Bloom Energy’s revenue rose 138% and its outlook went up for the third time this year, but the quarter used $181m of cash, and the binding customer orders in its filings total $0.72bn against about $5.8bn it has now committed to its own suppliers. On consensus to 2028 and a gradual slowdown after it, the shares are worth $108.01 to $164.34, most likely $130.24. At $206.20 the price needs revenue about 62% above that path from 2028, more than our bull case delivers, at the same margins. We rate the shares overvalued.
A record quarter that used cash. Revenue was $1,235.8m, up 138%, and non-GAAP operating margin reached 23.1%; the 2026 revenue outlook rose to $4.4–4.6bn, from $3.1–3.3bn in February. Operating cash flow was nevertheless a $27.1m outflow and free cash flow a $180.6m outflow, because contract assets — revenue recognised on delivered systems but not yet billed — rose $430.3m in the quarter to $858.6m. The offset is real: customers had paid $661.4m in advance, against $78.2m in December, and nine-month free cash flow was positive at $41.6m.
What is binding: $0.72bn from customers, $5.8bn to suppliers. Remaining performance obligations in the Form 10-Q are $720.7m, up from $494.1m in June. The backlog Bloom defines in its 10-K, about $6bn of product and $14bn of service, has not been updated since February. In the same quarter its long-term purchase commitments went from $16.4m to about $5.8bn, $2.6bn of it due in 2027 and some of it take-or-pay. That is the strongest evidence in the filing that management expects much larger volumes, and it stays fixed if demand pauses.
The price needs more than the bull case’s revenue. Our base case takes consensus to 2028 ($6.8bn in 2027, $9.9bn in 2028), lets growth fade from 32% to 4% by 2036 and charges stock pay as a cost: $130.24 a share. At $206.20 the market is paying for $26.4bn of revenue in 2030, against $16.3bn in ours and $20.9bn in our bull case, at base-case margins. The price is 25% above the top of our $108.01–$164.34 band.
At $206.20 Bloom’s market value is about $64.9bn on 314.6m shares, counting vested options, the small 2029 notes and the 12.8m shares on the $2.5bn of 0% notes, which are in the money above $194.97, and its enterprise value about $62.5bn. That is 13.9 times this year’s revenue at the guidance midpoint, 9.2 times next year’s consensus, and 63 times this year’s non-GAAP earnings guidance. A business growing at 120% a year can earn multiples like these; the question is how long the growth lasts and what it costs in cash. Our base case answers both explicitly: revenue reaches the consensus of $6.8bn in 2027 and $9.9bn in 2028, growth then fades from 32% to 4% by 2036, the non-GAAP operating margin peaks at 25.5% and settles at 23%, and stock pay is charged as a cost. That path is worth $130.24 a share, 37% below the price.
Run the model in Section 02 backwards and the price needs one of three things, each on its own with everything else held at the base case. Revenue about 62% higher than our path in every year from 2028: $16.0bn in 2028 against a $9.9bn consensus, $26.4bn in 2030 and $46.0bn in 2036, above our bull case’s $20.9bn and $38.9bn. Or a non-GAAP operating margin 11 points higher every year, 34% in the long run, against the 23–24% Vertiv guides to. Or a discount rate of 7.6% instead of 10.0%, for a stock whose thirty-day volatility is about 89%. Put plainly, the market is paying for a path at or beyond our bull case, at base-case margins. The price sits 25% above the top of our $108.01–$164.34 band.
The signature measure for Bloom is the gap between what customers have bound themselves to and what Bloom has bound itself to. The first bar is the customer side: remaining performance obligations, the part of the order book in the financial statements, inside the backlog the company defines; the second is Bloom’s own long-term commitments to its suppliers; the third is the revenue the consensus expects. All three are drawn to one scale.
Remaining performance obligations from the Form 10-Q, Note 3: $661.2m of product and installation expected within one to two years, $56.2m of deferred service and $3.3m of site restoration. Backlog as defined in the 2025 Form 10-K and quantified in the February 2026 supplemental; the company has not published a figure since. Purchase commitments from the Form 10-Q, agreements longer than twelve months; they were $16.4m at 30 June and nil in December. The customer side that can be checked is about an eighth of the supplier side; the backlog that would close the gap is the company’s own figure, eight months old. On our arithmetic, if materials are about three-quarters of product cost, the 2027 commitments alone support roughly $5.5–6bn of product revenue (derived).
The fourth-quarter results in February bring the backlog figure the chief executive promised and the first 2027 outlook. A 2027 outlook at or above the $6.8bn consensus, with a backlog that reconciles to the reported obligations, would confirm the base case; it would not by itself reach the price. Contract assets turning into cash in the fourth quarter is the second test: at $858.6m they are 69% of a quarter’s revenue, and two or three customers decide when they are paid.
The rate backdrop is not friendly to a long-duration growth stock. The 10-year Treasury yield was 5.00% on 23 October, having traded between 4.96% and 5.15% over the previous week. We use a 10.0% discount rate in every case: Bloom has almost no interest-bearing debt, and the specific risks of its business are carried in the scenarios and their weights rather than added to the rate. The band runs the rate from 9% to 11%.
2026–36 revenue CAGR · 2036 operating margin after share-based pay · 2036 free cash flow · terminal value is of enterprise value · shares, the 0% notes .
This is the model that produced every figure in this report, reduced to its nine moving parts. It is an unlevered discounted cash flow on a ten-year path, 2027 to 2036, with the fourth quarter of 2026 as a stub. Free cash flow to the firm is non-GAAP operating income less stock pay, after cash tax, plus depreciation, less capital spending, less the increase in working capital, which is carried as a stock at a share of revenue rather than as a share of growth. The valuation date is 26 October 2026; each year is discounted at its mid-point, and the Gordon terminal value at 3% on 2037 cash flow is discounted from the end of 2036. On base settings the terminal value is 60% of enterprise value. The scenario cards load whole published paths, and the sliders bend the loaded path rather than replace it, so each card reproduces its published value exactly.
Why these inputs. 2027 revenue of $6.8bn (assumed). The consensus of 27 analysts before the results, range $5.3–8.3bn; there is no company outlook for 2027 yet. 2028 of $9.9bn (assumed). The consensus of 25 analysts. 2030 of $16.3bn and 2036 of $28.5bn (assumed). Growth of 32% in 2029 and 25% in 2030, fading to 4% by 2036; between the settings the path keeps the shape of the loaded case. Long-run non-GAAP operating margin of 23% (assumed). 24.5% in 2027 and a 25.5% peak in 2028–29, giving back 2.5 points by 2034 as turbine lead times shorten; the 2026 guidance midpoint is 22% and the third quarter was 23.1%. Stock pay of 2.3% of revenue (assumed). From 3.8% in 2027; it was 4.9% in the third quarter, and it is charged as a cost. Working capital of 15% of revenue (assumed). From 18% in 2027; it was $919m at 30 September, about 20% of 2026 revenue. Discount rate 10.0% and terminal growth 3.0% (assumed), the same in every case.
