend yield 0%
52-week range $16.05 - $242.20 (14.4x off the lows)
Beta 3.19
50d / 200d MA $163.85 / $108.40 (price +41% / +113%)
Short interest 23.2M shares = 9.34% of float; 2.17 days to cover

The valuation is the single most important fact about this stock. Forward EV/Revenue of ~12x on FY2027 consensus revenue ($5.2B) is more reasonable, but you are paying for the consensus to be right.

Business

What it sells. The Bloom Energy Server is a solid oxide fuel cell — a 100kW+ on-site power generator that runs on natural gas, biogas, or hydrogen, sold or financed to data centers, hospitals, factories, and utilities that need power faster than the grid can connect them. Customers buy the boxes (or have them financed by infrastructure capital), Bloom installs them, then collects long-dated service revenue keeping them running for 15+ years. The business model is hardware-plus-recurring-service: each Energy Server sale drags a multi-decade service contract behind it.

Segments (FY2024 reported, $1.474B revenue):

Segment What it is Approx. %
Product Energy Server hardware sales (the box itself) ~74% (~$1,090M)
Service Long-term O&M contracts (fixed-price, 15+ year terms) ~14% (~$210M)
Installation Site engineering and commissioning (grew 31.8% YoY in FY24) ~10% (~$150M)
Electricity Managed-services / PPA-style revenue ~2% (~$25M)

The exact FY2025 ($2.024B) segment split is not yet broken out at line-item level in the headline release — flagged across all fragments as a not-verified item. The economic logic (and historical pain point) is that Bloom records service revenue but also bears the warranty replacement cost, so the service line was loss-making for years; margin recovery has been the operational story since 2024.

Geographic mix: US-dominant. Korea is the second meaningful market, served exclusively through SK ecoplant. India, Japan, and Europe are smaller. Detailed split not disclosed at segment level.

Assets & operations footprint:

  • Fremont, CA — multi-gigawatt factory. 164,000 sq ft, ~$200M investment, opened with >1 GW annual capacity. Houses ~2,000 California employees.
  • Newark, DE — electrolyzer and SOFC line. Electrolyzer capacity raised to 2 GW. ~715 employees.
  • Sunnyvale, CA — historical R&D / legacy site. Active-manufacturing status not verified.
  • Capacity guidance: on track to double annual production capacity to 2 GW by end of 2026 — this is the rate-limiter on revenue conversion.

Key customers & strategic partners

Oracle (April 2026 expansion). Master services agreement for up to 2.8 GW, with the initial 1.2 GW already contracted and being deployed. Oracle received a warrant on April 9, 2026 to buy up to 3.53M shares at $113.28 (~$400M strike value), expiring October 9, 2026. This is the deal that reset the BE narrative. Largest forward backlog driver.

American Electric Power (AEP). Announced November 2024: framework deal for up to 1 GW, ~$2.65B over the term. As of January 2026 the first 100 MW order has been placed; AEP is calling it the world's largest commercial fuel-cell procurement.

Brookfield (October 2025). Committed up to $5B in infrastructure capital to deploy Bloom systems across global AI factory sites. Stock spiked 20% on the announcement. Brookfield is the financing layer (project-finance, not direct purchase); revenue still flows from end-customers (hyperscalers), Brookfield underwrites them. Removes the customer-financing constraint for non-investment-grade buyers.

SK ecoplant (Korea — primary strategic partner since 2018). Environmental/energy-solutions arm of Korea's SK Group (formerly SK E&C). Bloom's exclusive Korea distributor and a 10%+ shareholder. Cumulative equity investment ~$566M, including original $255M into zero-coupon non-voting redeemable convertible preferred at $25.50/share in 2021. The 2023 expanded offtake locked in 500 MW through 2027 worth ~$1.5B product revenue plus ~$3B 20-year service revenue. In July 2025 SK ecoplant trimmed by selling 10M shares via OTC block trade at $27.60 for ~$276M proceeds but did not exit (residual ~13.49M shares held via Econovation LLC). See Management section for the four-vector related-party scrutiny and the Doosan channel-partner overlap.

Equinix. Decade-long customer since a 2015 1 MW pilot. February 2025 expansion took deployments past 100 MW across 19 IBX data centers, ~75 MW operational and ~30 MW under construction. Reference customer that demonstrates uptime more than it drives the P&L.

Concentration risk: Historically high. SK ecoplant was a disclosed 10%+ related-party customer in past 10-Ks. With the Oracle frame absorbing capacity through end-2027, single-customer concentration is rising not falling — and the shape of the risk is changing from one strategic partner (SK ecoplant) to one hyperscaler (Oracle) to one infrastructure capital provider (Brookfield) simultaneously. Top-1/top-5 % from the FY2024 10-K not pulled.

Moat and competitive position

  1. Operating fleet at gigawatt scale — 1.5+ GW deployed, 15+ years of reliability data. Hyperscaler procurement teams risk-rank vendors on operating history; nobody else in the SOFC space is close.
  2. Vertical integration — designs and manufactures the cells, stacks, system, and service model. Most peers outsource one or more layers.
  3. Customer reference list — Equinix, Oracle, AEP, SK. A reference set that takes a decade to build.
  4. Time-to-power — the structural feature that matters most right now. SOFC deploys in months and produces firm power on natural gas immediately.

