Bloom Energy
A maker of fuel cells that turn natural gas into electricity on the customer's own site, sold to data centres that cannot wait for the grid.
The scores
Growth4/4 ↑
Revenue up 138% in Q3 2026 and guidance raised again to roughly 122% growth for the year; the rate is real whatever its concentration.
Business quality2/4 →
A genuinely proprietary platform, but sold to unnamed financiers, threatened by cheaper engines and by generators signing grid contracts instead.
Financial position4/4 ↑
$2.49 billion of cash and positive nine-month operating cash flow, though Q3 2026 used $27 million; still the case's strongest part.
Execution record3/4 ↑
Delivered a real margin and cash-flow inflection and raised guidance, but the record is short and the beats reflect unmodellable revenue rather than proven forecasting.
Per share capture2/4 →
Share count up roughly 25% in fifteen months with convertible notes still outstanding, so per-share progress lags company progress by a wide margin.
Potential2/4 →
The demand shift is real and independently measured, but the prize is uncomposed and may be bounded by a materials supply nobody has sized.
Risk profile1/4 ↓
Two customers, one related, were 69% of Q3 2026 sales; a fraud allegation on the core input is only denied; no cohort checks it.
Valuation1/4 →
About 17.5 times sales against a prior median of 4.3, on revenue where two customers are 69% of the quarter and one is unnamed.
Evidence quality2/4 →
Financial claims are filed and solid, but the two questions that decide this case - the customer's identity and the scandium source - are undocumented.
If it works 0.8× · if it fails you keep about 4%.
The Verdict
We rate Bloom Energy unattractive. It makes fuel cells that turn natural gas into electricity on a customer's own site, and its latest quarters were a genuine turn to profit, but two customers bought about two-thirds of the sales, and a short seller and a lawsuit question its supply of a key metal, so the business is mixed. The price is stretched: it counts on many more quarters like that one, so a weak one would hurt. As a bet it is fragile: worth less than today even if it works, about a tenth left if sales fall back. The turn is real; what stands behind it is not disclosed.
The long caseAI data centres cannot wait years for a grid connection, so some build their own power; Bloom's fuel cells can be installed quickly, if buyers keep choosing them over turbines.
The evidence under this - 10 claims
The guidance is a forward-looking management figure, and its credibility rests on the same revenue concentration recorded in C06: if one unnamed customer was approximately 73% of first-half revenue, then the full-year figure is substantially a forecast about one buyer's build schedule rather than about market-wide demand, and the company's own release lists backlog recognition among its risk factors.
- 2026-10-22editedQ3 2026 results: full-year revenue guidance raised to $4.4-4.6B; the Q2 range of $3.9-4.2B is superseded.
- Concentration this extreme can be a sign of winning a very large customer rather than of fragility, and the customer is described as not a related party, which is an improvement on the FY2025 structure. But the direction is the wrong way: a year earlier the largest four customers were 30%, 18%, 15% and 11%, so the business became dramatically more concentrated as it grew, and the buyer whose orders are three-quarters of revenue is disclosed to the market only as a percentage.
- The largest customer is 43% of nine-month revenue in the Q3 10-Q, below the 73% single customer of the Q2 10-Q, so the worst reading of the first half is not repeated; but two customers at 69% of the September quarter, one of them the related party, is still a business that depends on two relationships, and the 10-Q's customer definition (the contractual counterparty, possibly a financing affiliate) means the filing does not say who the end buyers are.
- 2026-10-22editedQ3 2026 10-Q: largest customer 43% and second 26% of the quarter; the 73% six-month figure is not repeated and does not reconcile with the nine-month one.
Backlog measures that include expected tax credits are not unique to Bloom and the company discloses its definition plainly rather than hiding it, which is the opposite of concealment. The concern is not the disclosure but the arithmetic it permits: a figure that counts anticipated investment tax credits and 'commitments' that are not stated to be firm can be much larger than the contracted revenue an auditor would recognise, and Bloom's own Q2 2026 release lists 'the ability of current product and service backlog to ultimately be recognizable as revenue' among its risk factors.
- 2026-10-22editedQ3 2026 10-Q and call: still no filed contracted-revenue figure; backlog described as growing faster than revenue without a number.
Hunterbrook is a short seller with a financial position in the outcome and is therefore the least disinterested possible source on Bloom; its 220-ton and 240-ton figures also do not reconcile with the US Geological Survey's independent estimate of 40 tons of global production and 80 tons of capacity (C10), a discrepancy this thesis does not resolve and does not need to: the USGS figure is both authoritative and smaller, so the supply constraint it implies is tighter, not looser, than the short report's.
A filed complaint is an allegation and not a finding, plaintiffs' firms solicit class members as a matter of routine after any sharp price decline, and the class action may be dismissed. The 10-Q's silence is also defensible if the suit was filed after the July 28 filing date. What survives those caveats is that the company has answered the allegation only by denial, in a July 9, 2026 8-K and one answer on the Q2 call where the chief executive said that is all Bloom will say (C27), and has named no supplier or country of origin in any document.