What the controls hold fixed. The cash tax rate rises from 5% of operating profit in 2027 to 23% by 2032: $2.3bn of federal net operating losses at December 2025 shelter the early years, and the 15% minimum tax on book income applies once average book income passes $1bn. Capital spending falls from 7.4% of revenue to 3%, and depreciation rises from 1.3% to 2.5%. The $306.5m customer consideration asset, the Oracle warrants and shares, is amortised against reported revenue but was paid in shares, so cash collected exceeds reported revenue; it is added back, after tax, over 2027–28. The fair-value band in Section 03 is this model run across the discount rate and terminal growth, and every cell of that grid is reproducible from the two controls.
Derivations, base case. Fourth quarter of 2026: revenue $1,447.7m, the $4.5bn guidance midpoint less $3,052.3m in the first nine months; non-GAAP operating income $340.0m on the same basis; stock pay $62m; capital spending $120m; working capital held at the 30 September level; free cash flow $167m, discounted 0.09 years. Shares 314.55m: 295.66m on the cover of the Form 10-Q (19 October 2026), 4.98m vested options exercised for $76.6m of cash, and 1.09m shares on the $22.8m of 3% Green notes due 2029, which convert at $20.84. The $2.5bn of 0% notes due November 2030 convert at $194.97 (5.1290 shares per $1,000, 12.82m shares, no capped call); they are counted as shares when the value solved for is above the conversion price and as debt at face when it is below. Unvested awards (9.32m) are excluded because stock pay is charged as a cost. Net cash before the notes $2,358.8m: $2,486.7m of cash, $76.6m of option cash and $13.7m of Brookfield JV stakes net of unfunded commitments, less $185.0m of financing obligations, a $3.0m Korean JV loan and $30.2m of non-controlling interest.
The number above is not a fact about Bloom Energy. It is what these assumptions produce, and the one that decides it is not in any filing: how long revenue keeps growing after 2028. The rating holds in every cell of the sensitivity grid; it changes to fairly valued at a price of $164.34 or below, and the price is almost exactly our bull-case value of $205.84. Move the 2030 and 2036 revenue settings until they are yours, and use the number that falls out.
Each card loads its published case into the model above and prints its value: $26.42, $130.24, $205.84 and $293.22. All four use the same 10.0% discount rate and 3.0% terminal growth, so they differ only in how much Bloom sells, at what margin and with how much cash tied up; demand, margin and supply-chain risk sit in the cases and the weights, not in the rate. Weighting the four at 25 / 45 / 25 / 5 gives $131.33.
| Input | Bear (25%) | Base (45%) | Bull (25%) | Super-bull (5%) |
|---|---|---|---|---|
| 2027 revenue | $5.5bn | $6.8bn | $7.8bn | $8.3bn |
| 2028 revenue | $5.2bn | $9.9bn | $12.0bn | $14.0bn |
| Growth, 2029 / 2030 | 10% / 10% | 32% / 25% | 36% / 28% | 40% / 32% |
| 2030 revenue | $6.3bn | $16.3bn | $20.9bn | $25.9bn |
| 2036 revenue | $8.9bn | $28.5bn | $38.9bn | $52.5bn |
| Non-GAAP operating margin | 20% → 15% | 24.5% → 25.5% → 23% | 25.5% → 27.5% → 26.5% | 26% → 29% → 28.5% |
| Stock pay, % of revenue | 4.0% → 2.5% | 3.8% → 2.3% | 3.6% → 2.0% | 3.5% → 1.9% |
| Working capital, % of revenue | 23–24% → 18% | 18% → 15% | 17% → 14% | 16% → 13% |
| Capital spending, % of revenue | 6.5% → 3% | 7.4% → 3% | 7.5% → 3% | 7.6% → 3% |
| Common to all | Discount rate 10.0% and terminal growth 3.0%; cash tax 5% of operating profit in 2027 rising to 23% from 2032; depreciation 1.3% → 2.5% of revenue; valued at 26 October 2026 from the 30 September 2026 balance sheet; stock pay charged as a cost | |||
| Value per share | $26.42 | $130.24 | $205.84 | $293.22 |
2027 and 2028 base revenue are the consensus before the results (27 and 25 analysts). Third-quarter stock pay was 4.9% of revenue; working capital was about 20% of 2026 revenue at 30 September.
| $ millions, base case | Q4 2026 | 2027 | 2028 | 2029 | 2030 | 2032 | 2034 | 2036 |
|---|---|---|---|---|---|---|---|---|
| Revenue | 1,448 | 6,800 | 9,900 | 13,068 | 16,335 | 22,160 | 26,082 | 28,482 |
| Growth | — | 51% | 46% | 32% | 25% | 14% | 7% | 4% |
| Non-GAAP operating income | 340 | 1,666 | 2,524 | 3,332 | 4,084 | 5,318 | 5,999 | 6,551 |
| Less stock pay | (62) | (258) | (317) | (366) | (408) | (510) | (600) | (655) |
| Less cash tax | (6) | (70) | (265) | (475) | (698) | (1,106) | (1,242) | (1,356) |
| Plus depreciation | 14 | 88 | 158 | 261 | 359 | 532 | 652 | 712 |
| Less capital spending | (120) | (503) | (554) | (588) | (653) | (709) | (782) | (854) |
| Less increase in working capital | 0 | (305) | (360) | (376) | (490) | (408) | (256) | (164) |
| Plus customer consideration asset | 0 | 160 | 121 | 0 | 0 | 0 | 0 | 0 |
| Free cash flow to the firm | 167 | 778 | 1,308 | 1,789 | 2,193 | 3,117 | 3,771 | 4,233 |
Sonde estimates. The fourth quarter of 2026 is the stub to 31 December; the customer consideration asset is added back after tax.
| Base case | $m | Note |
|---|---|---|
| Present value of cash flows, Q4 2026–2036 | 15,611 | Mid-year convention, the fourth quarter as a stub |
| Present value of terminal value | 23,826 | 60% of enterprise value; 2037 free cash flow $4,401m at 10.0% less 3.0% |
| Enterprise value | 39,437 | |
| + Cash, option exercise cash and JV stakes | 2,577 | $2,486.7m + $76.6m + $13.7m |
| − 0% notes due 2030, at face | (2,500) | The value is below the $194.97 conversion price |
| − Financing obligations, other debt and minority interests | (218) | $185.0m + $3.0m + $30.2m |
| Equity value | 39,296 | |
| Shares, millions | 301.73 | Outstanding, vested options and the 2029 notes as converted |
| Value per share | $130.24 |
The grid holds the base case fixed and varies the discount rate against terminal growth. Its corners are the published fair-value band, $108.01–$164.34: a WACC of 9–11% by terminal growth of 2.5–3.5%.