The moat is "we got here first and built scale economics" plus switching costs once installed (10-yr+ service contracts) — NOT brand, NOT network effects, NOT regulatory. Competitive vulnerabilities: fuel cells run on natural gas in 2026 reality, not hydrogen, so the decarbonization story depends on H2 economics that are not yet there; reciprocating gensets are cheaper per MW upfront (Bloom's pitch is lifecycle cost, not nameplate); service-segment economics are still the weakest line in the model. Detailed SOFC peer analysis is in Industry landscape — the short version is the moat is wider than the original framing credited (only Doosan is a real near-term SOFC competitor).

Financials

The story is the operating-income inflection. After years of bleeding, Bloom flipped to GAAP operating profit in FY2024 and expanded materially in FY2025. Net income remains negative because of interest expense on the convertible stack and unusual items, but the underlying operating engine has turned.

Income statement & margins ($M):

Metric FY2022 FY2023 FY2024 FY2025 FY2026E FY2027E
Revenue 1,201 1,335 1,474 2,024 3,227 5,237
Revenue growth (YoY) +23% +11% +10.5% +37.3% +59% +62%
Gross margin ~13% (12.4%) ~15% (14.8%) ~27.5% ~29.0% (29.6%) ~32% guide n/a
Operating income (261) (209) 23 73 425-475 (initial guide) n/a
Operating margin -22% -16% +1.6% +3.6% ~14% guide n/a
Net income (301) (302) (29) (88) n/a n/a
Diluted EPS ($1.62) ($1.42) ($0.13) ($0.37) $1.40 $3.12

(Revenue figures vary trivially across fragments: the checklist lists FY22 $1,199M and FY23 $1,333M vs the profile's $1,201M / $1,335M — rounding, not a substantive discrepancy. Diluted EPS row from profile; checklist cites FY26E EPS ~$1.39 and the DCF uses forward EPS $4.02 / $4.00 at the post-print $288 price for its TECC math.)

Cash flow & balance sheet ($M):

Metric FY2022 FY2023 FY2024 FY2025
Operating cash flow (192) (373) 92 114
Capex (117) (84) (59) (57)
Free cash flow (309) (456) 33 57
Cash & equivalents 348 665 803 2,454
Total debt 1,021 1,455 1,530 2,992
Net debt 63 182 326 164
Stockholders equity 341 502 562 769
Shares outstanding (M) 205.7 224.7 229.1 280.0
EBITDA (199) (135) 89 20

FCF discrepancy across fragments (kept, both real): the profile and DCF report FY2025 FCF of $57M (full-year); the buy-checklist reports +$188M TTM FCF as of Apr 27, 2026. These are different measurement windows (FY2025 full-year vs trailing-twelve-month into Q1 2026), not a contradiction. Similarly net debt: profile/DCF cite $164M (FY2025); the checklist cites $542M ($3.0B debt / $2.5B cash) at the Apr 27 checklist date.

The EBITDA puzzle, resolved. EBITDA went DOWN in FY2025 ($89M → $20M) despite revenue +37%. This is a classification artifact, not fundamental (resolved 2026-04-28): FY25 had ~$99M in "Special Income Charges / Total Unusual Items" treated within operating income but excluded from EBIT (and therefore EBITDA = EBIT + D&A). Operating income +$50M YoY is the real read; use op income / OCF as the cleaner gauge of FY25 operating performance.

Balance-sheet expansion (FY2025). The balance sheet jumped because of the upsized ~$2.2B convertible senior notes priced late 2025, plus an October 2025 debt-for-equity exchange that retired ~$443M of old converts in exchange for ~$449M cash and 18.1M new shares (the mgmt-dd fragment characterizes the broader October exchange as ~$976M principal exchanged for ~42.4M shares + ~$988M cash — the larger figure spans the full 2028+2029 note exchange). Cash more than tripled, gross debt nearly doubled, share count grew 22% in one year. Net debt fell despite the gross-debt buildup because cash grew faster.

Capital structure detail:

  • $350M green convertible senior notes due 2029, 3.00% coupon (May 2024 issuance; $402.5M incl. greenshoe, $20.85 conversion strike)
  • ~$2.2B (upsized to $2.5B with greenshoe) 0% convertible senior notes due 2030, $194.97 conversion strike (Nov 2025)
  • ~280M shares outstanding plus 3.53M Oracle warrants (struck $113.28, exp Oct 9 2026)
  • SK ecoplant preferred previously converted; SK still holds residual common stake
  • Convertible overhang ~$2.55B principal, both tranches ITM at $288; implied conversion dilution ~12.5M shares (~4.4%) beyond the 280M base

SBC dilution — the real dilution story. Stock-based compensation jumped $82M → $139M FY24→FY25 (+70%) = 7% of revenue, plus 300k shares granted to CEO Sridhar on Feb 27, 2026 alone. The buy-checklist flags this as the genuine ongoing dilution concern, more than the open-market insider sales. Share count went from 205.7M (FY2022) to 280.0M (FY2025), a 36% increase in three years.