- 2026-09-24editedCaption and court added from the lead-plaintiff notices (S19); counter-evidence corrected, because the company has answered the allegation by denial in a July 9, 2026 8-K and on the Q2 call (C27), which the original pass recorded as unanswered.
- 2026-10-22editedQ3 2026 10-Q: the suit is now disclosed, lead plaintiff and counsel appointed October 15; no ruling on a motion to dismiss yet.
- 2026-10-22editedQ3 2026 balance sheet added; cash is $2.49B, little changed since December while contract assets and receivables grew.
Comparing today's multiples against prior-year medians struck while the company was loss-making is close to meaningless on the earnings-based ratios, so the own-history comparison is only informative on the sales multiples, where 17.5 against 3.67 to 4.29 is a real expansion and not an artifact of the loss-making base.
- 2026-09-24editedRe-struck on the research pack refreshed at the 2026-09-24 review (S21): the price rose 34.9% between the two reviews with no quarter reported in between, so every price multiple moved and the own-history medians did not.
- 2026-09-26editedRe-struck on the research pack refreshed at the 2026-09-26 review (S22) at the September 25 close: the price rose again with no quarter reported, so every price multiple moved and the own-history medians did not.
- 2026-10-29editedRe-struck at the October 28 close ($228.08, 21.0% below the September 25 close) on the trailing revenue the September 10-Q adds; the earlier multiples (27.31 and 27.36) described a price that no longer trades.
The quarter's cash use is working capital rather than weaker earnings: contract assets (revenue recognised before it is billed) stood at $792.8 million against $178.9 million at December 31, 2025, while deferred revenue and customer deposits rose $610.6 million over nine months. Management calls the business self-funding and expects positive free cash flow for the full year, which is a forecast, not a filed result.
The company
An AI data centre can wait years for a grid connection. Bloom sells it a way around the wait: a fuel cell system on the customer's own site that turns natural gas, biogas or hydrogen into electricity without burning it and with no moving parts.
The buyer's other choices are gas engines, small gas turbines, or solar panels with batteries. About 1,100 sites in nine countries run on Bloom systems.
The evidence under this - 2 claims
How it makes money
Bloom often sells the system not to the user but to a financier who owns it and sells the power, because most users would rather pay only for the electricity they use. Service adds the rest. In the June quarter, products were $935.4M of the $1.07B, and Bloom kept 33.4 cents of each sales dollar after the cost of making it.
The evidence under this - 2 claims
Where the money comes from, stream by stream
| Stream | Share of sales (H1 2026) | How it earns it |
|---|---|---|
| Product | 87.5% | Energy Server fuel-cell systems sold to customers or to financiers who own them and sell the power on. |
| Service | 7.2% | Maintenance contracts on the installed fleet, net of payments under performance guarantees. |
| Installation | 4.2% | Site work to install the servers where Bloom does it; customers may use a third-party installer instead. |
| Electricity | 1.1% | Power sold from servers Bloom keeps on its own books under managed-services and lease contracts. |
Who runs it
KR Sridhar founded Bloom and has run it since 2002, taking it from years of losses to its first operating profit in 2024. A new finance chief, Simon Edwards, joined in April 2026. Insiders own about 3%, two-thirds of it Sridhar's. His bonus rides on sales and adjusted operating profit, his shares on product sales growth, product margin and winning large AI data-centre deals; none of it on cash or on profit per share.
The evidence under this - 3 claims
Insiders filed Form 144 notices of proposed sales and Form 4s through August, September and early October 2026, so the February holdings are likely higher than today's; the sales were not read for this block.
- The PSU targets are not disclosed, the board citing competitive harm, so how hard the goals are cannot be checked; in the first year the financial targets for one-third of the CEO's PSUs were met at 300%.
- Revenue growth is rewarded twice, in the cash bonus and the PSUs, while the share count that dilutes it (C14) is rewarded nowhere.
Why it wins, and who could beat it
Why customers come back
Bloom designs and prints its own cells and builds the stacks, so the chemistry, the machine and the factory are one platform it owns rather than a kit of bought parts.
From outside, that shows up as more than a decade of working installations at about 1,100 sites: the reference list a buyer of critical power checks before signing.
The evidence under this - 1 claim
The threat
The larger threat is a buyer choosing not to generate on site at all. Constellation and Vistra have signed 20-year contracts to supply Microsoft and Meta from existing power plants, 2,609 megawatts in Vistra's case alone.
The evidence under this - 1 claim
These contracts and Bloom's product are not strictly substitutes for every buyer: a power purchase agreement delivers electrons over the grid and remains subject to interconnection timing, whereas an Energy Server delivers power on site in months, so a buyer facing a five-year queue may need both. The threat is to the marginal decision rather than to the whole market.
