| 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…
3-statement linked UFCF model. Stock $287.97 (Apr 29, 2026 close), mkt cap $81.9B post Q1 2026 beat. Companion to [[be]] (profile) and [[be-mgmt-dd]] (governance forensics).
| 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 stock price requires either bull-case revenue scaling AND bull-case margins AND a sub-9% discount rate. Each of those is independently defensible. Together, they require essentially flawless execution against the AI data-center power thesis for a decade. The setup is asymmetric: small misses on any of the three legs compress the multiple meaningfully; large beats are needed to expand it further from here.
This is the textbook valuation problem for a momentum stock. DCF math doesn’t drive the price; the AI infrastructure scarcity narrative does. The model is useful for quantifying what scenarios are baked in, not for signal-generation on entry/exit timing.
BE’s FY2024 revenue ($1,474M) breaks into four segments. FY2025 $2,024M segment split not yet broken out at line-item level in the headline release; using FY2024 mix as the base.
| Segment | FY2024 ($M) | % | FY26E mid | FY30E | FY35E | Notes |
|---|---|---|---|---|---|---|
| Product | 1,090 | 74% | 2,500 | 8,400 | 16,500 | Energy Server hardware. Volume scales linearly with capacity expansion. Oracle 2.8 GW + AEP 1 GW = primary drivers. |
| Service | ~210 | 14% | 350 | 1,200 | 3,200 | 15-year fixed-fee O&M contracts attached to each Product sale. Recurring, growing tail. Margin recovery from negative to ~30% over the model period. |
| Installation | ~150 | 10% | 600 | 1,800 | 2,500 | Site engineering. Grew 31.8% YoY in FY24. Lower-margin pass-through. |
| Electricity | ~25 | 2% | 150 | 600 | 800 | Managed-services / PPA. Long-dated. |
| Total | 1,474 | 100% | 3,600 | 12,000 | 23,000 | Mgmt FY26 guide $3.4-$3.8B post Q1 2026 raise |
Key revenue assumption: Product revenue compounds at ~28% CAGR through 2030 driven by Oracle/AEP/Brookfield contract drawdowns, then decelerates to mid-teens through 2035 as the AI buildout cycle matures. Service revenue grows as the installed base compounds (each new Product sale adds 15+ years of Service tail). Installation tracks Product. Electricity remains the smallest line.
FY2026 explicit: $3.6B (mgmt guide midpoint of raised range $3.4-$3.8B) FY2027 explicit: $5.24B (current consensus average) FY2028 explicit: $7.2B (+38%) FY2029 explicit: $9.5B (+32%) FY2030 explicit: $12.0B (+26%) FY2031-2035: declining growth from 21% to 7% as cycle matures
| Year | Revenue | Op Margin | Op Income | Notes |
|---|---|---|---|---|
| FY2022 (actual) | 1,201 | -22% | -261 | Pre-inflection era |
| FY2023 (actual) | 1,335 | -16% | -209 | Approaching breakeven |
| FY2024 (actual) | 1,474 | +1.6% | +23 | The inflection |
| FY2025 (actual) | 2,024 | +3.6% | +73 | Expansion |
| FY2026E | 3,600 | 19% | 685 | Mgmt guide $600-750M = ~17.6%-19.7% midpoint |
| FY2027E | 5,237 | 22% | 1,152 | |
| FY2028E | 7,200 | 24% | 1,728 | |
| FY2029E | 9,500 | 25% | 2,375 | Terminal margin reached |
| FY2030E | 12,000 | 25% | 3,000 | |
| FY2031E | 14,500 | 25% | 3,625 | |
| FY2032E | 17,000 | 25% | 4,250 | |
| FY2033E | 19,500 | 25% | 4,875 | |
| FY2034E | 21,500 | 25% | 5,375 | |
| FY2035E | 23,000 | 25% | 5,750 |
Margin justification: Terminal 25% op margin sits between hardware peers (Cummins ~13-15%, GE Vernova ~10-12%) and software/IP-heavy industrial peers (Watts Water ~18%, Roper ~30%). Bloom’s mix justifies the premium because: (i) Service is high-incremental-margin once installed-base scales; (ii) Product is increasingly software-rich (digital twin, predictive maintenance); (iii) hyperscaler customer concentration enables fixed-price contracts at favorable terms during the AI build cycle. The 25% terminal is closer to a base-bear blend; 30% would be the bull case.