| WACC ↓ / terminal growth → | 2.5% | 3.0% | 3.5% |
|---|---|---|---|
| 9.0% | $147.34 | $155.13 | $164.34 |
| 9.5% | $135.33 | $141.70 | $149.14 |
| 10.0% base | $124.96 | $130.24 | $136.32 |
| 10.5% | $115.93 | $120.34 | $125.38 |
| 11.0% | $108.01 | $111.72 | $115.93 |
The $206.20 price lies above every cell, 25% over the highest: overvalued. The rating would change to fairly valued at $164.34 or below. Run across the same grid, the probability-weighted value spans $108.82–$165.51. The second table holds the rates at the base case and moves one operating input at a time:
| Change to the base case, all else held | Value | Change |
|---|---|---|
| Revenue 10% higher from 2028 | $142.63 | +10% |
| Revenue 10% lower from 2028 | $117.84 | -10% |
| Non-GAAP margin 2 points higher every year | $144.21 | +11% |
| Non-GAAP margin 2 points lower every year | $116.26 | -11% |
| Working capital 3 points of revenue higher | $127.75 | -2% |
| Working capital 3 points of revenue lower | $132.72 | +2% |
| Cash tax at 23% from 2027, no NOL shelter | $127.57 | -2% |
| Stock pay treated as free | $147.03 | +13% |
Modal base case $130.24, 37% below the price. Fair-value band $108.01–$164.34, the WACC 9–11% by terminal growth 2.5–3.5% grid, which anchors the rating; the price is above it: overvalued. Probability-weighted value $131.33 across the four cases at our own weights, 36% below the price and close to the base case. There is no material skew to name: the bear sits $103.82 below the base and the bull $75.60 above it, and the super-bull’s small weight adds a long tail.
| Single input changed, all else at base | Price implies | Our base |
|---|---|---|
| Revenue from 2028, every year | 62% higher | Consensus, then a fade |
| 2030 revenue on that path | $26.4bn | $16.3bn |
| Non-GAAP operating margin, every year | +11 points; 34% long-run | 23% long-run |
| Discount rate | 7.6% | 10.0% |
The revenue reading is the most natural way to see the price: the market is paying for a path at or above our bull case, at base-case margins. A margin above 30% would be well beyond Vertiv’s guided 23–24% or GE Vernova’s 12–14% EBITDA margin, and a discount rate below 8% would be unusual for a stock with 89% volatility.
| Scenario | Value | Against price | Weight | Contribution |
|---|---|---|---|---|
| Bear | $26.42 | -87% | 25% | $6.60 |
| Base | $130.24 | -37% | 45% | $58.61 |
| Bull | $205.84 | -0% | 25% | $51.46 |
| Super-bull | $293.22 | +42% | 5% | $14.66 |
| Probability-weighted | $131.33 | -36% | 100% | $131.33 |
We weight the bear at 25% because Bloom sells into data-centre budgets that its own 10-Q describes as harder to finance, two or three customers decide each quarter, and $5.8bn of supplier commitments would stay fixed in a pause. We weight it no higher because customers have paid $661.4m in advance and the outlook has risen three times this year. Weights are our judgement, not a model output.
| Risk | Priced in | Not also in |
|---|---|---|
| AI spending pauses | The bear revenue path (2028 down about 5%) and its 25% weight | The discount rate |
| Supplier commitments | Bear working capital at 23–24% of revenue in 2027–28, the excess inventory a pause would leave | A liability deducted in the base case |
| The time-to-power premium fades | Base margin gives back 2.5 points after 2029; the bear falls to 15% | The terminal growth rate |
| Scandium and supply chain | Bear volumes and the bear weight | The discount rate |
| Fleet degradation and guarantees | Service economics inside the base margin; deeper in the bear margin | A separate reserve deduction |
| Customer concentration | The scenario weights | The discount rate |
| Dilution | Stock pay charged as a cost; the notes converted at the value solved for | The share count (unvested awards are not added) |
| Securities litigation | Not priced: no estimate is possible; disclosed in Section 07 | — |
| $ millions | Value | Note |
|---|---|---|
| Market value | 64,860 | 314.55m shares × $206.20: 295.66m outstanding (10-Q cover, 19 Oct), 4.98m vested options, 1.09m shares on the 2029 notes |
| + 0% convertible notes due 2030, at face | 0 | The price is above the $194.97 conversion price; counted as 12.82m shares |
| + Financing obligations and other debt | 188 | Sale-leaseback financing obligations $185.0m; Korean JV loan $3.0m |
| + Non-controlling interest | 30 | |
| − Cash | (2,487) | 30 Sep 2026, unrestricted |
| − Option exercise cash and JV stakes, net | (90) | Options $76.6m; Brookfield JV investments less unfunded commitments $13.7m |
| Enterprise value | 62,502 |
Source: Form 10-Q at 30 September 2026; reference price the last NYSE trade at 3:05 p.m. Eastern on 26 October 2026. The 0% notes would add 12.82m shares above $194.97.
| Measure | BE | Basis |
|---|---|---|
| EV / 2026E revenue | 13.9× | $4.5bn guidance midpoint |
| EV / 2027E revenue | 9.2× | $6.8bn consensus |
| EV / 2028E revenue | 6.3× | $9.9bn consensus |
| EV / 2026E non-GAAP operating income | 63× | $995m guidance midpoint |
| P/E, 2026E non-GAAP | 63× | $3.25 guidance midpoint |
| P/E, 2027E non-GAAP | 41× | About $5.0 consensus, before the results |
| Price / trailing free cash flow | n/m | Twelve-month free cash flow is negative |
| Company | Mkt value | EV | EV / 2026E revenue | EV / 2026E adj. profit | P/E 2026E | Growth, latest quarter | Backlog, latest |
|---|---|---|---|---|---|---|---|
| Bloom Energy $206.20 | $64.9bn | $62.5bn | 13.9× | 63× operating | 63× | +138% | ≈$20bn (Feb) |
| GE Vernova $859.11 | $228.8bn | $219.6bn | 4.8× | 37× EBITDA | n/m | +22% | $176bn |
| Vertiv $200.09 | $77.0bn | $76.9bn | 5.5× | 23× operating | 30× | +24% | $15.0bn (Dec 2025) |
| FuelCell Energy $11.59 | $0.9bn | ≤$0.4bn | ≤2.8× trailing | n/m | n/m | (29)% | $1.17bn |
Bloom Energy at the last NYSE trade at 3:05 p.m. Eastern on 26 October 2026; GE Vernova, Vertiv and FuelCell Energy at their last trades within a minute of it. Market value is price × the share count on the cover of each company’s latest Form 10-Q. GE Vernova: 266.33m shares; cash $13,120m, borrowings $2,794m, minority interests $1,158m at 30 June 2026; 2026 guidance from its second-quarter release (revenue $45.5–46.5bn, adjusted EBITDA margin 12–14%). Vertiv: 384.99m shares; cash and short-term investments $3,110.6m, debt $2,939.8m; 2026 guidance from its second-quarter release (net sales $13.8–14.2bn, adjusted operating margin 23.3–24.3%, adjusted EPS $6.65–6.75). FuelCell Energy: 79.95m shares, cash $658.1m and debt and other liabilities of up to $165.5m at 31 July 2026, so its EV is an upper bound. GE Vernova and Vertiv report their third quarters after this date; their figures are as at 30 June.