The rivals: CAT, CMI, CEG, VST, GEV, PLUG.
Who else does this, rival by rival
| Ticker | Rival | How far along | Priced against it |
|---|---|---|---|
| CAT · CMI | Caterpillar and Cummins, the listed makers of natural-gas reciprocating gensets that Bloom's own Item 1 names as its first competitive classPriced against it: Same product | On sale | Same product |
| CEG · VST | Constellation Energy and Vistra, selling long-term co-located and grid-delivered power purchase agreements off existing nuclear and gas fleetsPriced against it: Different product | On sale | Different product |
| GEV | GE Vernova and Mitsubishi Power, in the small-gas-turbine class Bloom's Item 1 names secondPriced against it: Different product | On sale | Different product |
| PLUG | Plug PowerPriced against it: One segment only | On sale | One segment only |
How big it could get
How big this could get
Demand is measured: US government projections have data-centre servers rising from about 7% of commercial-building electricity in 2025 to 22% to 33% by 2050.
Supply may be the limit. The cells use scandium, and the world produced about 40 tons of it in 2024, mostly in China. A short seller estimates Bloom's five-gigawatt goal would need about 220 tons a year. Bloom has not said how much it uses, so demand and supply never meet in one number.
The evidence under this - 3 claims
Hunterbrook is a short seller with a financial position in the outcome and is therefore the least disinterested possible source on Bloom; its 220-ton and 240-ton figures also do not reconcile with the US Geological Survey's independent estimate of 40 tons of global production and 80 tons of capacity (C10), a discrepancy this thesis does not resolve and does not need to: the USGS figure is both authoritative and smaller, so the supply constraint it implies is tighter, not looser, than the short report's.
Who has signed
The firm buyers the filings show are customers Bloom does not name. One bought about 73% of first-half 2026 sales; in the September quarter two bought 43% and 26%, the second a related party. A year earlier, no single customer was more than 30% of a quarter's sales.
The evidence under this - 1 claim
- Concentration this extreme can be a sign of winning a very large customer rather than of fragility, and the customer is described as not a related party, which is an improvement on the FY2025 structure. But the direction is the wrong way: a year earlier the largest four customers were 30%, 18%, 15% and 11%, so the business became dramatically more concentrated as it grew, and the buyer whose orders are three-quarters of revenue is disclosed to the market only as a percentage.
- The largest customer is 43% of nine-month revenue in the Q3 10-Q, below the 73% single customer of the Q2 10-Q, so the worst reading of the first half is not repeated; but two customers at 69% of the September quarter, one of them the related party, is still a business that depends on two relationships, and the 10-Q's customer definition (the contractual counterparty, possibly a financing affiliate) means the filing does not say who the end buyers are.
- 2026-10-22editedQ3 2026 10-Q: largest customer 43% and second 26% of the quarter; the 73% six-month figure is not repeated and does not reconcile with the nine-month one.
Where the growth is coming from, business by business
Where the growth is coming from
| Business | Share of revenue | Growth: last quarter | Last half-year | Last year |
|---|---|---|---|---|
| Product | 87% | +179.4% | +195.0% | +170.4% |
| Installation | 7% | +35.2% | +36.3% | +40.0% |
| Service | 5% | +15.3% | +20.8% | +17.9% |
| Electricity | 1% | -5.2% | -14.7% | -37.4% |
| Whole company | 100% | +138.9% | +150.5% | +126.7% |
The filed split through 2026-09-30. These rates are not carried forward: the reading below says why.
How we read it: Believes guided ~100% growth, not the 188% blend: the rate rests on two customers taking 69% of Q3 2026, measured against a weak first-half 2025 base; guidance implies a slower second half.
The results
The numbers
Sales grew 111% over the twelve months to Sep 2026, faster than the 37% in the year to Dec 2025.
Operating margin is 15%, up from −16% in the year to Dec 2023.
Running costs take 18¢ of each sales dollar, down from 32¢ in the year to Dec 2021.
Shareholders' money earned 40% a year, up from −11× in the year to Dec 2021.
It turned 11% of sales into free cash, $437M over the last twelve months.
It holds $2.49B of cash against $2.48B of debt, $7.25M more cash than it owes.
The share count rose 87% in 4.8 years.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-22. Fiscal years end in December.