Risk to terminal margin: AI data center demand intensity declines after 2030 → Service margin holds but Product mix shifts toward more competitive utility procurements at lower margin. Pricing pressure from Korean/Chinese SOFC entrants by 2030. Hydrogen transition adds capex without immediate revenue.
This is the methodologically interesting part. Two cost-of-equity frameworks give very different numbers.
| Component | Value |
|---|---|
| Risk-free rate (10Y UST) | 4.3% |
| Equity risk premium | 5.0% |
| Industry-adjusted beta | 1.5 (Yahoo raw beta is 3.19; the high beta reflects momentum-volatility, not fundamental risk) |
| Cost of equity | 11.8% |
| Pre-tax cost of debt | 3.0% (BE’s recent 0% convert + 3% green convert) |
| After-tax cost of debt | 2.4% |
| Capital structure (market values) | ~95% equity / 5% debt |
| WACC (CAPM) | 11.3% |
TECC = 1 / forward P/E.
At $287.97 and forward EPS $4.02, forward P/E = 71.6x. TECC ≈ 1.4%.
The market is currently pricing BE at a cost of equity of ~1.4%, which is structurally indefensible long-term but reflects the AI infrastructure scarcity premium. The gap between CAPM-WACC (11.3%) and TECC (1.4%) is the 9.9-point momentum premium that the market is awarding to “pick-and-shovel AI” stories right now.
WACC = 10% (round-number sensible mid-point that gives roughly 5-year-Treasury-plus-equity-premium math while not crediting the full TECC compression). Sensitivity table runs WACC 8% / 10% / 12% to show the bull/base/bear envelope.
| Item | FY26E | FY30E | FY35E (terminal) |
|---|---|---|---|
| D&A as % of revenue | 2.5% | 4.0% | 4.0% |
| Capex as % of revenue | 4.0% | 5.0% | 5.0% |
| Working capital as % of ΔRevenue | 10% | 10% | 10% |
| Net reinvestment rate (capex+WC-D&A) / NOPAT | ~30% | ~25% | 16% (terminal) |
Capex justification: Bloom is currently capital-light because Fremont and Newark capacity additions are largely complete. Capex/revenue runs ~3% in 2025. Once capacity hits 2 GW (end 2026), growth requires another factory or expansion. Long-run capex/revenue 5% is reasonable for a hardware-plus-service business at scale.
Working capital: Bloom historically runs material WC drag during ramp (inventory + receivables). 10% of ΔRevenue is consistent with industrials at this growth rate.
Use the implied-reinvestment formulation: TV = NOPAT_t+1 × (1 - g/ROIC) / (WACC - g)
| Input | Value |
|---|---|
| Terminal NOPAT (FY36, 4% growth from FY35) | $4,720M |
| Terminal ROIC | 25% |
| Terminal growth (g) | 4% |
| Reinvestment rate (g/ROIC) | 16% |
| Terminal FCF (NOPAT × (1 - 16%)) | $3,965M |
| WACC - g | 6% |
| Terminal value (end of year 10) | ~$66B |
The 4% terminal growth is on the high side of conventional modelers’ bands (which typically cap at 2.5-3.5%) but reasonable for a structurally-tailwinded business. AI compute demand will plausibly grow faster than nominal GDP for the next 15-20 years; even after AI growth normalizes, the on-site power TAM is structurally tied to electrification trends.
