Initiated 2026-10-02 · Price $289.15 (close 2026-10-02, stockanalysis.com) · Mkt cap $85.2bn basic / $93.5bn on 323.3mm diluted · Industrials / electrical equipment (solid-oxide fuel cells, on-site power) · Model: verified

Rating: Sell — Conviction: High

Bloom is in the middle of a real inflection. Q2 2026 revenue was $1.07bn (+166%), GAAP operating income $182mm, and FY2026 guidance was raised to $3.9–4.2bn, roughly double FY2025. The price discounts far more than that. At $289.15 the stock trades at ~226x LTM GAAP EBITDA and ~136x FY2026E on the model's own basis. That is about twice what the Bull case is worth, and the Bull case already assumes revenue grows ~6x in five years at 25% margins.

Conviction tests (3a-v-c): T1 pass (Base −83.4% at +1pp, −82.0% at −1pp; Bull −54.6% / −49.2%; Sell at both) · T2 pass (Gordon $23.32, exit-multiple $76.65; Sell on either) · T3 pass · T4 pass

The rule's High describes how robust the valuation call is, not when it pays. A beta-3.8 stock with momentum and a 25 GW prospect list can stay far above fundamentals for a long time.

Model value range vs price
Bear $12.08Base $49.99Bull $138.69Price $289.15

Business overview

Bloom makes solid-oxide fuel-cell "Energy Servers" that turn natural gas (or biogas/hydrogen) into electricity on site, without combustion. In Q2 2026 product revenue was $935mm (88%), service $69mm, installation $51mm and electricity (legacy PPAs) $10mm. The customers are now mainly AI data centres, hyperscalers, neoclouds and colocation operators, plus utilities and commercial and industrial sites. Two customers made up ~44% and ~21% of Q2 revenue, one of them a related party, and 90% of H1 revenue was US. Earnings are driven by three things: AI data-centre power demand that the grid can't connect fast enough, factory throughput (capacity is going from 1 GW to 2 GW by end-2027, per the Q2 call), and price versus stack cost, which sets gross margin (33.4% in Q2, ~34% non-GAAP guided).

Competition

Bloom's competition for behind-the-meter data-centre power is mostly gas generation, not other fuel cells. Cleanview's tracker of data-centre projects (mid-2026) gives Caterpillar 33% of projects (reciprocating engines and Solar Turbines), with more than 8.8 GW permitted, against Bloom's 14%. Bloom's share rose after the Oracle deal (2.8 GW, April 2026). Natural-gas turbines are ~75% of identified generation. GE Vernova's gas backlog plus slot reservations reached 116 GW at Q2 2026, and new heavy-duty turbine orders reportedly won't deliver until ~2031. That delivery gap is Bloom's speed-to-power advantage. Pressure would show up first in price, once aeroderivative and engine supply catches up and that gap narrows. FuelCell Energy is the only listed pure-play fuel-cell peer, and it is sub-scale and loss-making. The closest profitable comparables trade at 26.4x (CAT) and 65.0x (GEV) trailing EV/EBITDA, against Bloom's 204x on the same aggregator definition.

Bull case

  1. Speed to power becomes a multi-year franchise. The 25 GW prospect list, Oracle's 2.8 GW and the Brookfield programme (expanded from $5bn to $25bn) convert. Capacity goes past 2 GW, and revenue reaches ~$12.6bn by FY2030. Plays out if grid interconnection queues and turbine lead times stay long through 2029. Model: rev_growth (bull)
  2. Scale lifts margins. Fixed-cost leverage and falling SBC intensity take GAAP EBITDA margin from 18.4% (Q2 2026) to 25%. Plays out if pricing holds while the factory doubles. Model: ebitda_margin (bull)
  3. The market keeps paying for AI power. Net cash and the 0% converts leave nothing to refinance. A peer-beta discount rate (13.6%) and a 40x exit multiple hold. Plays out if AI capex is treated as a decade-long build. Model: wacc, exit_ev_ebitda (bull)

Bear case

  1. Concentration and digestion. Two customers were ~65% of Q2 revenue. If one pauses after FY2026 while hyperscalers digest capacity, growth slows to +15% and revenue then falls 10% and 5%. Plays out if 2027 AI capex guidance flattens or a large customer re-sources. Model: rev_growth (bear)
  2. Price competition on a doubled factory. Once turbine and engine supply catches up, on-site power becomes a price contest. With deleverage on 2 GW of capacity and inventory built ahead of orders, EBITDA margin falls to 9% and working capital rises to 27% of revenue. Model: rev_growth, ebitda_margin, nwc_pct_rev (bear)
  3. Multiple normalisation. A 226x LTM multiple gives way to cyclical power-equipment pricing (12x exit) at the stock's own beta of 3.78 (24.1% discount rate). Plays out in any growth scare. Model: wacc, exit_ev_ebitda (bear)

Valuation & balance sheet

Metric (definition) Current Own history (range or 5y avg) Peers Source, as-of
EV/EBITDA, model basis (diluted mkt cap − cash + debt + financing obligations + NCI, ex-leases; GAAP EBITDA after SBC) 738x FY2025; ~226x LTM; ~136x FY2026E n/m (EBITDA negative FY2021–23) — Model inputs; edgar, q2rel, 2026-06-30
EV/EBITDA, trailing (aggregator) 204.3x n/a (unverified) GEV 65.0x · CAT 26.4x · FCEL n/m (negative EBITDA) stockanalysis.com, 2026-10-02
EV/Sales, trailing (aggregator) 27.5x n/a (unverified) GEV 6.2x · CAT 5.7x · FCEL 6.1x stockanalysis.com, 2026-10-02
Forward P/E (aggregator consensus) 81.6x n/a (unverified) GEV 47.2x · CAT 28.8x · FCEL n/m stockanalysis.com, 2026-10-02

