Bloom Energy (BE) — Question Tree to 2029 Cash Flows

Recursive decomposition of "What could BE's cash flows look like in 2029?" — click any node to expand
New to Bloom Energy? Read the business & numbers summary first — it explains what BE does, how it makes money, and walks through the actual financial history before this tree decomposes the 2029 question.

Live Snapshot (May 18, 2026)

Stock: $258.71 Market cap: $73.6B EV: $79.0B Net debt: $0.5B (Cash $2.5B − Debt $3.0B; most debt is converts) Shares: 284M 52w range: $17.01 – $310.00 TTM revenue: $2.45B (+130% YoY) TTM EBITDA: $232M (9% margin) TTM FCF: $266M Gross margin: 30.1% Fwd P/E: 62x P/B: 80x
KNOWN grounded in disclosed data   PARTIAL estimated with reasonable confidence   UNKNOWN needs research / not yet decomposed
ROOT: What could Bloom Energy's cash flows look like in 2029?

Pinning a number requires answering: 2029 revenue × 2029 operating margin × FCF conversioncash dilution from converts. Each of these decomposes into a sub-tree. Then we assemble a bear/base/bull/lottery range and translate to per-share value.

Preliminary scenario skeleton (to be refined by deeper branches):

Scenario2029 RevOp MarginOp IncomeFCF (~70%)Shares (M)FCF/share
Bear$5.5B8%$440M$310M320$0.97
Base$9B14%$1.26B$880M305$2.89
Bull$15B20%$3.0B$2.1B295$7.12
Lottery$22B25%$5.5B$3.85B285$13.51

At today's $258 stock: bear case implies ~250x FCF (no), base implies ~90x FCF (still high), bull implies ~36x FCF (reasonable for high-growth), lottery implies ~19x FCF (cheap). The market is pricing in something between bull and lottery.

Now decompose each input.

Q1: What is 2029 revenue? PARTIAL

Revenue = MW shipped × $/MW (hardware) + installed-base × $/MW/yr (service) + electricity-as-service margin (PPA business) + electrolyzer revenue. Need to size each leg.

Q1.1: How many MW will BE ship per year by 2029?

2025 shipments: roughly 700-900 MW (estimate from $2.45B revenue at ~$3M/MW system price). Need to grow this 5-10x for the bull case. After the manufacturing deep dive, the bottleneck looks more like demand/site-readiness and supplier lead times than factory square footage, but the 5 GW capacity claim still needs shipment proof.

Q1.1.1: What is BE's current and planned manufacturing capacity? RESEARCHED   → deep-dive page

Updated conclusion: the SEC-filed hard baseline is Fremont going from ~1 GW to 2 GW annual production capacity by end-2026. The Q1 2026 earnings call added a much stronger management claim: current manufacturing footprint can support 5 GW/yr of commercial product output and Bloom is not currently capacity constrained. Treat that 5 GW number as management-stated nameplate/footprint capacity, not audited realized output.

Evidence:

  • 2025 10-K: principal U.S. manufacturing/R&D/warehousing footprint is 958k sq ft excluding HQ: Fremont 326k leased, Newark 454k leased, Newark 178k owned.
  • Newark owned facility is purpose-built for fuel-cell and Energy Server assembly, designed for copy-exact duplication, with 25 additional acres for expansion or supplier co-location.
  • PP&E purchases were $116.8M (2022), $83.7M (2023), $58.9M (2024), $56.8M (2025), and $26.2M in Q1 2026. The old $200-400M/GW capex assumption was too high.
  • Q1 2026 10-Q says Bloom will keep investing at Fremont and Delmarva; timing depends on implementation milestones, supplier lead times, and customer demand.

Implication: manufacturing no longer kills the bull case by itself. I would still discount the 5 GW figure to ~3-3.5 GW/yr of achievable 2026-27 shipping capacity until quarterly delivery data proves otherwise.

Q1.1.2: How much MW of demand will materialize from AI data centers by 2029?