Cheap, fair or expensive? Expensive on every absolute measure, and closer to fair on growth-adjusted ones. Bloom trades at 13.9 times this year’s revenue against 4.8 times for GE Vernova and 5.5 times for Vertiv, while growing three to six times faster this year; on 2027 consensus the gap narrows to 9.2 times. On earnings, 41 times next year’s consensus compares with the 30 times Vertiv pays on this year’s, with growth that is higher but less visible: Vertiv and GE Vernova carry disclosed backlogs of $15bn and $176bn against revenue of $14bn and $46bn, while Bloom’s reported obligations are $0.72bn against $4.5bn. Bloom trades at a premium to its peers, not a discount, so the risks in Section 07 are not already paid for in the multiple; the DCF prices them once, in the scenarios and their weights.
| What would expand the multiple | What would compress it |
|---|---|
| A backlog figure in February that confirms growth well above revenue | A fourth quarter at the bottom of the $4.4–4.6bn range |
| A 2027 outlook above $7bn | Hyperscaler capital spending moderating, or a quarter of slower bookings |
| Contract assets turning into cash | Supplier commitments running ahead of deliveries |
| International and utility orders; a second hyperscaler on a gigawatt-scale project | Progress in the securities case, or fleet degradation costing more than the guarantee reserves allow |
Consensus inputs are secondary and pre-date the quarter: 2027 ($6.8bn) and 2028 ($9.9bn) come from aggregators before the third-quarter results. The cash-tax path is assumed: the 10-K gives $2.3bn of federal net operating losses, but deductions from stock pay and the 15% minimum tax make the timing uncertain; taxing fully from 2027 gives $127.57. Fourth-quarter working capital is held at the 30 September level. The customer consideration asset is added back as cash on an assumed schedule. Bloom’s backlog, last quantified in February, is not used as an input, and the $5.8bn of supplier commitments is not treated as a liability in the base case. Terminal value is 60% of enterprise value. Unmodelled: carbon-capture economics, international demand beyond the bull case, and the outcome of the securities case.
Bloom reported its third quarter, the three months to 30 September 2026, on 22 October after the close, and filed its Form 10-Q the same day. KR Sridhar, founder, chairman and chief executive, and Simon Edwards, chief financial officer, took questions. The shares closed at $217.77 that day and at $193.30 the next, down 11.2%.
| Q3 2026 | Value | Change y/y | Why it matters |
|---|---|---|---|
| Revenue | $1,235.8m | +138% | About 17% above the $1.06bn consensus |
| Product revenue | $1,073.6m | +179% | First $1bn product quarter |
| Non-GAAP gross margin | 34.5% | +4.1 pts | Product 38.4%, service 21.4%, installation (3.5)% |
| Non-GAAP operating income | $285.7m | from $46.2m | 23.1% margin; more than all of 2025 |
| GAAP operating income | $225.4m | from $7.8m | |
| Non-GAAP EPS | $0.94 | from $0.15 | Consensus $0.69; GAAP $0.78 |
| Stock-based compensation | $60.1m | +57% | 4.9% of revenue; $173.5m in nine months |
| Operating cash flow | $(27.1)m | from $19.7m | Working capital absorbed $356m |
| Free cash flow | $(180.6)m | from $7.4m | Nine months: +$41.6m |
| Contract assets | $858.6m | from $428.3m in June | $463.9m of the quarter’s revenue unbilled |
| Customer deposits | $661.4m | from $78.2m in December | Cash in advance on new agreements |
| Remaining performance obligations | $720.7m | from $494.1m in June | The order book in the financial statements |
| Cash and equivalents | $2,486.7m | (180.2) in the quarter | Against $2.5bn of 0% notes due 2030 |
Sources: Q3 2026 results release, supplemental and Form 10-Q, 22 October 2026; consensus from Investing.com. Free cash flow is operating cash flow less capital spending (derived).
One: a third raise, and the range now depends on a few deliveries. Revenue guidance went to $4.4–4.6bn, about 122% growth at the midpoint. The chief financial officer said the outlook is built from backlog, checked customer by customer against site readiness, plus a probability-weighted view of pipeline, and that the timing of a small number of year-end deliveries decides where in the range 2026 lands.
Two: the quarter used cash, and management asks to be judged over the year. Billing follows contract milestones, so revenue on delivered systems can be collected a quarter or more later. Management says contract assets will turn as milestones are reached, and that the business has funded several times 2025’s capital spending this year while holding cash roughly flat.
Three: capacity is approaching 5 GW, and Fremont is expanding to go past it. The chief executive called 5 GW a milestone on a longer path and declined to give a timeline or a quarterly cadence.
Four: new channels were the commercial news. AEP’s 702 MW plant near Cheyenne for Google, with carbon capture, under a 20-year power agreement; utilities buying Bloom systems for their own networks; a conditional US Army award to a partner; first units shipping to Texas; and international wins described as near.
Five: backlog is growing faster than revenue, but there is still no number. Asked how to read that claim with revenue up about 140%, the chief executive said the figure will come with the fourth-quarter results and would not bracket it.
| 2026 outlook | Feb 2026 | Apr 2026 | Jul 2026 | Oct 2026 |
|---|---|---|---|---|
| Revenue | $3.1–3.3bn | $3.4–3.8bn | $3.9–4.2bn | $4.4–4.6bn |
| Non-GAAP gross margin | ~32% | ~34% | ~34% | ~34% |
| Non-GAAP operating income | $425–475m | ~$675m | $800–900m | $960–1,030m |
| Non-GAAP EPS | $1.33–1.48 | $1.85–2.25 | $2.55–2.85 | $3.10–3.40 |
The fourth quarter implied by the new range is $1.35–1.55bn of revenue, $1.45bn at the midpoint, up 86% on a year earlier, and about $340m of non-GAAP operating income. Gross-margin guidance was held at about 34% despite higher revenue, on mix and the cost of ramping new lines. There is no 2027 outlook yet.