FY2026 guide history (clean beat-and-raise): initial Feb 2026 guide was revenue $3.1-3.3B, non-GAAP operating income $425-475M, gross margin ~32% non-GAAP. Raised after Q1 2026 to $3.4-3.8B revenue and $600-750M operating income. ROIC vs WACC is almost certainly negative today; the bull case is FY26-27 turns it positive. Capex efficiency is exceptional (graded A in mgmt-dd) — FY25 generated +$550M incremental revenue on -$2M capex change, partly mix (Product scales without proportional capex once factories are built), partly capacity funded by service-revenue retention.

Industry landscape

On-site / behind-the-meter power for AI data centers, with SOFC as Bloom's specific niche. The bottleneck has migrated from silicon (2023) to data (2024) to power (2025+): NVDA Rubin/Feynman racks draw ~1MW each (vs 50-100kW Hopper/Blackwell), forcing migration from 48V to 800V DC distribution; regulatory/community resistance has blocked or delayed ~$64B of data center projects; and structural power-price inflation has gas turbines sold out through 2029 at $100-150+/MWh forward.

Bloom's SOFC niche is near-monopoly at gigawatt scale. Of the four named SOFC competitors, only Doosan Fuel Cell (KRX: 336260) is a real near-term threat — a 50 MW SOFC factory (mass production July 2025, Ceres-licensed, Korea-focused), ~20-40x smaller than Bloom's 1-2 GW capacity. GE Vernova has no SOFC program (gas-turbine alternative only). Cummins is sub-commercial (20kW DOE-grant scale). MHI is niche cogen (MEGAMIE hybrid, 2 sites in 7 years at ~250kW). FuelCell Energy (FCEL) uses carbonate cells, not SOFC — lateral, not direct. The real competitive threat is gas turbines (GEV, MHI/Siemens) and future SMRs (NuScale/Oklo/X-energy) — and the thesis is precisely that those alternatives are increasingly non-viable for hyperscaler RFPs.

See sector page: power-semiconductors

Management

Overall management grade: B-, improving from C. The headline question is whether the people running Bloom are aligned with shareholders, competent, and honest with the street. The short answer: yes, with a one-multiple-turn governance discount baked into valuation. Four yellow flags worth monitoring; no red flags at the deal-breaker tier.

Leadership

KR Sridhar — Founder, Chair & CEO (combined role since 2001, CEO 25 years). BS Mechanical Engineering, NIT Tiruchirappalli (1982); MS Nuclear / PhD Mechanical, University of Illinois Urbana-Champaign (1989). Former NASA Ames researcher developing solid oxide electrolysis for Mars-mission O2 generation; after NASA cancelled Mars Surveyor 2001 he inverted the architecture (run hydrocarbons through the SOFC stack to generate electricity) and co-founded Ion America in 2001. National Academy of Engineering inductee 2016. Single-entity continuous founder, no prior failed companies, no SEC enforcement action against him personally, no personal bankruptcy. The "20-year overnight success" arc obscures that the unit economics did not work for most of that time — BE only crossed positive operating income in FY2024.

Simon Edwards — CFO since April 13, 2026. Joined from Groq, where he was CFO then briefly CEO. A narrative-aligned hire for the AI infrastructure rebrand, not for accounting depth — Groq is private and pre-revenue at scale, so Edwards has no public-company SOX battle scars. Whether that is an asset (fresh thinking) or a liability (no playbook for restatement landmines given BE's 2020 MSA history) is genuinely uncertain.

Aman Joshi — Chief Commercial Officer. Owns the customer relationships including Oracle and AEP. Has never been CFO (an earlier profile assumption to that effect was incorrect). Sold 10,000 shares at $135.88 on Apr 1, 2026 = $1.36M; current direct holdings 180,521 shares.

Satish Chitoori — Chief Operating Officer. Manufacturing scale-up lead for Fremont and Newark. Sold 20,000 shares at $204.23 on Apr 14, 2026 = $4.08M; 212,365 shares held.

Maciej Kurzymski — Chief Accounting Officer (interim CFO May 2025 - April 2026). CAO since 2021; stayed through the entire 11-month interim period and continues as CAO under Edwards — a positive signal on transition continuity.

The CFO carousel — three CFOs in 24 months (the loudest governance flag)

CFO Tenure Why they left Signal
Greg Cameron Aug 2020 → Apr 29, 2024 To CF Industries; 6-week gap suggests negotiated voluntary An Immelt hire (ex-GE Capital). Departed right before FY24 op-income inflection — odd timing
Daniel Berenbaum Apr 29, 2024 → May 1, 2025 Twelve months exactly; boilerplate "not the result of any disagreement on accounting" language; no transition period The loudest specific signal. No 60-90 day handoff usually means a broken relationship. Most likely root cause: capital-policy disagreement preceding the $2.2B convert + Oracle warrant program
Maciej Kurzymski (interim) May 2025 → Apr 13, 2026 11-month interim is abnormally long (typical searches close in 4-6 months) Either candidates kept turning the role down, or Sridhar was picky
Simon Edwards Apr 13, 2026 → From Groq Narrative hire for AI infrastructure positioning