The thesis: data centers can't wait 5 years for grid interconnection. BE deploys 90 days. That speed-of-deploy gap is the entire AI bull case for fuel cells.

Q1.1.2.a: How big is the data-center power gap that BE specifically addresses? PARTIAL

Per knowledge from prior sector work: US data center demand will go from ~25 GW today to ~50-70 GW by 2030 (45+ GW of incremental demand). Of that incremental, a chunk needs bridge power between site-ready and grid-ready dates — typically 1-5 years per site.

BE's addressable share = (bridge MW) × (BE win rate vs alternatives).

Bridge MW estimate: if 40 GW of new AI data center capacity needs bridge power averaging 3 years, that's 40 GW × 3 = 120 GW-years of bridge demand. Spread over 2026-2030 = ~24 GW-years/yr addressable.

BE win rate vs alternatives: see Q1.1.2.b — guess 10-20%, so BE could ship 2.4-4.8 GW/yr from data centers alone by 2028-29. Caps at manufacturing capacity (Q1.1.1).

Q1.1.2.b: Why would a hyperscaler buy BE vs alternatives?

The alternatives, ranked by relevance:

AlternativeDeploy timeCost per MWEmissionsBE's edge
BE solid oxide fuel cell~90 days$3M/MW system + fuel~50% lower CO2/MWh than grid avg, no NOx/SOxbaseline
On-site gas turbine (GE/Mitsubishi)12-18 months$1-1.5M/MWHigher CO2, real NOx/SOx (local permit risk)BE faster, cleaner permitting
On-site reciprocating gas (CAT)6-12 months$0.8-1.2M/MWWorst emissions profileBE far cleaner, similar speed
Grid + nuclear PPA3-7 years interconnectvariesLowBE much faster, on-site
SMRs (Oklo, Kairos, X-energy, TerraPower)2030+ first units$5-8M/MW estLowestBE ships now, SMRs are 2030+

The honest answer: BE wins on speed + clean permitting. It loses on $/MW upfront and ongoing fuel cost. The willingness to pay BE's premium scales directly with how desperate the hyperscaler is for "right now" power.

Q1.1.2.b.i: What proves hyperscalers actually buy BE? KNOWN

Disclosed BE wins:

  • AEP (American Electric Power) — 1 GW master supply agreement (2024) for data center deployments. Largest single fuel-cell order in history.
  • Equinix — long-standing customer, multiple sites
  • Quanta Computing (Taiwan AI server maker) — disclosed in 2025
  • SK ecoplant (Korea) — large international anchor
  • CoreWeave / Anthropic / various hyperscaler-adjacent — partly disclosed

The AEP 1 GW alone is roughly 1.3x BE's 2025 shipments and is multi-year — so the demand is real. The question is whether it sustains and scales.

Q1.1.2.b.ii: How does BE's value prop change once grid catches up?

This is the critical bear scenario. If by 2028-29 the grid interconnection queue clears (PJM/ERCOT rule changes, hyperscaler-funded transmission, etc.), the "90-day deploy" edge collapses. Then BE has to compete on $/MWh delivered against grid power at $40-60/MWh — and BE delivers at maybe $100-150/MWh all-in. BE loses that comparison.

Counter: many BE installations are permanent primary power, not bridge. Especially in healthcare, retail, telecom, manufacturing. The grid-catch-up risk applies mostly to the AI bridge segment.

I need to find: What % of BE's backlog is "bridge to grid" vs "permanent primary power"? This is probably disclosed in their 10-K or investor day deck. UNKNOWN

Q1.1.2.c: What about behind-the-meter SMR competition by 2029? PARTIAL

From the VST/CEG analysis: Meta committed 4 GW to Oklo + TerraPower SMRs, Amazon committed up to 960 MW with X-energy, Google committed 500 MW to Kairos. But all of these have first units targeting 2030+.

Through 2029, SMRs are not a real competitive threat for BE. The competitive window is 2030-2032. So BE has a clear ~3-4 year runway before SMRs eat into the bridge-power use case.