| Year | Revenue | Op Inc | NOPAT | D&A | Capex | ΔWC | UFCF | Discount factor | PV |
|---|---|---|---|---|---|---|---|---|---|
| FY26 | 3,600 | 685 | 541 | 90 | 144 | 158 | 329 | 0.909 | 299 |
| FY27 | 5,237 | 1,152 | 910 | 157 | 236 | 164 | 667 | 0.826 | 551 |
| FY28 | 7,200 | 1,728 | 1,365 | 252 | 360 | 196 | 1,061 | 0.751 | 797 |
| FY29 | 9,500 | 2,375 | 1,876 | 380 | 475 | 230 | 1,551 | 0.683 | 1,059 |
| FY30 | 12,000 | 3,000 | 2,370 | 480 | 600 | 250 | 2,000 | 0.621 | 1,242 |
| FY31 | 14,500 | 3,625 | 2,864 | 580 | 725 | 250 | 2,469 | 0.564 | 1,394 |
| FY32 | 17,000 | 4,250 | 3,358 | 680 | 850 | 250 | 2,938 | 0.513 | 1,508 |
| FY33 | 19,500 | 4,875 | 3,851 | 780 | 975 | 250 | 3,406 | 0.467 | 1,589 |
| FY34 | 21,500 | 5,375 | 4,246 | 860 | 1,075 | 200 | 3,831 | 0.424 | 1,624 |
| FY35 | 23,000 | 5,750 | 4,543 | 920 | 1,150 | 150 | 4,163 | 0.386 | 1,605 |
| PV of explicit | 11,667 | ||||||||
| Terminal value | 0.386 | 25,476 | |||||||
| Enterprise value | 37,143 | ||||||||
| Less: net debt | (164) | ||||||||
| Plus: cash optionality (Oracle warrant, partly offset by dilution) | (300) | ||||||||
| Equity value | 36,679 | ||||||||
| Diluted shares (post-Oracle warrant exercise) | 290M | ||||||||
| Implied fair value per share | ~$126 |
(All $ in millions except per-share)
Adjustments to think about: - Oracle warrant intrinsic value of ~$620M at current price is a transfer to Oracle; net dilution to existing holders embedded - Convertible note overhang (~$2.55B principal) needs to convert at strike ~$195 — already in the money; conversion adds ~12.5M shares = 4.4% additional dilution beyond the 290M base - NOLs from accumulated losses give effective tax shield in early years (model uses 21% throughout for conservatism)
Range with adjustments: $125-$140 base-case fair value per share.
At $287.97 × 290M diluted = $83.5B equity, $83.7B EV.
Backing out the implicit terminal value: - PV of explicit (10 years) at base assumptions = $11.7B - Implied PV of terminal = $83.7B - $11.7B = $72B - Implied terminal value (year 10) = $72B / 0.386 = $186B - At 6% (WACC - g) and 84% FCF/NOPAT conversion, implied terminal NOPAT = $186B × 0.06 / 0.84 = $13.3B - Backing into FY35 NOPAT = $13.3B / 1.04 = $12.8B - Implied FY35 op income = $12.8B / (1 - 21%) = $16.2B
To justify $288, the model needs FY35 op income of $16.2B — roughly 3x the base-case $5.75B.
That requires one of: 1. FY35 revenue of $65B at 25% margin (vs. base case $23B) 2. FY35 revenue of $46B at 35% margin (impossible margin) 3. A combination plus lower discount rate (8%)
The first option requires BE to be doing $65B in revenue by 2035 — that is roughly today’s Cummins revenue scale. It implies that on-site fuel cells become a 25% market share solution to ~30% of the AI data center power buildout. That’s not a fantasy, but it requires the Oracle/AEP/Brookfield triangle to be replicated 10x over by other hyperscalers.
The third option (lower WACC) is what the market is actually doing — accepting TECC of 1.4% as the implicit cost of capital. That works as long as the AI scarcity narrative holds. When it cracks, the multiple compresses by half in weeks.