Ownership & skin in the game

Total insider ownership 5.82%. Sridhar holds ~2,692,921 shares direct + ~2.6M unvested (the profile's earlier figure of ~4.72M direct+trust conflates these; mgmt-dd splits them) = ~$775M direct + ~$750M unvested at $288, likely >80% of his net worth in BE — strong alignment. He has done 24 transactions over 5 years, all sells, zero open-market buys — but the largest sales were forced 10-year option expirations (e.g., Aug 2025: 216,955 shares at $48.97-$53.79 = $13.04M, exercising options struck $30.89 expiring Sep 10, 2025), not discretionary signaling. The Feb 24, 2026 sale of 200,000 shares at ~$170 (~$34M) was discretionary (~7% of his stake) — rational diversification after a 6x, not a flight. No 10b5-1 plan filed right before bad news appears in the public record.

The director group's wealth is largely elsewhere (Immelt at NEA portfolios, Chambers at JC2 Ventures, Snabe at Siemens) — normal for outside directors but means board personal financial conviction is moderate. All director holdings are stock awards, not open-market purchases.

Board & governance

10-member board, Lead Independent Director structure. 5 of 7 outside directors are genuinely independent.

  • KR Sridhar — Chair (combined with CEO) — not independent
  • Jeffrey Immelt — Lead Independent Director since May 2020; former GE Chairman/CEO; NEA Venture Partner (NEA was a pre-IPO Bloom investor; portfolio incl. Desktop Metal, Tuya, Bright Health, Cleo, Collective Health, Formlabs, Sila Nano, TAE Technologies) + Twilio. Quasi-independent — functions partly as an NEA vehicle; hired Greg Cameron (first of the CFO carousel)
  • Jim Snabe — since Aug 2025; Chairman Siemens AG, ex-co-CEO SAP; genuinely independent, strong recent addition
  • John T. Chambers — since May 2018; former Cisco Chairman/CEO; JC2 Ventures (OpenGov, Pindrop, Quantum Metric, Sprinklr, Uniphore)
  • Michael Boskin — long-tenured; former Chair US Council of Economic Advisers; Stanford Hoover (entrenchment risk)
  • Mary K. Bush — long-tenured; Bush International LLC
  • Cynthia Warner — since June 2023; ex-CEO Renewable Energy Group, prior BP exec
  • Eddy Zervigon — pre-IPO; Quantum Xchange CEO; Chair of Nominating Committee (long-tenured)

Governance flags: combined Chair/CEO; Lead Independent is an NEA partner; no dual-class structure disclosed; staggered-board status not verified; no poison pill in effect. Specific committee chairs (Audit/Comp/Nom-Gov) not extracted — critical to verify the Audit Committee has a designated financial expert with restatement-experience pedigree given the 2020 MSA history.

Compensation & alignment

CEO pay 2024: base $876,923 + bonus $1,579,500 + stock awards $42,394,800 + other $110,522 = $44,961,745 total (CAP $39,968,710). Peer group is the Nasdaq Clean Edge Green Energy Total Return Index — an index, not a custom peer set, a soft governance flag that avoids naming peers (PLUG, FCEL, STEM). Performance metrics: Revenue (Product and Service), Non-GAAP gross margin, Non-GAAP operating income/loss, Cash flow from operations. Max payout uplift raised to 200% (from 150%); one tranche of Sridhar's 2024 PSUs paid out at 300% in year one. A 600,000-PSU package included a 300K "strategic priorities" tranche with undisclosed criteria vesting before Dec 31, 2027. Verdict: generous comp design with low-transparency hurdles — "pay for the easy beat," not "pay for plan execution" — worth haircutting management quality.

Related-party scrutiny — SK ecoplant (four-vector)

SK ecoplant is simultaneously (1) a 10%+ Korean Class A shareholder (sold 10M shares July 10, 2025 at $27.60; residual ~13.49M via Econovation LLC), (2) a major customer with a 500 MW commitment through 2027 (~$1.5B product + $3B service over 20 years), (3) exclusive global distributor, and (4) co-investor in a Korean SOFC manufacturing JV with BE. The textbook configuration for usurpation-of-corporate-opportunity scrutiny. The Preferred Distributor Agreement (PDA) is not publicly filed unredacted — most-favored-customer / ASP discount / take-or-pay terms are not extractable. 2024 10-K Note 17 discloses an SK-receivable balance of ~$93.5M. New live concern: SK ecoplant became Doosan's Korean data-center channel partner (Nov 2025) — if SK is dual-sourcing between Bloom and Doosan, Bloom's Korean monopoly has been quietly broken; track via 10-K Note 17 and any updated PDA terms.

Litigation & enforcement

2019-2020 Securities Class Action (filed May 28, 2019, N.D. Cal.): alleged false/misleading MSA accounting in the IPO registration through Sept 16, 2019. Bloom restated financials in February 2020 for periods Jan 2018 - Sept 2019. Settlement: $3.0 million (small, almost certainly insurance-funded; material in signal value, not dollars). No SEC enforcement action against BE or executives; no personal litigation against Sridhar, Joshi, Chitoori, or Edwards.