This is good news for the 2029 cash-flow question specifically. But it caps the terminal multiple — investors will start discounting in 2027-28 as SMR commercialization gets clearer.

Q1.1.3: How much MW from non-data-center demand?

BE's legacy business: utilities, retail (Walmart was an old anchor), healthcare hospitals, telecom (towers), manufacturing, education. Pre-AI, this was running ~300-500 MW/yr.

This is the "floor" demand. Plausibly grows 10-15%/yr organically = 500-700 MW/yr by 2029. Should not change much under any scenario.

What I need to verify: historical MW shipped by segment. Probably in 10-K disclosures. UNKNOWN

Q1.1.4: International — particularly Korea SK?

BE-SK Group joint venture has been a meaningful chunk of international revenue. Korea has high power costs + tight emissions rules + ambitious hydrogen agenda — natural fit for SOFCs.

Risk: Doosan (Korean conglomerate) acquired Ceres Power's UK SOFC tech and is building a domestic competitor. Could displace BE from Korea over 2026-29.

Plus: Italy (Enel), Japan (potential), India (BE-Mukesh Ambani / Reliance announced 2024 hydrogen partnership).

International is probably 15-25% of 2029 revenue. Not the dominant lever but adds optionality.

Q1.2: What is $/MW for hardware? (Pricing power)

Currently ~$3M/MW system price. Question: does this rise (scarcity premium for AI bridge) or fall (cost-down curve, competition)?

Q1.2.1: Cost-down curve — what's the trajectory of BE's COGS per MW? PARTIAL

BE has guided to ~25% cost reduction per doubling of cumulative volume (typical "experience curve" for manufactured energy products — similar to solar PV which famously hit ~20-30% per doubling).

If they ship 700 MW in 2025 and 2 GW in 2029, that's ~3 doublings of cumulative volume. Implies cost-down of ~50%. Combined with current 30% gross margin, this is the main driver of margin expansion in the 2029 scenarios.

Risk: rare-earth / yttrium / scandium / nickel input cost inflation. SOFC stacks use specific ceramic materials that may not benefit from broad commodity scale.

Q1.2.2: Can BE charge a scarcity premium right now?

Probably yes, for AI bridge orders. The AEP 1 GW deal terms aren't disclosed but anecdotally BE got better pricing than legacy contracts. If demand >> supply through 2027-28, BE could maintain or slightly raise $/MW even as costs fall — that's pure margin expansion.

This is the biggest single uncertainty in the model. If pricing power holds, gross margin goes from 30% → 40%+. If it doesn't (i.e., BE has to pass through cost savings to win deals), gross margin stays at 30-35%.

Q1.3: How much is service / recurring revenue?

BE sells multi-year service contracts on every system. As installed base grows, this becomes a substantial recurring revenue stream at much higher margin than hardware.

Q1.3.1: What's the installed base trajectory? PARTIAL

Cumulative MW installed ~2-3 GW today (estimate). At 1 GW/yr shipments through 2029, installed base reaches 6-8 GW by end-2029.

Service revenue ~$50-100K/MW/yr based on typical SOFC service contracts. So 7 GW × $75K avg = ~$525M/yr of service revenue by 2029, at ~40-50% gross margin = ~$210-260M of service gross profit.

Small relative to hardware but meaningful and high-quality. Adds ~5% to total revenue and ~10% to gross profit.

Q1.4: Electrolyzer / hydrogen revenue?

BE pivoted in 2021-22 to also sell electrolyzers (running fuel cells in reverse to make H2 from electricity). Heavily IRA-subsidy-dependent.

Q1.4.1: Is electrolyzer demand real in 2026? UNKNOWN

The hydrogen economy has under-delivered vs 2021 expectations. Most green H2 projects have been delayed or cancelled. The IRA 45V tax credit rules have been contentious.

Honest take: probably $0-200M in electrolyzer revenue by 2029. Small. Could be 0. Will treat as a free option, not part of base case.