| Dec 2021 | Dec 2022 | Dec 2023 | Dec 2024 | Dec 2025 | 12 mo to Sep 2026 | |
|---|---|---|---|---|---|---|
| Size and growth | ||||||
| Sales | $972M | $1.20B | $1.33B | $1.47B | $2.02B | $3.83B |
| Sales growth | +22% | +23% | +11% | +11% | +37% | +111% |
| Profit | ||||||
| Gross margin | 20% | 12% | 15% | 27% | 29% | 32% |
| Operating margin | −12% | −22% | −16% | 1.6% | 3.6% | 15% |
| Net margin | −20% | −26% | −23% | −1.8% | −4.3% | 14% |
| What growth costs, per sales dollar | ||||||
| Research and development | 11% | 13% | 12% | 10% | 9.2% | 6.2% |
| Sales and marketing | 8.9% | 7.6% | 6.7% | 4.6% | 6.4% | 4.6% |
| General and admin | 13% | 14% | 12% | 11% | 9.8% | 6.8% |
| Capital spending | 5.1% | 9.7% | 6.3% | 4.0% | 2.8% | 6.6% |
| Cash | ||||||
| Cash from operations | −$60.7M | −$192M | −$373M | $92.0M | $114M | $691M |
| Free cash flow | −$110M | −$309M | −$456M | $33.1M | $57.2M | $437M |
| Balance sheet | ||||||
| Cash and short-term investments | $396M | $348M | $665M | $803M | $2.45B | $2.49B |
| Debt | $527M | $412M | $847M | $1.24B | $2.62B | $2.48B |
| Net cash | −$131M | −$63.1M | −$182M | −$440M | −$168M | $7.25M |
| Returns and shares | ||||||
| Return on capital | - | - | - | - | - | 18% |
| Shares (diluted) | 173M | 186M | 213M | 227M | 240M | 324M |
What growth costs
Growth is making Bloom more profitable: research, selling and office costs grow much more slowly than sales, so each new sale costs less to win. The catch is that this rests on the margin on each server holding up; when that slipped in 2022, losses deepened.
The evidence under this - 2 claims
- In 2022, when gross margin fell to 12.4%, R&D rose to 12.6 cents per sales dollar and operating margin to -21.8%, so the leverage depends on the gross margin holding.
- Sales and marketing rose from 4.6 cents in 2024 to 6.4 in 2025, and stock pay from 5.6 to 6.9 cents (C42), so not every overhead line is shrinking.
The last report (28 Jul)
Quarter ended June 2026, filed figures.
In the quarter ended September 30, 2026 sales were $1,235.8M, up 138%. Bloom kept 33.8 cents of each sales dollar after making the product, operating profit was $225.4M against $7.8M, and the yearly forecast rose to $4.4B to $4.6B.
The cash did not follow the quarter: operating cash flow was minus $27M, because billing lagged the sales Bloom booked. Two customers, one a related party, bought 69% of the quarter.
What it did to the thesisIt confirmed the profit turn and the forecast. It did not answer the doubts behind the Unattractive view: two buyers still hold most of the sales, no filing says where the scandium comes from, and the class action has a lead plaintiff.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $0.78 against $0.39 expected; $0.62 as filed, the gap being what the company adds back.
The evidence under this - 5 claims
The quarter's cash use is working capital rather than weaker earnings: contract assets (revenue recognised before it is billed) stood at $792.8 million against $178.9 million at December 31, 2025, while deferred revenue and customer deposits rose $610.6 million over nine months. Management calls the business self-funding and expects positive free cash flow for the full year, which is a forecast, not a filed result.
The guidance is a forward-looking management figure, and its credibility rests on the same revenue concentration recorded in C06: if one unnamed customer was approximately 73% of first-half revenue, then the full-year figure is substantially a forecast about one buyer's build schedule rather than about market-wide demand, and the company's own release lists backlog recognition among its risk factors.
- 2026-10-22editedQ3 2026 results: full-year revenue guidance raised to $4.4-4.6B; the Q2 range of $3.9-4.2B is superseded.
- Concentration this extreme can be a sign of winning a very large customer rather than of fragility, and the customer is described as not a related party, which is an improvement on the FY2025 structure. But the direction is the wrong way: a year earlier the largest four customers were 30%, 18%, 15% and 11%, so the business became dramatically more concentrated as it grew, and the buyer whose orders are three-quarters of revenue is disclosed to the market only as a percentage.
- The largest customer is 43% of nine-month revenue in the Q3 10-Q, below the 73% single customer of the Q2 10-Q, so the worst reading of the first half is not repeated; but two customers at 69% of the September quarter, one of them the related party, is still a business that depends on two relationships, and the 10-Q's customer definition (the contractual counterparty, possibly a financing affiliate) means the filing does not say who the end buyers are.
- 2026-10-22editedQ3 2026 10-Q: largest customer 43% and second 26% of the quarter; the 73% six-month figure is not repeated and does not reconcile with the nine-month one.
A company is not required to name every input material in its annual report, so this silence is not itself proof of dependence. Nor is it silence in every filing: on July 9, 2026 Bloom filed an 8-K (C27) rejecting the short report's conclusions and stating that its scandium oxide supply is sufficient for current demand and backlog and is not dependent on China, which also settles that Bloom's cells use scandium oxide, a point this ledger previously recorded as unconfirmed. What keeps the absence load-bearing is that the 8-K names no supplier and no country of origin, and the annual report's affirmative China statement (C08) still never names the input the allegation concerns.