Sensitivity of fair value per share ($) to terminal op margin and WACC (base case 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 |
(Values rounded; back-of-envelope reproducibility ±5%)
Read: Current price $288 is consistent with terminal margin of 28-32% and WACC 8-9%. Both ends of that band are bullish but not crazy. The price is defensible if you believe the AI data center power TAM is large enough to sustain ~28%+ operating margins for a decade and that the appropriate discount rate is closer to a high-quality industrial than a typical small-cap growth name.
Sensitivity to revenue path (terminal margin 25%, WACC 10%):
| FY30 Revenue / Terminal growth | 3% | 4% (base) | 5% |
|---|---|---|---|
| $10B (slower ramp) | $98 | $115 | $138 |
| $12B (base) | $115 | $135 | $162 |
| $15B (faster ramp) | $138 | $162 | $195 |
| $18B (bull) | $158 | $185 | $222 |
Probability-weighted blend across three scenarios:
| Scenario | Probability | Fair value | Probability-weighted |
|---|---|---|---|
| Bull (Oracle deal replicates 5x; FY30 rev $18B; terminal margin 30%; WACC 8%) | 25% | $295 | $74 |
| Base (current trajectory holds; FY30 rev $12B; terminal margin 25%; WACC 10%) | 50% | $136 | $68 |
| Bear (one major hyperscaler walks; backlog conversion slips 18 months; terminal margin 18%; WACC 11%) | 25% | $72 | $18 |
| Probability-weighted fair value | $160 |
Probability-weighted base ~$160 vs. current $288 = ~45% downside to weighted-fair. Even the 25%-weighted bull case ($295) only roughly matches current price. The asymmetry is unfavorable from here.
What would change this picture: - Bull weight increases if a second hyperscaler signs a 1+ GW frame in next 6 months - Bear weight increases if AEP draws against the 1 GW frame slip beyond initial 100 MW, or if the Korean/Chinese SOFC competition materializes faster than expected
The DCF cannot justify the current price under base-case assumptions. The price requires bull-case revenue, bull-case margins, and bull-case discount rate simultaneously. Probability-weighted fair value sits ~45% below the current quote.
That does not automatically mean “sell.” Momentum stocks can sustain price-to-DCF gaps for years if the narrative compounds. The reverse-DCF tells you what the market is buying: a 10-year story where Bloom becomes a $50-65B revenue industrial supplier of choice for AI data center power, and where the market continues to accept TECC of 1.4% during that buildout.
Right way to use this DCF: 1. Scenario weighting is the actual decision. If your subjective probability that BE replicates the Oracle deal 3-5 times is >40%, the price makes sense. Below 25%, the DCF math says exit. 2. Watch for the catalysts that move the bull weight — second hyperscaler frame, accelerated AEP draw, Brookfield project pipeline announcements. 3. The implied multiple compression on a missed catalyst is severe. A single quarter where backlog growth flatlines or AEP draw slips would compress the multiple from 75x forward EPS to 40-50x — call it a 30-40% correction.
Position-sizing implication: This is a stock for a small-to-medium position size with a defined exit discipline, not a core long-term hold at this multiple. The asymmetric payoff requires you to manage the position actively.
| Item | Amount | Notes |
|---|---|---|
| Cash & equivalents (FY25) | $2,454M | Funded by Nov 2025 $2.5B convert (upsized) |
| Short-term debt | $378M | |
| Long-term debt | $2,614M | Includes 2029 and 2030 converts |
| Total debt | $2,992M | |
| Net debt | $164M | Cash buffer covers gross debt almost completely |
| Convertible overhang (strike-weighted) | ~$2.55B principal | 2029 notes (3% coupon, $20.85 strike) + 2030 notes (0% coupon, $194.97 strike). Both ITM at $288 |
| Oracle warrant | 3.53M shares at $113.28 strike | Expires Oct 9, 2026; ~$620M intrinsic value at current price |
| Implied conversion dilution | ~12.5M shares (~4.4%) | Beyond the 280M FY25 base |