Credibility scorecard

Guide-vs-actual 2024-2026 is clean beat-and-raise (the one miss was FY2024 revenue: $1.6B initial guide vs $1.474B actual, -8%). FY2025 beat top of the $1.65-1.85B guide at $2.024B. Q1 2026: revenue $751.1M vs ~$540M consensus (+39%), EPS $0.44 vs $0.13 consensus (+242%) — a massive beat that drove FY26 guide up to $3.4-3.8B. The 2019-2023 era was the opposite (guide-and-cut). Estimated follow-through rate ~75-85%; guidance tendency conservative/sandbagger trending straight-shooter; weasel-language frequency low-moderate (below PLUG levels). Watch-item: whether Edwards continues the conservative posture — the "fully funded… from current resources" language repeated ~6 months after the $2.2B convert is the one to track. Capital allocation grade C+ → B-: Nov 2025 0% convert at $194.97 (52.5% premium, 0% coupon) was the textbook EXCELLENT action; the Oct 2025 debt-for-equity exchange forcing conversion at ~$23 effective strike when the stock had 10x'd was the single worst-timed action (BAD); capex efficiency A.

Catalysts & risks

Catalysts (bull)

  1. Oracle / AI data-center pull — 1.2 GW already contracted under the 2.8 GW master agreement; the single biggest revenue driver in 2026-2027 and the underpinning of the FY2027 consensus jump to $5.2B.
  2. AEP utility procurement — 100 MW initial order placed against the 1 GW / ~$2.65B frame; sets precedent for other regulated utilities to procure SOFC behind-the-meter. Cadence of follow-on orders is the swing factor for FY2027-2028.
  3. Backlog buildout — reportedly ~$20B post-Oracle vs ~$6B at end of FY2025; 4-5 years of revenue visibility at FY2027 consensus.
  4. Capacity doubling — 2 GW annual nameplate by end of 2026 (the rate-limiter on revenue conversion).
  5. Brookfield $5B financing partnership — removes the customer-financing constraint for non-IG buyers.
  6. A second hyperscaler signing a 1+ GW frame in the next 6 months — the specific event that would raise the DCF bull weight; Dong's 800V-DC architecture argument is the underappreciated structural driver (Crusoe's Wyoming 900 MW filing, Sept 2025, is the marquee Bloom-powered all-DC proof point).

Risks (bear)

Risk Likelihood Note
Valuation / momentum / multiple compression High The multiple compresses on any execution miss; structural feature of the setup. Beta 3.19, $66B MC at 32x sales — a 2x compression to 16x sales (still rich) = -50%. A single quarter of flat backlog growth or slipped AEP draw compresses fwd P/E from ~75x to 40-50x = 30-40% correction.
Customer concentration (Oracle as new top customer) High Concentration is rising not falling; shifts shape (SK → Oracle → Brookfield) but cannot be eliminated at this growth rate. Improves naturally as new hyperscalers convert.
Capital-stack expansion / dilution High $2.45B cash buffer and 2 years of FCF-positive offset; Oracle warrant (3.53M shares, ~1.3% dilution) is the next event before Oct 2026; SBC at 7% of revenue is the larger ongoing drag.
CFO turnover / governance Medium Three CFOs in 2 years; closes if Edwards stays 24+ months with clean prints.
Natural-gas price exposure (customer ROI sensitivity) Medium Structural; can be hedged not eliminated; mitigates as gas/power spread widens.
Service-segment economics Medium Closes if Service crosses to GAAP-positive contribution by FY2026.
Permitting / interconnect risk for AEP-style deals Medium Behind-the-meter siting avoids ISO interconnect, but state PSC processes still gate volume.
Margin compression on competitor entry Structural GE Vernova / Cummins / MHI / Doosan named — but only Doosan is a real near-term SOFC threat (20-40x smaller, Korea-focused). "Your margin is my opportunity" — 29.6% GM / 13.3% op margin in fuel-cell hardware is exceptional and structurally vulnerable over the long run.

Insider-selling cluster — resolved from yellow flag to false alarm. The original Apr 27 read flagged ~$32M sold by COO+CFO+officers in the 6 weeks pre-print (Soderberg ~$19.7M, Chitoori ~$7.0M, Joshi ~$4.5M, Kurzymski ~$1.2M). Resolved Apr 28: sales clustered 1-3 days after grants across multiple insiders on the same days = textbook 10b5-1 RSU sell-to-cover, NOT discretionary bearish signal. Drop from the bear case.

Key-person risk. Sridhar is founder, CEO, Chair, and the public face of the technology story; his departure or incapacity would be material. No public succession plan beyond the bench (Joshi CCO, Chitoori COO) and the Immelt/Snabe/Chambers board backstop.

Behavioral-trap audit (checklist): 3 of 6 boxes checked — FOMO (14x off lows), narrative seduction ("AI data center power" is the cleanest narrative in markets), recency bias (extrapolating Q4'25's $778M as a permanent run-rate). The setup is the textbook behavioral trap: parabolic stock + irresistible narrative + earnings catalyst. Discipline says size small, set a hard stop.