- 2026-09-24editedCounter-evidence corrected in the open: it said Bloom had never confirmed its cells use scandium, but the July 9, 2026 8-K (S14), which the original pass did not open, discusses Bloom's scandium oxide supply directly. The claim text itself remains true.
- 2026-10-22editedQ3 2026 10-Q, release and call read: scandium still unnamed in any filing; the denial is repeated in shorter form.
A filed complaint is an allegation and not a finding, plaintiffs' firms solicit class members as a matter of routine after any sharp price decline, and the class action may be dismissed. The 10-Q's silence is also defensible if the suit was filed after the July 28 filing date. What survives those caveats is that the company has answered the allegation only by denial, in a July 9, 2026 8-K and one answer on the Q2 call where the chief executive said that is all Bloom will say (C27), and has named no supplier or country of origin in any document.
- 2026-09-24editedCaption and court added from the lead-plaintiff notices (S19); counter-evidence corrected, because the company has answered the allegation by denial in a July 9, 2026 8-K and on the Q2 call (C27), which the original pass recorded as unanswered.
- 2026-10-22editedQ3 2026 10-Q: the suit is now disclosed, lead plaintiff and counsel appointed October 15; no ruling on a motion to dismiss yet.
What management did with the money
Bloom is not short of money, with $2.67B of cash at June. But the share count has risen about 25% since March 2025, so each share owns a smaller slice of the turn.
Insiders in 2026Q2: 13 purchases and 18 sales, from their Form 4 filings.
The evidence under this - 2 claims
- 2026-10-22editedQ3 2026 balance sheet added; cash is $2.49B, little changed since December while contract assets and receivables grew.
Share growth at a company transitioning from losses to profit is partly the mechanical effect of stock compensation vesting and convertible instruments moving in the money as the price rises, rather than distress financing, and the equity raised has funded a balance sheet now holding $2.67 billion of cash. The offsetting point is that roughly 25% share growth in fifteen months means per-share progress lags company progress by a wide margin, and the convertible notes add further potential dilution the share count does not yet reflect.
- 2026-09-24editedWording made explicit that the reference price is the August 24, 2026 pack's, since the pack beside the thesis was refreshed at the 2026-09-24 review.
- 2026-10-22editedQ3 2026 10-Q: 295.5 million shares at September 30, 2026, up from 280.0 million at December 31, 2025.
The price, and where we disagree
In short. Priced for perfection. Bloom's turn to profit is real, but the price assumes it repeats many times, with most sales going to one unnamed buyer nobody outside can check.
The report is in. The shares closed at $228.08 on 28 October, 21% below the 25 September close our earlier price claims used, so those claims are now re-struck at that close.
The next test is the full-year report in early 2027. The lines to read are who bought, whether a filing says where the scandium comes from, and whether cash from operations turns positive again.
The evidence under this - 4 claims
- 2026-09-24editedRefreshed at the 2026-09-24 review on the refreshed research pack (S21); the 20-day range fell out of the 'Very volatile' bucket.
- 2026-09-26editedRefreshed at the 2026-09-26 review on the refreshed research pack (S22) at the September 25 close.
- 2026-10-29editedRe-struck at the October 28 close after the September-quarter report; the price fell 21.0% from the September 25 strike, over the 12% drift trigger. The moving-average, range and return figures were dropped because they were not recomputed.
Comparing today's multiples against prior-year medians struck while the company was loss-making is close to meaningless on the earnings-based ratios, so the own-history comparison is only informative on the sales multiples, where 17.5 against 3.67 to 4.29 is a real expansion and not an artifact of the loss-making base.
- 2026-09-24editedRe-struck on the research pack refreshed at the 2026-09-24 review (S21): the price rose 34.9% between the two reviews with no quarter reported in between, so every price multiple moved and the own-history medians did not.
- 2026-09-26editedRe-struck on the research pack refreshed at the 2026-09-26 review (S22) at the September 25 close: the price rose again with no quarter reported, so every price multiple moved and the own-history medians did not.
- 2026-10-29editedRe-struck at the October 28 close ($228.08, 21.0% below the September 25 close) on the trailing revenue the September 10-Q adds; the earlier multiples (27.31 and 27.36) described a price that no longer trades.
- Concentration this extreme can be a sign of winning a very large customer rather than of fragility, and the customer is described as not a related party, which is an improvement on the FY2025 structure. But the direction is the wrong way: a year earlier the largest four customers were 30%, 18%, 15% and 11%, so the business became dramatically more concentrated as it grew, and the buyer whose orders are three-quarters of revenue is disclosed to the market only as a percentage.
- The largest customer is 43% of nine-month revenue in the Q3 10-Q, below the 73% single customer of the Q2 10-Q, so the worst reading of the first half is not repeated; but two customers at 69% of the September quarter, one of them the related party, is still a business that depends on two relationships, and the 10-Q's customer definition (the contractual counterparty, possibly a financing affiliate) means the filing does not say who the end buyers are.