Q1.5: Roll up — what's my revenue range for 2029?
ComponentBearBaseBullLottery
Hardware MW shipped1.0 GW1.8 GW2.5 GW3.5 GW (above current capacity)
Hardware $/MW$2.8M$3.2M$3.5M$4.0M (scarcity premium)
Hardware revenue$2.8B$5.8B$8.75B$14.0B
Service revenue$400M$525M$650M$800M
Installation / other$400M$700M$1.0B$1.5B
PPA / electricity$150M$300M$500M$1.0B
Electrolyzer / H2$0$100M$300M$700M
International incremental$200M$500M$1.0B$2.0B
Total 2029 revenue$4.0B$7.9B$12.2B$20.0B
Implied CAGR from $2.45B13%34%50%69%

2025 revenue grew 130% YoY. Sustaining 50%+ CAGR for 4 more years is possible but rare. Base case at 34% CAGR is reasonable.

Q2: What is 2029 operating margin? PARTIAL

2025 operating margin: 9.6%. To hit base case 14% by 2029 requires gross margin expansion + opex leverage. To hit bull 20% requires both AND meaningful service-revenue mix shift.

Q2.1: What's the gross margin trajectory?

Current: 30.1%. Management has guided "long-term target ~35%+" for several years but actual trajectory has been bumpy.

Q2.1.1: What drives gross margin up?
  • Volume / experience curve — see Q1.2.1, ~50% cost-down through 2029 on stacks
  • Mix shift to service — service is ~45-50% margin vs hardware ~25-30% margin
  • Pricing power — Q1.2.2, scarcity premium adds 200-500 bps
  • Material cost decline — if rare earths stay stable or decline

Realistic 2029 gross margin: bear 30%, base 36%, bull 40%, lottery 44%.

Q2.2: What about opex leverage?

2025 opex (R&D + SG&A): ~$500M on $2.45B revenue = 20% of revenue. If revenue triples but opex grows only 30-50%, opex falls to ~10% of revenue. This is the second big lever for op margin expansion.

Q2.2.1: How much R&D does BE need to keep spending? UNKNOWN

BE has spent $200-300M/yr on R&D for stack development, electrolyzer development, hydrogen co-generation, etc. As products mature, R&D % of revenue typically declines.

Plausible path: R&D stays absolute-flat at $300M but falls from 12% to 4% of revenue by 2029. SG&A scales sublinearly with revenue (1.4x revenue growth → 1.0x SG&A growth).

Combined opex: $500M → $700M from 2025 to 2029. As % of revenue: 20% → 7-9%.

Q2.2.2: Stock-based compensation — how much dilution / opex hit?

BE's SBC has historically been ~$100-150M/yr. Excluded from "adjusted" margins but real cost to shareholders. At 285M shares this is ~0.5%/yr of dilution, not catastrophic.

Need to verify: SBC trajectory in recent filings. Has it grown with the stock price? UNKNOWN

Q2.3: Roll up — operating margin 2029
ComponentBearBaseBullLottery
Gross margin30%36%40%44%
Opex as % rev18%11%8%7%
SBC as % rev4%3%2%2%
GAAP op margin8%22%30%35%
Op margin (excl SBC)12%25%32%37%

This is meaningfully higher than my preliminary skeleton at the top of the document. If revenue scales and opex leverages as expected, op margin gets to 20%+ which is genuinely impressive for industrial hardware.

Q3: FCF conversion from operating income?

FCF = Operating CF − CapEx. Need to estimate maintenance + growth capex.

Q3.1: How much capex to support 2.5+ GW/yr of shipments?

Current capex run rate is lower than I originally assumed. PP&E purchases were $56.8M in 2025 and $26.2M in Q1 2026, while management says the current footprint can scale toward 5 GW/yr. This points to process/throughput/copy-exact scaling, not a solar/battery-style greenfield buildout.

Updated estimate: $75-100M/GW is a better implied incremental-capex signal than the old $200-400M/GW assumption, but it is not an audited per-GW factory budget. Use $100-150M/yr of growth capex through 2028 until Bloom proves it can run near 5 GW nameplate without a bigger PP&E step-up.