Dong's downside case ($55-80): even the bear scenario has BE positive cash flow and growing — 2026E rev only +14%, 2027E +58%, 2028E +40%, op margin compressing to 16% (vs 22% base), exit P/E 16-20x. The downside is "BE grows 25%+ at 16% margin and the multiple compresses," not "BE goes back to losses."

Valuation / DCF

The DCF cannot justify the current price under base-case assumptions, and that is the consistent finding across the DCF, checklist, and thesis fragments. Probability-weighted fair value sits ~45% below the post-Q1 quote.

Three-scenario DCF (3-statement linked UFCF, stock $287.97 Apr 29 close)

Scenario FY35 Revenue Terminal Op Margin WACC Terminal Growth Implied Fair Value vs. $287.97
Bear $15B 18% 11% 3% ~$72 -75%
Base $23B 25% 10% 4% ~$136 -53%
Bull $35B 30% 8% 5% ~$295 +2%

The base-case fair value is roughly half the current price. The current price requires bull-case revenue scaling AND bull-case margins AND a sub-9% discount rate — each independently defensible, together demanding essentially flawless decade-long execution.

Base-case build (UFCF, $M)

Year Revenue Op Inc NOPAT UFCF PV
FY26 3,600 685 541 329 299
FY27 5,237 1,152 910 667 551
FY28 7,200 1,728 1,365 1,061 797
FY29 9,500 2,375 1,876 1,551 1,059
FY30 12,000 3,000 2,370 2,000 1,242
FY31 14,500 3,625 2,864 2,469 1,394
FY32 17,000 4,250 3,358 2,938 1,508
FY33 19,500 4,875 3,851 3,406 1,589
FY34 21,500 5,375 4,246 3,831 1,624
FY35 23,000 5,750 4,543 4,163 1,605

PV of explicit = $11,667M; terminal value (ROIC-based, ~$66B undiscounted) PV = $25,476M; enterprise value $37,143M; less net debt ($164M); less ~$300M Oracle-warrant/dilution adjustment → equity value $36,679M ÷ 290M diluted shares = implied fair value ~$126 (range $125-$140 with adjustments). Revenue decomposition assumes Product compounds ~28% CAGR through 2030 (Oracle/AEP/Brookfield drawdowns) then decelerates to mid-teens; Service grows with the installed base; terminal 25% op margin sits between hardware peers (Cummins ~13-15%, GEV ~10-12%) and IP-heavy industrials (Watts Water ~18%, Roper ~30%).

Two cost-of-equity frameworks (the methodologically interesting part)

CAPM-WACC = 11.3% (rf 4.3%, ERP 5.0%, industry-adjusted beta 1.5 vs Yahoo raw 3.19, cost of equity 11.8%, after-tax cost of debt 2.4%, ~95% equity / 5% debt). Base case uses WACC = 10%. TECC (True Equity Cost of Capital = 1 / forward P/E) ≈ 1.4% at $287.97 and forward EPS $4.02 (forward P/E 71.6x). The gap between CAPM-WACC and TECC is the 9.9-point momentum premium the market awards "pick-and-shovel AI" stories. That works as long as the AI scarcity narrative holds; when it cracks, the multiple compresses by half in weeks.

Reverse DCF — what the price implies

At $287.97 × 290M diluted = $83.5B equity / $83.7B EV. To justify $288, the model needs FY35 op income of $16.2B — roughly 3x the base-case $5.75B. That requires either FY35 revenue of $65B at 25% margin (≈ today's Cummins revenue scale; implies on-site fuel cells become a 25% share of ~30% of the AI data-center power buildout), or $46B at an impossible 35% margin, or the bull combination plus an 8% discount rate. "Not a fantasy, but it requires the Oracle/AEP/Brookfield triangle to be replicated 10x over."

Sensitivity (fair value per share, base revenue path)

Terminal margin / WACC 8% 9% 10% 11% 12%
18% $175 $138 $115 $94 $76
22% $230 $185 $150 $122 $100
25% (base) $275 $215 $175 $145 $115
28% $325 $255 $205 $170 $140
32% (bull) $385 $305 $240 $200 $165

Current price $288 is consistent with terminal margin 28-32% and WACC 8-9% — bullish but not crazy.

Monte Carlo (probability-weighted)

Scenario Probability Fair value Weighted
Bull (Oracle replicates 5x; FY30 rev $18B; 30% margin; WACC 8%) 25% $295 $74
Base (current trajectory; FY30 rev $12B; 25% margin; WACC 10%) 50% $136 $68
Bear (one major hyperscaler walks; conversion slips 18mo; 18% margin; WACC 11%) 25% $72 $18
Probability-weighted fair value $160

~45% downside to weighted-fair from $288; even the 25%-weighted bull ($295) only roughly matches current price.