- 2026-10-22editedQ3 2026 10-Q: largest customer 43% and second 26% of the quarter; the 73% six-month figure is not repeated and does not reconcile with the nine-month one.
The quarter's cash use is working capital rather than weaker earnings: contract assets (revenue recognised before it is billed) stood at $792.8 million against $178.9 million at December 31, 2025, while deferred revenue and customer deposits rose $610.6 million over nine months. Management calls the business self-funding and expects positive free cash flow for the full year, which is a forecast, not a filed result.
How we read the multiple
Buyers pay about 17.5 times a year's sales for the whole company, debt included, against about 4 in Bloom's own recent past. Bloom has only just turned profitable.
That does not pay for the turn that happened; it pays for several more like it, from a customer list nobody outside can see.
No listed company does closely enough the same thing to check the price against.
Against its own history. Far dearer than its own past on every measure. The business did turn profitable, but the price has moved much further than the business has.
The evidence under this - 2 claims
Comparing today's multiples against prior-year medians struck while the company was loss-making is close to meaningless on the earnings-based ratios, so the own-history comparison is only informative on the sales multiples, where 17.5 against 3.67 to 4.29 is a real expansion and not an artifact of the loss-making base.
- 2026-09-24editedRe-struck on the research pack refreshed at the 2026-09-24 review (S21): the price rose 34.9% between the two reviews with no quarter reported in between, so every price multiple moved and the own-history medians did not.
- 2026-09-26editedRe-struck on the research pack refreshed at the 2026-09-26 review (S22) at the September 25 close: the price rose again with no quarter reported, so every price multiple moved and the own-history medians did not.
- 2026-10-29editedRe-struck at the October 28 close ($228.08, 21.0% below the September 25 close) on the trailing revenue the September 10-Q adds; the earlier multiples (27.31 and 27.36) described a price that no longer trades.
Priced against its rivals
There is no group to price it against. No filed cohort exists, so the sales multiple is judged against Bloom's own history alone, and that history was struck while the company was loss-making. The price is judged against its own history instead.
Private or foreign rivals that cannot be priced: Caterpillar Inc. and Cummins Inc. (gas reciprocating engines), Mitsubishi Power and Siemens Energy (small gas turbines).
The bet: what it pays if it works
Even if this works, the shares look expensive. Our best case is sales doubling again after this year's doubling, which needs Bloom to find about three more buyers as large as its one big unnamed customer. Even then, at a generous price for its sales, the company would be worth less than today.
If it fails, Bloom is still a real, profitable business with plenty of cash, but sales could fall back toward last year's level and the shares would be worth a fraction of today's price.
| Revenue in two to three years | Buyers that size needed | At the stated multiple |
|---|---|---|
| $6.00B | about 2 at 100% kept | 0.6× |
| $8.00B (the big win) | about 3 at 100% kept | 0.8× |
| $10.84B (market forecast) | about 5 at 100% kept | 1.1× |
| $12.00B | about 5 at 100% kept | 1.2× |
Forecasts put US on-site power for data centres near 20 GW a year by 2029; at 15% Bloom ships 3 GW, half again its 2026 capacity, about $8.8 billion plus today's business. On that forecast, $10.84B pays 1.1× from today's value: an Expensive bet on the market case.
To pay 2× (a Fair bet) it would need about $20.43B of revenue (about 10 buyers that size); to pay 5× (a Cheap bet), about $51.08B of revenue (about 26 buyers that size), at the same multiple.
The steps it has to take
- Done Operating profit and positive operating cash flow in a quarter
- Done A named hyperscaler site chooses Bloom: Oracle's Project Jupiter permit application
- Slipped Q3 2026 10-Q: concentration falls below the 73% one-buyer share, with customers named due 2027-02
- Slipped Project Jupiter's gas pipeline in service and Bloom's units running there due 2027-02
- In progress Full-year 2026 revenue inside the $3.9B to $4.2B guidance due 2027-02
- Not started A filed remaining-performance-obligation figure in place of the company-defined order book due 2027-02
- Not started A filed scandium supply chain outside China, and the class action dismissed due 2027-12
- Not started Three or more anchor-sized buyers, revenue near $8B a year due 2028-12
The detail behind the label
The evidence under this - 6 claims
The guidance is a forward-looking management figure, and its credibility rests on the same revenue concentration recorded in C06: if one unnamed customer was approximately 73% of first-half revenue, then the full-year figure is substantially a forecast about one buyer's build schedule rather than about market-wide demand, and the company's own release lists backlog recognition among its risk factors.
- 2026-10-22editedQ3 2026 results: full-year revenue guidance raised to $4.4-4.6B; the Q2 range of $3.9-4.2B is superseded.
- Concentration this extreme can be a sign of winning a very large customer rather than of fragility, and the customer is described as not a related party, which is an improvement on the FY2025 structure. But the direction is the wrong way: a year earlier the largest four customers were 30%, 18%, 15% and 11%, so the business became dramatically more concentrated as it grew, and the buyer whose orders are three-quarters of revenue is disclosed to the market only as a percentage.