The tone was expansive and confident. The chief executive framed Bloom as leading the first architectural change in the power industry in 150 years. The chief financial officer was precise on what had happened and careful about what had not: cash was to be judged over the year, margins over the long run. Specific figures were absent where investors wanted them most: backlog, the capacity timeline and 2027.
| Firm | Question | Answer, in short |
|---|---|---|
| J.P. Morgan | Why expand Fremont now; why hold gross-margin guidance? | Preparing for life after the 5 GW milestone, no timeline; margins held back by ramp and expedite costs. |
| Evercore ISI | What drove the operating cash outflow? How big is the utility opportunity? | Judge cash over the year; the business is self-funding. Utilities can place Bloom systems on their own networks. |
| Morgan Stanley | Where do carbon capture, heat recovery and 800 V DC stand? | The AEP project for Google is the first large carbon-capture case; 800 V DC depends on new data-centre designs. |
| RBC | How large is the first Nebius block? | Servers are ready for all 300 MW; the customer builds block by block and sets the timing. |
| UBS | Total cost of power against the grid; water use | Cost per token, not cost per kilowatt-hour, is the right comparison. |
| Baird | Capacity cadence toward 5 GW; international demand | No quarterly figures; capacity follows demand; international news expected soon. |
| BMO | Largest single site today | Already in the 100 MW class; the Wyoming project is 702 MW. |
| Mizuho | How fast is backlog growing? | The number comes with the fourth-quarter results; no range given. |
| Bernstein | Hyperscaler interest in carbon capture | Every hyperscaler customer is asking; expects capture-ready systems to become standard. |
| Oppenheimer | When do contract assets turn into cash? | Balances may grow, but they turn as milestones are billed. |
Paraphrased from a published transcript of the 22 October call (The Motley Fool); every figure checked against the release and the Form 10-Q.
| Positive surprises | Negative surprises |
|---|---|
| Revenue about 17% above the $1.06bn consensus; non-GAAP EPS $0.94 against $0.69 | Operating cash outflow of $27m and free cash outflow of $181m in a record quarter |
| Operating expenses up 33% against revenue up 138% | Gross-margin guidance not raised with revenue |
| Customer deposits of $661.4m, eight times December’s | No backlog figure, eight months after the last one |
| A named hyperscaler, Google, behind AEP’s 702 MW order | $5.8bn of new supplier commitments, suppliers and materials not named |
| Product margin up again while deliveries were expedited | Performance-guarantee costs rising, attributed in the 10-Q to fleet degradation |
What to watch next. Where fourth-quarter revenue lands in the range and whether contract assets fall as billing milestones arrive; the backlog figure and the first 2027 outlook in February; whether 2027 guidance is consistent with the $2.6bn of supplier purchases committed for that year; the lead plaintiff’s amended complaint in the securities case; and the New Mexico air-permit hearing for Oracle’s Project Jupiter.
Bloom Energy makes solid oxide fuel cells, sold as modular Energy Servers. A fuel cell turns natural gas, biogas or hydrogen into electricity electrochemically rather than by burning it, so there is almost no NOx or SOx, little water use and little noise. The units are built in Fremont, California and Newark, Delaware, shipped as factory-tested skids and installed beside the building they power. The pitch to a data-centre developer is speed and permitting: Bloom can deliver power in months where a grid connection takes years, and a non-combustion source is easier to site near people than a turbine or an engine.
Most customers prefer to pay for power over time, so Bloom often sells the equipment to a financier, frequently a Brookfield vehicle, which owns it and sells power to the end user. In the filings the “customer” can therefore be the financier, not the data-centre operator that created the demand. Bloom then runs and maintains every system under contracts of 5 to 20 years.
| Revenue stream | Q3 2026 | Q3 2025 | Non-GAAP gross margin | Durability |
|---|---|---|---|---|
| ProductEnergy Servers | $1,073.6m | $384.3m | 38.4% | Recognised on delivery and acceptance; the growth engine |
| Installation | $84.9m | $65.8m | (3.5)% | Loss-making; moving to a consult-only model with certified contractors |
| ServiceOperations and maintenance | $67.6m | $58.6m | 21.4% | Recurring; 5–20 year terms, renewable annually |
| ElectricityLegacy power contracts | $9.8m | $10.4m | 25.7% | A shrinking book |
| Total | $1,235.8m | $519.0m | 34.5% | +138% year on year |
Source: Q3 2026 results release, GAAP to non-GAAP gross-margin reconciliation, 22 October 2026.
Who is paying. Revenue is concentrated, and the concentration moves between quarters. In the second quarter one customer that is not a related party accounted for 73% of revenue, a disclosure the company corrected in a Form 10-Q/A filed the day after the original. In the third quarter two customers accounted for 43% and 26%; the second is a related party, the Brookfield Fund JVs in which Bloom holds 6.2% and 15.0% stakes. Related-party revenue was $321.6m in the quarter. US customers were 97% of revenue. Oracle’s master agreement, Nebius in New Jersey (300 MW) and AEP’s 702 MW plant near Cheyenne for Google (announced 14–15 October; full operation by the end of 2028) are the names behind the numbers.
From customers. Remaining performance obligations were $720.7m at 30 September, against $419.4m in December and $494.1m in June. The measure excludes contracts that can be cancelled and contracts billed as work is done, so it understates what customers intend to buy. The 2025 Form 10-K defines product backlog as the revenue attributable to existing contractual commitments by a financier or an end customer, including expected tax incentives, and service backlog as contracted operation and maintenance, terminable for convenience each year. In February the company put the two at about $6bn and $14bn; on the call the chief executive said backlog is growing faster than revenue and that the number will come with the fourth-quarter results.
To suppliers. At 30 September Bloom had about $5.8bn of commitments under supply agreements longer than twelve months, covering raw materials and components, with take-or-pay or minimum-quantity terms in some: $0.1bn in the rest of 2026, $2.6bn in 2027, $1.3bn in 2028 and $1.8bn in 2029. For scale, product cost of revenue was $1,690.2m in the first nine months of 2026. The 10-Q does not name the suppliers or the materials. A company does not commit $2.6bn of 2027 purchases unless it expects 2027 volume well above 2026; if demand pauses, the obligation does not.
| Working capital, Q3 2026 | $m | What it is |
|---|---|---|
| Contract assets | (430.3) | Revenue recognised on delivered systems ahead of billing milestones |
| Accounts receivable | (175.2) | Billed, not yet collected |
| Inventories | (170.4) | Raw materials bought ahead of production |
| Other long-term assets | (178.0) | Not explained in the 10-Q |
| Deferred cost of revenue | (42.7) | Costs of systems shipped but not yet accepted |
| Deferred revenue and customer deposits | +309.3 | Cash received in advance |
| Accounts payable | +184.7 | Suppliers paid later |
| Accrued expenses and other | +158.8 | Purchase-order accruals $329.0m, from $76.9m in December |
| Accrued warranty | (7.5) | Specific reserve drawn down to $42.5m from $58.3m in June |
Source: Form 10-Q, statement of cash flows, three months to 30 September 2026. Negative figures used cash.
Net income of $249.0m and $57.2m of stock pay were more than absorbed by working capital. Of the quarter’s revenue, $463.9m had been recognised and not billed by 30 September. Over nine months the business funded itself: operating cash flow of $272.9m against capital spending of $231.3m. What the quarter shows is that the cash cost of growth rises with the size and lumpiness of the projects, and that two or three customers decide when it comes back.