Dong thesis (exit-PE method — the bull counterweight)

Jianshu Dong's Oct 27, 2025 update ("Highway to Direct Current") sets targets via exit-PE, not DCF, assuming EPS at ~4x consensus: 1-year (YE2026) $200+, 5-year (YE2030) $600+, 10-year blue-sky ~$2,000, downside $55-80. EPS path ~$2.6 (2026) / ~$8 (2027, vs consensus $2.11, +279%) / ~$10 (2028); long-term growth 35%+. Stock was $105 when Dong wrote; already 45% above his 1-year $200+ target with 8 months still on the clock. Three pillars: (1) AI compute deflation 100x/year (not per decade) — Sora-2 text-to-video the next leg; British Rail 1840s at 8% GDP his preferred bubble analogue; (2) power is THE binding constraint (1MW/rack forcing 800V DC; ~$64B of projects blocked; gas turbines sold out through 2029); (3) Bloom's native 800V DC output is unique among baseload tech — TCO $51/MWh DC vs $96 large-turbine AC vs $156 small-turbine AC, a ~15%+ TCO saving that roughly equals a hyperscaler's entire AI-cloud op margin. Dong runs no DCF, ignores the convertible/warrant dilution, and does not address Doosan or governance — these are the gaps to push back on. The DCF (~$136 base) and Dong's exit-PE ($200-606) diverge because Dong assumes EPS at 4x consensus.

Analyst sentiment (Apr 27, 2026; current price $234.68)

24 analysts: 13 Buy (incl. 4 Strong Buy) / 12 Hold / 2 Sell; consensus Buy/Moderate Buy. Mean PT $166.96 (checklist: $165.96, ~28% BELOW current), median $174.00, high $251 (UBS), low $55. The price sat above the mean target and ~90th percentile of the range — diagnostic of a bimodal stock. Recent revisions clustered above $225 (Baird $242, UBS $251, Citi $229, JPM $231), suggesting stale low-side targets drag the mean down; the alternative read is the rally overshot.

Position-sizing implication: small-to-medium size with defined exit discipline, not a core long-term hold at this multiple. Right way to use the DCF: scenario weighting is the actual decision — if your subjective probability that BE replicates the Oracle deal 3-5x is >40%, the price makes sense; below 25%, the math says exit.

Decision log

  • 2026-04-27 Pre-buy checklist run at $231.17 (mkt cap $65.7B). FundamentEdge: 2 of 5 gates fail/warn (margin durability fails; revenue-growth durability and estimate-revision breadth warn). Behavioral-trap audit 3/6. Insider-selling cluster (~$32M COO+CFO+officers, 6 weeks pre-print) flagged as MAJOR YELLOW FLAG. Verdict: WATCH — thesis real, entry bad (14x off lows, +41% above 50dMA, +28% above analyst PT mean). Preferred path: wait for post-print pullback to ~$164 (50dMA), enter 1.5-2% with stop $148; or 0.5% catalyst flyer with hard stop $204. Do not size as a 3-5% conviction position at $231.
  • 2026-04-28 Profile created at $234.68 (mkt cap $66.7B). Oracle 2.8 GW frame + Brookfield $5B = the story; capital structure bloated but cash buffer strong; CFO turnover the biggest governance flag. Trading above mean PT ~$167.
  • 2026-04-28 Checklist forensic resolutions: (a) insider sales almost certainly 10b5-1 RSU sell-to-cover (sales 1-3 days after grants, multiple insiders same days) — drop the $32M yellow flag; (b) FY25 EBITDA decline ($89M→$20M) is a classification artifact (~$99M unusual items in operating, excluded from EBIT) — op income +$50M YoY is the real read; (c) NEW dilution flag — SBC $82M→$139M (+70%) = 7% of revenue + 300k CEO grant Feb 27. Net: thesis slightly cleaner, valuation/technical concerns unchanged. Still WATCH at $231.
  • 2026-04-28 (after close) Q1 2026 print BLEW OUT: revenue $751M vs $540M consensus (+39%); EPS $0.44 vs $0.13 (+242%). Stock $226 → $290 (May 1, +28%). The "wait for $164 retest" entry call was wrong on direction.
  • 2026-04-30 DCF written (post-print, stock $287.97). Base-case fair value ~$136 (-53%); probability-weighted ~$160 (-45%); bull ~$295 (+2%); bear ~$72 (-75%). Reverse DCF: $288 needs FY35 op income $16.2B (~3x base). Position for small-to-medium size with exit discipline, not a core hold at this multiple.
  • 2026-04-30 Mgmt-DD written (stock $287.97, mkt cap $81.9B). Overall management grade B-, improving from C. Green: Sridhar ~$775M direct stake, beat-and-raise, Snabe + Edwards governance upgrades, exceptional capex efficiency. Yellow (4 to monitor): CFO carousel + Berenbaum no-transition exit; combined Chair/CEO with NEA-partner Lead Independent; generous comp (200% max, 300% paid year-one, undisclosed "strategic priorities" grant, index peer benchmark); SK ecoplant four-vector related party with non-public PDA terms; Oct 2025 debt-for-equity exchange = single worst-timed capital action. No red flags at deal-breaker tier. Would trust them with capital — with a one-multiple-turn governance discount.
  • 2026-05-02 Dong thesis + SOFC landscape synthesized. Dong 1Y $200+ already overshot; 5Y $600+; 10Y ~$2k; downside $55-80. SOFC moat WIDER than the Apr 27 Risk #4 framing — only Doosan is a real near-term competitor (20-40x smaller, Ceres-licensed, Korea-focused); GEV has no SOFC, Cummins sub-commercial, MHI niche cogen. NEW concern: SK ecoplant became Doosan's Korean DC channel partner Nov 2025 — quiet erosion of Bloom's Korean monopoly; track via 10-K Note 17.
  • 2026-05-02 Consolidated verdict: valuation-based WATCH still holds at $290. Asymmetry favors waiting — DCF base $136, Dong bull 1Y already overshot, current price at the optimistic end of even Dong's bull projection. Three positioning options: (1) continue WATCH (highest expected value on conventional asymmetry math); (2) tiny exposure 0.5-1% with hard stop ~$220; (3) full thesis position 2-3% only on a pullback to $190-220 (preferred — best entry expected value). Do not chase at $290.
  • 2026-05-30 Consolidated all five BE fragments into this canonical page (W3 reorg). Sector reassigned to power-semiconductors. No facts changed; discrepancies preserved inline (price snapshots $234.68/$287.97/$290.52 by fragment date; FCF $57M FY25 vs $188M TTM; net debt $164M vs $542M; Sridhar share count 4.72M conflated vs 2.69M direct + 2.6M unvested split).