- The largest customer is 43% of nine-month revenue in the Q3 10-Q, below the 73% single customer of the Q2 10-Q, so the worst reading of the first half is not repeated; but two customers at 69% of the September quarter, one of them the related party, is still a business that depends on two relationships, and the 10-Q's customer definition (the contractual counterparty, possibly a financing affiliate) means the filing does not say who the end buyers are.
- 2026-10-22editedQ3 2026 10-Q: largest customer 43% and second 26% of the quarter; the 73% six-month figure is not repeated and does not reconcile with the nine-month one.
Backlog measures that include expected tax credits are not unique to Bloom and the company discloses its definition plainly rather than hiding it, which is the opposite of concealment. The concern is not the disclosure but the arithmetic it permits: a figure that counts anticipated investment tax credits and 'commitments' that are not stated to be firm can be much larger than the contracted revenue an auditor would recognise, and Bloom's own Q2 2026 release lists 'the ability of current product and service backlog to ultimately be recognizable as revenue' among its risk factors.
- 2026-10-22editedQ3 2026 10-Q and call: still no filed contracted-revenue figure; backlog described as growing faster than revenue without a number.
- 2026-10-22editedQ3 2026 balance sheet added; cash is $2.49B, little changed since December while contract assets and receivables grew.
Comparing today's multiples against prior-year medians struck while the company was loss-making is close to meaningless on the earnings-based ratios, so the own-history comparison is only informative on the sales multiples, where 17.5 against 3.67 to 4.29 is a real expansion and not an artifact of the loss-making base.
- 2026-09-24editedRe-struck on the research pack refreshed at the 2026-09-24 review (S21): the price rose 34.9% between the two reviews with no quarter reported in between, so every price multiple moved and the own-history medians did not.
- 2026-09-26editedRe-struck on the research pack refreshed at the 2026-09-26 review (S22) at the September 25 close: the price rose again with no quarter reported, so every price multiple moved and the own-history medians did not.
- 2026-10-29editedRe-struck at the October 28 close ($228.08, 21.0% below the September 25 close) on the trailing revenue the September 10-Q adds; the earlier multiples (27.31 and 27.36) described a price that no longer trades.
Where we disagree with the market
Bloom is the cleanest way to own on-site power for AI data centres, and the July short report was noise the results answered.
The profit turn is real, but it rests on one unnamed buyer, and the supply questions go to whether Bloom can grow.
Coverage reads the beat and the pile of future orders Bloom advertises. The filings show what those leave out: the concentration, an order tally Bloom defines itself that counts tax credits it expects to receive, and a denial on scandium that names no supplier.
Seventeen analysts rate it a buy, and the shares are up 317% in a year despite the July report. The market has decided the results answered the questions; the filings say they did not.
What analysts say, and the evidence
The evidence under this - 6 claims
- Concentration this extreme can be a sign of winning a very large customer rather than of fragility, and the customer is described as not a related party, which is an improvement on the FY2025 structure. But the direction is the wrong way: a year earlier the largest four customers were 30%, 18%, 15% and 11%, so the business became dramatically more concentrated as it grew, and the buyer whose orders are three-quarters of revenue is disclosed to the market only as a percentage.
- The largest customer is 43% of nine-month revenue in the Q3 10-Q, below the 73% single customer of the Q2 10-Q, so the worst reading of the first half is not repeated; but two customers at 69% of the September quarter, one of them the related party, is still a business that depends on two relationships, and the 10-Q's customer definition (the contractual counterparty, possibly a financing affiliate) means the filing does not say who the end buyers are.
- 2026-10-22editedQ3 2026 10-Q: largest customer 43% and second 26% of the quarter; the 73% six-month figure is not repeated and does not reconcile with the nine-month one.
Backlog measures that include expected tax credits are not unique to Bloom and the company discloses its definition plainly rather than hiding it, which is the opposite of concealment. The concern is not the disclosure but the arithmetic it permits: a figure that counts anticipated investment tax credits and 'commitments' that are not stated to be firm can be much larger than the contracted revenue an auditor would recognise, and Bloom's own Q2 2026 release lists 'the ability of current product and service backlog to ultimately be recognizable as revenue' among its risk factors.
- 2026-10-22editedQ3 2026 10-Q and call: still no filed contracted-revenue figure; backlog described as growing faster than revenue without a number.
The statement names no supplier, no country of origin and no tonnage, so it cannot be checked against the short report's specific allegation that Chinese-origin scandium reaches Bloom through intermediaries in Thailand, Japan and South Korea (C11); 'not dependent on China' is compatible with some Chinese-origin supply. The US Geological Survey names China as the leading producer of a market of about 40 tons a year (C10). And with the chief executive declining to say more, the class action (C12) rather than a company filing is now the likeliest place the sourcing gets tested.