Model-implied value range (from model-summary.json; generic module, Gordon-growth DCF and exit-EV/EBITDA DCF, midpoints): Bear $12.08 · Base $49.99 · Bull $138.69 per share, i.e. implied returns of −95.8% / −82.7% / −52.0% vs $289.15. These ranges show how the bull and bear drivers translate into value; they are not price targets. The price is about twice the Bull midpoint. To justify it, the market needs either revenue well beyond ~$12.6bn by FY2030 or a terminal multiple far above GE Vernova's today. The Base 25x exit is a ~97% de-rating from today's same-basis 738x FY2025 multiple (~82% from ~136x FY2026E). The two methods differ widely (Base Gordon $23.32, exit $76.65), so the exit method carries the higher terminal multiple, and even it is 73% below the price. The tail sensitivity puts the largest customer's loss in FY2027: revenue −30% and −15%, EBITDA margin 4–6%, NWC 30%, a 10x exit and a 25.1% discount rate. It gives $9.38 per share. With the stock already above the Bull case, the tail barely changes the call.

Balance sheet: not meaningful (net cash). At 2026-06-30 Bloom held $2,666.9mm of cash against $2.6mm of non-recourse term loan (October 2026) and $206.4mm of financing obligations. The $2.5bn 0% convertible notes due November 2030 (conversion ~$194.97) and ~$28mm of 3% green converts (2028/2029) are in the money and are treated as equity: their shares are in the 323.3mm diluted count. Below ~$195 the 2030 notes would become a $2.5bn cash claim. Ratings: n/a (unverified).

Model note: Verified: LibreOffice recalculation matched all 3,207 formula cells. No unverified inputs, no assumptions without basis, no scenario consistency CHECKs. EBITDA is GAAP, after ~$56mm a quarter of SBC. Base Y1 EPS of $2.01 is GAAP and sits well below the $2.55–2.85 non-GAAP guide, which adds back SBC (~$0.70 per share). The WACC of 18.9% equals its CAPM arithmetic. Beta 2.73 blends the stock's own 3.78 with the peer mean of 1.67, a deliberate haircut to the own beta. Convertible terms come from a search summary of the November 2025 8-K, not a direct fetch. FY2021–25 NWC includes current financing obligations, which can't be separated in EDGAR.

Scenario stress test

Reasoned from the bull/bear drivers above. The model column comes from the scenario overlays (Base case + shock).

Scenario Effect Mechanism Magnitude Model Δ value vs Base ($/sh)
S1 Fast equity crash − Beta-3.8 momentum name de-rates in a liquidity flush; orders unaffected Med −$9.19
S2 Slow bear / recession − Data-centre and C&I projects deferred; deleverage on a doubling factory. Transitory, so milder than the permanent Base − Bear gap ($37.91) High −$15.44
S3 Rapid rate shock − Long-duration equity; customers' and Brookfield's project-finance costs rise Med −$9.87
S4 Slow rate grind − Same channel, grinding Low −$4.04
S5 Soft-landing cuts + Lower discount rate, cheaper project finance, demand intact Med +$6.00
S6 Recession-driven cuts − Capex cuts outweigh the lower rate Med −$5.03
S7a Credit liquidity shock − Net cash, so no funding need; crowded high-beta name sold, deployment financing pauses Med −$5.51
S7b Slow default cycle − Debt-funded neoclouds and developers lose financing; orders slip Med −$6.17
S8 Stagflation − Materials and labour inflation on fixed-price orders; higher real rates Low −$2.62
S9a Dollar spike − ~10% non-US revenue; small translation drag Low −$0.20
S9b Dollar slide + Mirror of S9a Low +$0.20
S10 Melt-up + High-beta AI-power leader; momentum extends the multiple Med +$7.35
S11 Energy supply shock ± Energy-security demand rises, but higher gas prices hurt customer economics and input costs Low +$0.26
S12 Mega-cap/AI derating − The direct channel: hyperscaler capex cuts hit the core end market and the AI multiple High −$14.49

Currently active/on watch per the playbook: S3 partially active (state.md; the 2026-10-01 macro log has both pace legs met at 10Y 5.24%); S8, S10, S11 on watch.

Model value change vs Base, by scenario
S2 Slow bear / recession−$15.44S12 Mega-cap/AI derating−$14.49S3 Rapid rate shock−$9.87S1 Fast equity crash−$9.19S7b Slow default cycle−$6.17S7a Credit liquidity shock−$5.51S6 Recession-driven cuts−$5.03S4 Slow rate grind−$4.04S8 Stagflation−$2.62S9a Dollar spike−$0.20S9b Dollar slide+$0.20S11 Energy supply shock+$0.26S5 Soft-landing cuts+$6.00S10 Melt-up+$7.35

What would change the call

Upgrades if: the price falls toward the Base–Bull range without a change in fundamentals; or FY2027 guidance (likely February 2027) implies revenue of $8bn or more with GAAP EBITDA margin above 22%, backed by disclosed multi-GW orders from more than two customers. Downgrades if: n/a (already Sell). The call strengthens if Q3/Q4 customer concentration rises further, deferred revenue falls while sales grow, or gross margin slips below 30% as capacity ramps.

Watch items

Sources