New data-centre load now needs new generation. Gas turbines remain the low-cost option at scale, but GE Vernova reported gas backlog and slot reservations of 116 GW at the end of June, up from 100 GW three months earlier, so a new turbine order is a multi-year wait. Reciprocating engines from Caterpillar, Cummins and Wärtsilä are quick to deploy but noisy and high-emission. Fuel-cell rivals, FuelCell Energy and licensees of Ceres Power’s design, are small in data centres today; Vertiv sells the power and cooling inside the building rather than the generation. The grid itself is the long-run substitute. Opposition to data centres is the new variable: the 10-Q notes New York’s July 2026 moratorium on new hyperscale sites and similar measures introduced in many other states.
| $ millions | Q3 2026 | Q2 2026 | Q3 2025 |
|---|---|---|---|
| Revenue | 1,235.8 | 1,065.4 | 519.0 |
| Non-GAAP gross margin | 34.5% | 34.3% | 30.4% |
| GAAP operating income | 225.4 | 182.2 | 7.8 |
| Non-GAAP operating income | 285.7 | 239.6 | 46.2 |
| Adjusted EBITDA | 299.5 | 253.4 | 59.0 |
| Stock-based compensation | 60.1 | 56.4 | 38.2 |
| Operating cash flow | (27.1) | 226.4 | 19.7 |
| Capital expenditure | (153.5) | (51.6) | (12.3) |
| Free cash flow | (180.6) | 174.8 | 7.4 |
| Cash and cash equivalents | 2,486.7 | 2,666.9 | — |
Sources: Q3 2026 and Q2 2026 results releases and Forms 10-Q. Free cash flow derived.
Disclosed drivers: capacity approaching 5 GW a year with Fremont expanding beyond it; utilities placing Bloom systems on their networks; a conditional US Army award to a partner; first shipments to Texas; international wins described as close; carbon capture, cheaper on Bloom’s concentrated exhaust than on a turbine; and every system sold adding a maintenance contract. Assumed in our base case: consensus to 2028 and a gradual fade after it; nothing for carbon capture or international beyond that.
Risks, in order of weight: demand timing and concentration, with two customers 69% of the quarter; AI capital spending, which the 10-Q describes as increasingly debt-funded and harder to finance; the $5.8bn of fixed supplier commitments; scandium and the supply chain, where a short seller’s allegations are now the subject of a securities class action; fleet performance, with guarantee costs rising; permitting and community opposition; the time-to-power premium narrowing as turbine supply catches up; and dilution, with shares outstanding up 5.6% in nine months.
| Source | Score / 5 | Assessment |
|---|---|---|
| Brand | 3.5 | The default name in on-site fuel cells for data centres; little known outside power procurement |
| Network effects | 1.5 | None in the product; the financier ecosystem is an access advantage |
| Switching costs | 4 | Every system carries a long maintenance contract, and permits are written around Bloom’s emissions; the contracts are terminable annually |
| Cost advantage | 3 | Product margin 38% while expediting; turbines are cheaper per kilowatt at scale |
| Scale | 3.5 | Approaching 5 GW of annual capacity and $2.6bn of 2027 supply committed; small against the turbine and engine makers |
| Intellectual property | 4 | Two decades of stack chemistry and life engineering; non-scandium designs exist |
| Distribution | 3.5 | Brookfield and other financiers pre-arrange funding; utilities emerging as a channel; concentrated in a few counterparties |
| Regulation and permitting | 3.5 | No combustion; regulators can still block the gas supply instead |
| Data | 1.5 | Fleet telemetry improves service margins; not defensible |
| Customer loyalty | 3.5 | Repeat orders from Oracle and hyperscalers, who dual-source by habit |
Scores are our judgement. Composite about 3.2 out of 5: narrow but real.
Strongest advantages. Speed and permits together. A developer that cannot wait four years for a turbine slot, and cannot get an air permit for combustion near people, has few alternatives, and once a site is permitted around Bloom’s emissions profile, changing supplier restarts the clock. A quarter’s delay to a large AI campus costs its owner far more than the price difference between power technologies, so Bloom can price to the value of time rather than the cost of electricity. The third quarter’s product margin, rising while deliveries were expedited, is the evidence.
Biggest threats. The advantage is a scarcity rent. Turbine makers are adding capacity and interconnection reform is moving; when lead times shorten, the premium for speed narrows. Scandium is concentrated in China, and designs that do not use it would gain if supply tightened. Hyperscalers are the most capable buyers in the world and will fund a second source. And if the efficiency and output data a short seller compiled are representative, the total-cost argument weakens over a system’s life.
Direction: widening now, likely to narrow from around 2029. Scale, customer deposits and the carbon-capture design wins strengthen the position today; turbine supply catching up erodes it later. Our base case reflects that by giving back 2.5 points of operating margin between 2029 and 2034.
| Concern | Severity | Evidence and why it matters |
|---|---|---|
| Cash conversion | HIGH | Free cash flow $(180.6)m in the third quarter on record revenue. Contract assets $858.6m, from $241.2m in December; $463.9m of the quarter’s revenue recognised and not billed at its end. Nine-month free cash flow +$41.6m |
| Order-book disclosure | HIGH | Remaining performance obligations $720.7m. Bloom’s own backlog (about $6bn product, $14bn service) is defined in the 10-K but sits outside the financial statements, was last quantified in February, and has never been reconciled to the reported figure |
| Supplier commitments | HIGH | About $5.8bn committed beyond twelve months, from $16.4m in June; take-or-pay terms in some; suppliers and materials not named |
| Customer concentration | HIGH | One unrelated customer was 73% of second-quarter revenue (corrected in a 10-Q/A the day after filing); two customers 43% and 26% in the third; three customers 70% of receivables |
| Supply chain and litigation | HIGH | Hunterbrook alleges China-linked scandium supply (8 July); Bloom rejects it. A securities class action on the same allegations had a lead plaintiff and lead counsel appointed on 15 October |
| Fleet performance | MED-HIGH | Bloom attributes an $8.3m year-on-year rise in third-quarter performance-guarantee costs to fleet degradation. The cap on guarantee payments is $847.2m, from $585.4m in December; at most $466.1m remains payable |
| Related-party revenue | MEDIUM | $321.6m of third-quarter revenue (26%) from Brookfield Fund JVs in which Bloom holds equity |
| Dilution and non-GAAP | MEDIUM | Stock pay $173.5m in nine months, excluded from non-GAAP; shares outstanding up 5.6% since December; $2.5bn of 0% notes convert at $194.97 |
| Equity issued to an end customer | MEDIUM | Oracle received 2.15m shares worth $323.9m on its warrant; $306.5m sits as an asset that reduces reported revenue as Oracle systems deliver ($17.3m so far) |
| Warranty | MEDIUM | A specific warranty reserve, $58.3m in June and $42.5m in September, charged to product cost; its cause is not described |
| Insider activity | MEDIUM | About $21.8m of sales by officers and directors between 3 August and 1 October, mostly under 10b5-1 plans; no open-market purchases |
Severity is our judgement. Sources: Form 10-Q and 10-Q/A, Forms 4 filed 4 August–2 October 2026, Hunterbrook reports of 8 and 30 July 2026.