The real constraints are likely supplier qualification, specialty materials, line yield, field installation labor, and customer site readiness rather than funding the buildings.

Q3.2: Working capital — is BE building inventory / receivables fast?

2025 saw a big inventory build (data center backlog ramp). This is a real cash drag — receivables stretch on large multi-year contracts.

For 2029 estimates, assume working capital grows ~10% of incremental revenue. On $5B of revenue growth from 2025 to 2029, that's ~$500M of cumulative working capital absorption ≈ $125M/yr.

Q3.3: Net FCF conversion

Op income → +D&A − CapEx − WC build = FCF

D&A approx equals CapEx in steady state. After capacity expansion (2026-28), FCF should converge toward (Op income − WC absorption).

Realistic 2029 FCF conversion: 60-70% of op income if capacity expansion is largely done; only 40-50% if still mid-expansion.

Q4: How much equity dilution by 2029? KNOWN (mostly)

This matters a lot. BE has a $2.5B convertible debt issue maturing 2030 with conversion price ~$195. Stock is at $258 — converts are deep in the money.

Q4.1: Convertible bond mechanics

The $2.5B converts, if fully converted at $195, issue 2.5B / 195 ≈ 12.8M new shares. Existing 284M shares → ~297M shares. ~4.5% dilution.

This is built into the scenario tables already — base case assumes 305M shares (accounts for converts + ongoing SBC).

Q4.2: Will BE need to raise more equity? PARTIAL

If revenue scales as base/bull case suggests, BE is FCF positive through 2029. Cash balance grows. No equity raise needed.

If bear case (revenue stalls), BE could burn cash and need to raise. But $2.5B current cash gives substantial runway.

Bear case dilution scenario: $500M equity raise at $150/share = 3.3M additional shares. ~1% additional dilution. Not catastrophic.

Q4.3: Net share count trajectory by 2029
  • 2025: 284M shares
  • + Convert dilution: ~13M
  • + SBC dilution (~1.5%/yr × 4 yr): ~17M
  • + Possible raises (bear case only): 3-5M
  • 2029 shares: 314M (bear), 305M (base), 297M (bull/lottery if converts called early via cash redemption)
Q5: What macro / policy factors swing the answer?
Q5.1: Natural gas prices through 2029

Fuel is the customer's recurring cost, not BE's. But high gas prices make BE-delivered electricity more expensive than grid alternatives → reduces demand.

Conversely, very low gas prices make on-site gas turbines (BE's competitor) cheaper to operate, narrowing BE's TCO advantage.

The "sweet spot" is current $3-5/MMBtu gas, which makes BE competitive on a TCO basis with grid + lower-emission than turbines.

Q5.2: IRA / ITC / clean energy tax credits

BE customers benefit from the 30% Investment Tax Credit for fuel cell systems (extended under IRA through ~2032). This effectively cuts customer purchase price by 30%, making BE more competitive.

Risk: A future administration repeals or weakens ITC. This is a real overhang — BE's stock has historically tracked clean-energy policy sentiment.

Q5.3: Carbon pricing / state climate rules

BE's emissions advantage over gas turbines is meaningful (no NOx/SOx, ~50% less CO2). In CA, NY, MA, NJ where local air permits matter, BE has structural advantage.

If federal carbon price emerges (unlikely under current admin), BE's relative cost falls.

Q5.4: White House BYOG — does it help or hurt BE? PARTIAL

From the VST/CEG analysis: White House Jan 2026 mandated hyperscalers fund $15B+ of new generation themselves. BE is a primary beneficiary of "BYOG" — fuel cells are exactly the kind of fast-to-deploy on-site generation hyperscalers will build under that mandate.

This is materially positive for BE demand. Was probably part of why BE ran from $90 to $310 in late 2025 / early 2026.