Sources

Vault fragments folded into this page (W3 consolidation, 2026-05-30):

  • be — corporate profile (created 2026-04-28, updated 2026-05-15)
  • be-mgmt-dd — management & governance forensics (2026-04-30, B- grade)
  • be-buy-checklist — pre-buy checklist (2026-04-27 WATCH + Apr 28 / May 2 updates)
  • be-dcf — 3-statement DCF valuation (2026-04-30, base ~$136)
  • be-thesis-and-sofc-landscape — Dong thesis review + SOFC competitive landscape (2026-05-02)
  • (be-filings.md deliberately NOT merged — empty/stub placeholder per W3 instructions)

Primary / external sources cited across fragments:

Topics: ai-infrastructure · nuclear-smr · green-finance Briefing: 2026-04-30 · Bloom Energy (BE) · vault

Consolidation queue (merged 2026-05-30)

These fragment files were folded into this canonical page and remain live pending Pink's archive confirmation.

  • [ ] be-mgmt-dd.md
  • [ ] be-buy-checklist.md
  • [ ] be-dcf.md
  • [ ] be-thesis-and-sofc-landscape.md
  • [ ] be.md

Source updates (auto-maintained)

Drop/BE (Short Case) (Mar 11, 26) - [Updated] Bloom Energy Investment Thesis [Oct 27, 2025]

Dong's October 2025 update raises his 2026/2027/2028 non-GAAP EPS targets to ~$2.6/$8/$10, lifts his 1-year price target to $200+ and 5-year upside to $600+, and introduces a downside case of $55–$80, driven by the all-DC datacenter thesis, Crusoe and Brookfield wins, and anticipated video-generation AI demand.

Relevant to your thesis: Confirms the $600+ bull / $55–80 bear bimodal framing cited in the wiki's asymmetry section, and anchors the 13x 2027E non-GAAP PE entry-point that the current ~$290 price has now run well past.

Source: dropfile://BE (Short Case)/[Updated] Bloom Energy Investment Thesis [Oct 27, 2025].pdf

Drop/BE (Short Case) (Mar 11, 26) - [Updated] Bloom Energy Investment Thesis - Not Investment Ad...

Dong revised his 1-year base case to $100+ and 5-year upside to $500+ (from $50+/$100+ two months prior), citing Bloom's already-achieved LCOE cost parity for AI datacenter RFPs, the Oracle/AEP/AWS partnerships, ITC extension, and 2 GW capacity target — with a downside floor of $35-45.

Relevant to your thesis: This is the September 2025 predecessor to the $600+ 5-year target cited in the wiki's thesis section, confirming that Dong's targets escalated rapidly and the $290 price already reflected his bullish scenario by Q1 2026.

Source: dropfile://BE (Short Case)/[Updated] Bloom Energy Investment Thesis - Not Investment Advice.pdf

Drop/BE (Short Case) (Mar 11, 26) - Bloom Energy Investment Thesis [Not investment advice]

Dong's original June 2025 thesis (at $21.50/share) projected base-case non-GAAP EPS of ~$2 by 2026 growing 30%+ compounded, with a 1-year target of ~$50+ and 5-year target of ~$100+, framing downside as ~$20-30 even in a no-AI/no-ITC scenario.

Relevant to your thesis: Confirms the Dong bull/bear framing already embedded in the wiki — the $600+ bull and $55-80 bear targets reference this same analyst, whose base case was written at $21.50, making the current ~$290 price a 13x move above his original entry point.

Source: dropfile://BE (Short Case)/Bloom Energy Investment Thesis [Not investment advice].pdf

Pages that link here (4)
BE — Bloom Energy CorporationPower Semiconductors & Power Electronics2025-Data-Center-Power-ReportBloom Energy (BE)
Open in Obsidian