Third-party claims. Hunterbrook Capital, which disclosed a short position, published two reports in July. The first (8 July) alleged that Bloom’s scandium reaches it from China through intermediaries, questioned whether enough scandium exists for the capacity Bloom plans, and contrasted the $20bn backlog with $492.6m of remaining performance obligations at 31 March. Bloom said in an 8-K that it rejects the claims, that its supply is sufficient for current demand and backlog, and that it is not dependent on China. The second (30 July) used public generation data to argue that Bloom’s systems fall below their efficiency and output benchmarks earlier than the company has indicated. A shareholder class action filed in July in the Northern District of California repeats the supply-chain allegations; Bloom calls it without merit. We take no view on any of these claims and have not verified them. What can be checked is in Bloom’s own filings: the 10-Q attributes rising performance-guarantee costs to fleet degradation, and the cap on those payments has risen with the installed base.
| Strength | Weight | Evidence |
|---|---|---|
| Customer deposits | HIGH | $661.4m received in advance, from $78.2m in December, attributed to new agreements and milestone payments |
| Operating leverage | HIGH | Operating expenses +33%, revenue +138%; third-quarter operating income exceeded all of 2025’s |
| Forward volume signal | MED-HIGH | $2.6bn of supplier purchases committed for 2027 and $1.8bn for 2029: management is underwriting multi-year volume with its own balance sheet |
| Named demand | MED-HIGH | AEP for Google (702 MW, 20-year power agreement), Nebius (300 MW), Oracle |
| Balance sheet | MEDIUM | $2.49bn of cash against $2.52bn of convertible notes, almost all at 0%; $510m of undrawn revolver |
| Service annuity | MEDIUM | Service margin above 20% for two quarters; every new system adds a maintenance contract |
| Tax shelter | MEDIUM | Federal net operating losses of $2.3bn at December 2025 |
| Carbon-capture option | MEDIUM | Concentrated CO2 exhaust makes capture cheaper than on combustion plant; worth little in most models, including ours |
The balance of the evidence. The reported income statement is consistent with the cash generated over nine months and backed by customers paying deposits. The concerns come from the gap between what is binding and what is expected, all resting on a handful of counterparties. The strongest facts in Bloom’s favour are costly signals: customers paying cash before delivery, and management committing its own cash to suppliers three years ahead. Neither is consistent with a demand story that exists only in presentations.
| Dimension | Score / 5 | Assessment |
|---|---|---|
| CEO track record | 4 | KR Sridhar founded the company in 2001 and kept it going through two decades of losses; backed on-site power before the data-centre demand arrived. The 2019 restatement of earlier results is part of the record |
| CFO credibility | 3.5 | Simon Edwards, new this year (new-hire awards granted May 2026); clear on revenue recognition and cash timing |
| Guidance accuracy | 4.5 | Revenue outlook raised three times in 2026, from $3.1–3.3bn in February to $4.4–4.6bn |
| Transparency | 2.5 | Backlog not quantified since February; no capacity timeline; a 10-Q/A to correct transposed concentration figures; suppliers behind the $5.8bn not named |
| Capital allocation | 3.5 | Capacity added in steps funded from operations; projects financed off balance sheet through Brookfield; warrants and shares worth $323.9m issued to Oracle |
| Dilution | 2.5 | Shares outstanding up 5.6% in nine months; $173.5m of stock pay; 2.15m shares to Oracle; no buybacks |
| Insider ownership and dealing | 2.5 | Sales of about $21.8m by officers and directors since early August, no purchases; the chief executive gifted 300,000 shares |
| Compensation | 3 | The chief executive’s June 2026 performance award vests only after 3.5 years on annual targets, and the shares cannot be sold before the end of 2031 |
| Board quality | 3.5 | Experienced independent directors, including John Chambers and Jeffrey Immelt; combined chairman and chief executive |
Composite about 3.2 out of 5. Sources: Form 10-Q (executive awards), Forms 4 for transactions 3 August–1 October 2026, Q3 2026 call.
Do they act like long-term owners? In operations, yes. Committing $5.8bn to suppliers for three years, expanding Fremont ahead of the 5 GW milestone, and expediting deliveries at a cost to near-term margin are the decisions of people building for a decade, and the chief executive’s performance award cannot be sold before the end of 2031. In disclosure, less so: the company asks investors to rely on a backlog it has not quantified for eight months, while the measure in its financial statements is a small fraction of it, and the suppliers behind the largest new commitment in the filing are not named. Insider selling has continued through a rising and then falling share price, with no buying.
| Window | Event | What it decides | Impact · source |
|---|---|---|---|
| Late Oct 2026 | Hyperscaler third-quarter results and capital-spending guidance | The demand signal for 2027 | High · company calendars |
| Late Oct 2026 | GE Vernova and Vertiv third-quarter results | Turbine lead times; peer multiples | Medium · company calendars |
| Dec 2026 | Fourth-quarter deliveries | Where 2026 lands in $4.4–4.6bn | High · CFO, 22 Oct call |
| Q4 2026 | International orders described as near | A second leg outside US data centres | Medium · CEO, 22 Oct call |
| Unscheduled | Project Jupiter air-permit hearing, New Mexico | Oracle’s revised microgrid design | Medium · Oracle, 15 Oct 2026 |
| Late 2026 | Amended complaint in the securities case | Whether new allegations are made | Medium · court docket |
| Feb 2027 | Fourth-quarter results, backlog figure and first 2027 outlook | Whether the base case holds; consensus 2027 revenue $6.8bn | Very high · CEO, 22 Oct call |
| Feb 2027 | 2026 Form 10-K | Remaining obligations, the backlog definition, tax losses | Medium · filing calendar |
| From Jan 2027 | First full year of the new supplier commitments | $2.6bn due in 2027, some take-or-pay | Medium · Form 10-Q |
| Q4 2026 – Q1 2027 | Contract assets turning into cash | The self-funding claim | High · CFO, 22 Oct call |
| H1 2027 | Motion to dismiss in the securities case | Litigation risk | Medium · timing typical |
| 2027 | 5 GW capacity milestone and the Fremont expansion | Supply beyond 5 GW | High · CEO; no timeline given |
| 2027 | Nebius New Jersey (300 MW) ramp | Deliveries on the customer’s schedule | Medium · 22 Oct call |
| By end-2028 | AEP’s Cheyenne plant for Google at full operation | Carbon capture at scale | High · AEP, Oct 2026 |
| 2027–29 | Turbine supply normalising | How long the time-to-power premium lasts | High, slow · GE Vernova disclosures |
Dated where the company or a counterparty has given a date; “timing typical” where we infer it. Upside in each case is the event arriving as described or better; downside is a delay, a weaker figure or a new allegation. The single most informative item is the February backlog figure with the 2027 outlook, because the base case already assumes the consensus. Impact reflects how much the event can move the value, not the direction of its outcome.
BULLDemand. Look at what customers did with their own money. Deposits went from $78m in December to $661m in September. Nobody prepays for power equipment they do not intend to take. Revenue grew 138%, operating margin is 23%, and the outlook went up for the third time.
BEARAnd the quarter used $181m of cash. Contract assets doubled to $859m in three months. When two customers are 69% of a quarter, cash arrives when they decide, not when Bloom ships. The order book in the financial statements is $721m; the $20bn backlog has not been updated since February.