Q6: Refined 2029 cash flow estimate

Pulling Q1 (revenue), Q2 (operating margin), Q3 (FCF conversion), Q4 (shares) together with the deeper analysis:

MetricBearBaseBullLottery
2029 revenue$4.0B$7.9B$12.2B$20.0B
Gross margin30%36%40%44%
Op margin (GAAP)8%22%30%35%
Op income$320M$1.74B$3.66B$7.0B
Cash tax (~20%)$64M$348M$732M$1.4B
After-tax op income$256M$1.39B$2.93B$5.6B
FCF conversion50%65%70%72%
2029 FCF$128M$904M$2.05B$4.03B
Share count (M)314305297297
2029 FCF / share$0.41$2.96$6.91$13.57
At $258/share today: bear ~600x FCF, base ~87x, bull ~37x, lottery ~19x.
At a "fair" 25x FCF on 2029 numbers: Bear $10, Base $74, Bull $173, Lottery $339.
Bear case implies ~96% downside from $258. Base case implies ~71% downside. The market is pricing bull / lottery — meaning the buyer at $258 needs revenue to compound at 50%+ for four more years AND gross margins to expand to 40% AND op leverage to deliver. Triple-stacked optimism.
Q7: What are the biggest unknowns I still need to resolve? UNKNOWN
  1. Validate the 5 GW capacity claim. Manufacturing no longer looks like a 1.5-2 GW hard ceiling, but the 5 GW number is still nameplate/footprint until quarterly shipments and backlog conversion prove it.
  2. Backlog composition. What % is bridge-to-grid vs permanent? If 80% is bridge, BE has a 3-5 year window then revenue falls.
  3. Pricing per MW on the AEP 1 GW deal. Sets the new benchmark for AI-bridge contracts.
  4. Service contract economics. Term length, escalators, customer churn.
  5. Doosan / Ceres competitive threat in Korea + global. Could be a structural margin compressor.
  6. SBC trajectory. Has it scaled with stock price? Could surprise to the upside (positive) or downside (more dilution than I modeled).
  7. Convertible bond detailed terms. Could there be early forced conversion if stock holds above $195 for X days?
  8. Real-world fuel cell stack life / replacement cycle. Drives long-term service revenue trajectory.
  9. Hyperscaler announcements specifically with BE. The Anthropic / OpenAI / xAI rumors need verification.
  10. What happens when SMRs commercialize (2030+). Caps terminal multiple.
Q8: What does this say about whether to own BE at $258?

Honest first read: the stock is pricing in the bull-to-lottery scenario. Asymmetry is bad at current prices — limited upside (maybe +30% to lottery, +0% to bull which roughly = today's price) and big downside if base case plays out (-70%).

Important caveat: BE may be in a multi-year demand boom where 2030, 2031, 2032 cash flows continue to compound. If you believe the AGI-power-shortage story extends well past 2029, the relevant multiple is on 2031-32 FCF not 2029 FCF. At 50% CAGR through 2032, base case 2032 FCF could be $2.5B+ which justifies a much higher stock price.

Where I'd want to spend more research time:

  • 5 GW nameplate vs realized shipments (Q1.1.1) — watch quarterly product revenue, delivery timing, PP&E, supplier-lead-time language
  • Backlog bridge-vs-permanent mix (Q1.1.2.b.ii) — directly drives terminal value
  • Pricing power evidence from AEP and recent contracts (Q1.2.2) — drives gross margin upside
  • BE-specific hyperscaler contract pipeline beyond AEP and Oracle — drives 2027-2029 demand visibility

Comparison to VST/CEG: BE has a much shorter monetization window (2025-2030) before SMRs arrive. It also has the cleanest behind-the-meter exposure — it is the BYOG mandate. But it's priced for perfection.

Sources: yfinance live (May 18, 2026), BE Q4 2025 earnings, AEP 1 GW agreement coverage, IRA Section 48 fuel cell ITC, prior sector knowledge from data-center-power.md and fuel-cells-onsite-power.md. Convertible bond terms from BE 10-K. Win-rate vs alternatives estimated from public bid data and industry-research notes. Many sub-questions tagged UNKNOWN need follow-up research.