BULLVisibility. The reported figure leaves out everything cancellable or billed as delivered, which is most of how this industry contracts. The real tell is the supplier book: management committed $5.8bn of purchases through 2029 in one quarter. You do not sign take-or-pay contracts for a business you expect to shrink.
BEARYou also do not get out of them. If 2027 is a digestion year for AI spending, Bloom owns the inventory and the hyperscaler owns the option. The 10-Q itself says data-centre financing is getting harder and permits are taking longer.
BULLThe moat. The permitting problem is why Bloom wins. Turbine slots are booked years out, combustion near people is hard to permit, and every quarter of delay leaves a gigawatt campus idle. That is why product margin rose while Bloom was paying to expedite.
BEARThat is a scarcity price, and scarcity ends. GE Vernova is selling slot reservations by the gigawatt; when lead times shorten, the premium for speed goes with them. Meanwhile Bloom’s own filing attributes rising guarantee costs to fleet degradation, and a short seller says the fleet underperforms.
BULLQuality. Guarantee payments were $21m in nine months on about $3bn of revenue. And a 702 MW plant for Google with carbon capture under a 20-year contract is the kind of diligence a short report cannot buy.
BEARI will grant the business. I will not grant the price. At $206.20 the market needs revenue above our bull case, $26.4bn by 2030, at today’s margins. That leaves nothing for a single bad year.
BULLValuation. Consensus for 2028 is already $9.9bn. Companies on this kind of curve tend to beat the fade analysts pencil in.
BEARSome do. The ones priced for it have to.
The bull has the better of the argument on the business: the third quarter added costly signals that are hard to fake — deposits from customers, multi-year commitments to suppliers and a named hyperscaler behind a 702 MW order — and the cash outflow is explained by billing milestones and offset by positive free cash flow over nine months. The bear has the better of it on the price: our base case already accepts consensus to 2028 and takes revenue to 3.6 times 2026 guidance by 2030, and it is worth $130.24; the bull case is worth $205.84, and the price is close to it. Overvalued is the evidence-based conclusion. Most uncertain: whether the backlog is growing as fast as management says, and whether 2027 volume matches the supplier commitments. What would change it: a price of $164.34 or below; or a 2027 outlook and backlog that put the bull path in reach. Verify next: the February backlog figure and 2027 outlook, fourth-quarter contract assets and operating cash flow, and performance-guarantee payments as a share of service revenue.
Bloom Energy builds fridge-sized boxes that make electricity from natural gas without burning it. You put them next to a building, and the building makes its own power instead of waiting for the grid. AI data centres need huge amounts of power quickly, and a new grid connection can take years, so demand for Bloom’s boxes has exploded.
Mostly by selling the boxes: about 87% of last quarter’s revenue. Every buyer also signs a contract for Bloom to maintain them for years, which is a steady income. Installing them, a small part of the business, loses money.
Sales more than doubled this year and the company keeps raising its forecast. Big names are signing up: a power company is building a Bloom plant for Google, and Oracle and Nebius are customers.
Bloom has promised to buy about $5.8 billion of materials from its suppliers over the next three years, while the orders customers are firmly bound to in its accounts are much smaller. That promise shows how confident management is. If customers pause, Bloom still has to buy.
Even assuming sales grow about three and a half times by 2030, we think the shares are worth about $108 to $164 each, most likely $130.24. At $206.20 we call them overvalued: you would be paying for a better outcome than our optimistic case. Expensive shares can stay expensive for a long time; it means there is little cushion if something disappoints. The shares fell 11% in one day after these results.
Right: the backlog figure in February is large, the 2027 forecast beats expectations, and the unpaid bills turn into cash.
Wrong: AI spending slows and Bloom’s orders slow first; the supplier promises leave it with too much stock; or the short seller’s claims about where a rare material called scandium comes from, and how well the boxes perform over time, turn out to matter. Bloom says those claims are wrong, and some shareholders have sued.
Where the December quarter lands in the forecast range. The backlog figure and the first 2027 forecast, in February. And whether the unbilled sales from this quarter turn into cash over the winter.
| Status | What it covers |
|---|---|
| Verified from Bloom Energy filings and releases | Revenue by stream, margins, operating income, stock-based compensation, cash flow, capital spending, working capital, cash, the convertible notes and their terms, financing obligations, options and awards, remaining performance obligations, purchase commitments, the backlog definition, customer concentration, related-party revenue, warranty and performance-guarantee disclosures, the Oracle warrant and customer consideration asset, tax losses, insider transactions (Forms 4) and the 2026 outlook |
| Verified from market data and peer filings | Bloom Energy at the last NYSE trade at 3:05 p.m. Eastern on 26 October 2026; GE Vernova, Vertiv and FuelCell Energy at their last trades within a minute of it; peer share counts, debt and cash from their latest Forms 10-Q; the 10-year Treasury yield from the US Treasury daily par yield curve |
| Third-party, attributed | Call remarks and analysts’ questions from a published transcript of the 22 October 2026 call (The Motley Fool); consensus revenue and EPS (Investing.com, Yahoo Finance, 24/7 Wall St, Simply Wall St); Hunterbrook’s reports of 8 and 30 July 2026 and the class action, described without a view; Oracle and AEP project announcements |
| Our estimate or judgement | The revenue, margin, stock-pay, working-capital and capital-spending paths in every scenario; the cash-tax path; the customer consideration asset release schedule; the fourth-quarter stub; the scenario weights; the supplier-commitment arithmetic |
| Still open | The current backlog figure and its reconciliation to remaining obligations; the suppliers and materials behind the $5.8bn of commitments; the cause of the specific warranty reserve; the 2027 outlook; the outcome of the securities case |
Method. Unlevered discounted cash flow on a ten-year explicit path for a cash-generative operator, methodology version 1.3. The fourth quarter of 2026 is a stub to 31 December and each later year is discounted at its mid-point; the Gordon terminal value at 3.0% is discounted from the end of 2036; everything is valued at 26 October 2026 from the 30 September 2026 balance sheet. One discount rate (10.0%) and one terminal growth rate are used across all scenarios, and each material risk is priced once. Stock-based compensation is charged as a cost and unvested awards are excluded; vested options are exercised for cash; the 0% convertible notes convert at the implied value when it is above $194.97 and are debt at face below it. Scenario weights are our judgement, the fair-value band is the corners of the WACC 9–11% by terminal growth 2.5–3.5% grid, and the reverse valuation is run against the market price. The rating changes to fairly valued if the price falls to $164.34 or below with no change in the evidence, and to undervalued below $108.01; the conclusion is most sensitive to revenue after 2028, the long-run margin and the discount rate, in that order. Ratings are undervalued, fairly valued or overvalued according to where the price sits relative to the band; Sonde does not issue buy, sell or hold recommendations or price targets. The reference price used throughout is $206.20, the last NYSE trade at 3:05 p.m. Eastern on 26 October 2026, and every multiple in this report scales directly with it. Ratings are reviewed each quarter on results and immediately on a material contract, a financing event or a share-price move of more than 25%.