- 2026-10-22editedQ3 2026 10-Q and call read: the denial is repeated more narrowly and nothing further was said.
A company is not required to name every input material in its annual report, so this silence is not itself proof of dependence. Nor is it silence in every filing: on July 9, 2026 Bloom filed an 8-K (C27) rejecting the short report's conclusions and stating that its scandium oxide supply is sufficient for current demand and backlog and is not dependent on China, which also settles that Bloom's cells use scandium oxide, a point this ledger previously recorded as unconfirmed. What keeps the absence load-bearing is that the 8-K names no supplier and no country of origin, and the annual report's affirmative China statement (C08) still never names the input the allegation concerns.
- 2026-09-24editedCounter-evidence corrected in the open: it said Bloom had never confirmed its cells use scandium, but the July 9, 2026 8-K (S14), which the original pass did not open, discusses Bloom's scandium oxide supply directly. The claim text itself remains true.
- 2026-10-22editedQ3 2026 10-Q, release and call read: scandium still unnamed in any filing; the denial is repeated in shorter form.
Beating consensus by 100% and 255% in consecutive quarters is not primarily evidence of operational outperformance; it is evidence that the analyst community cannot model this revenue line, which is what one would expect when three-quarters of revenue comes from a single undisclosed customer whose delivery schedule outsiders cannot see.
- 2026-09-24editedRe-checked at the 2026-09-24 review against the refreshed pack (S21): the rating split and the surprise history are unchanged.
- 2026-09-24editedRefreshed at the 2026-09-24 review on the refreshed research pack (S21); the 20-day range fell out of the 'Very volatile' bucket.
- 2026-09-26editedRefreshed at the 2026-09-26 review on the refreshed research pack (S22) at the September 25 close.
- 2026-10-29editedRe-struck at the October 28 close after the September-quarter report; the price fell 21.0% from the September 25 strike, over the 12% drift trigger. The moving-average, range and return figures were dropped because they were not recomputed.
What could go wrong, and how we would know
The risks
- Two big customers. Two buyers took about 69% of the September quarter, one of them a related party, and nobody outside can see who they are.
- Cash lags the sales. Operating cash flow was minus $27M in the quarter while billed-but-unpaid and not-yet-billed sales grew by about $880M over nine months.
- Future orders nobody can check. Bloom's own tally counts tax credits it expects to receive; the audited figure for contracted revenue appears nowhere in its annual report.
- A fraud allegation. A short seller says Bloom buys Chinese scandium through middlemen, and a class action followed. Bloom denied it without naming a supplier.
- Its biggest named site is late. Oracle's New Mexico campus will run on Bloom cells, but its gas pipeline slipped to February 2027 and Oracle sent a force majeure notice.
- Insiders selling, none buying. 55 open-market sales and no purchases in a year, including four senior officers on 16 September.
The evidence under this - 8 claims
- Concentration this extreme can be a sign of winning a very large customer rather than of fragility, and the customer is described as not a related party, which is an improvement on the FY2025 structure. But the direction is the wrong way: a year earlier the largest four customers were 30%, 18%, 15% and 11%, so the business became dramatically more concentrated as it grew, and the buyer whose orders are three-quarters of revenue is disclosed to the market only as a percentage.
- The largest customer is 43% of nine-month revenue in the Q3 10-Q, below the 73% single customer of the Q2 10-Q, so the worst reading of the first half is not repeated; but two customers at 69% of the September quarter, one of them the related party, is still a business that depends on two relationships, and the 10-Q's customer definition (the contractual counterparty, possibly a financing affiliate) means the filing does not say who the end buyers are.
- 2026-10-22editedQ3 2026 10-Q: largest customer 43% and second 26% of the quarter; the 73% six-month figure is not repeated and does not reconcile with the nine-month one.
Backlog measures that include expected tax credits are not unique to Bloom and the company discloses its definition plainly rather than hiding it, which is the opposite of concealment. The concern is not the disclosure but the arithmetic it permits: a figure that counts anticipated investment tax credits and 'commitments' that are not stated to be firm can be much larger than the contracted revenue an auditor would recognise, and Bloom's own Q2 2026 release lists 'the ability of current product and service backlog to ultimately be recognizable as revenue' among its risk factors.
- 2026-10-22editedQ3 2026 10-Q and call: still no filed contracted-revenue figure; backlog described as growing faster than revenue without a number.
Hunterbrook is a short seller with a financial position in the outcome and is therefore the least disinterested possible source on Bloom; its 220-ton and 240-ton figures also do not reconcile with the US Geological Survey's independent estimate of 40 tons of global production and 80 tons of capacity (C10), a discrepancy this thesis does not resolve and does not need to: the USGS figure is both authoritative and smaller, so the supply constraint it implies is tighter, not looser, than the short report's.
A filed complaint is an allegation and not a finding, plaintiffs' firms solicit class members as a matter of routine after any sharp price decline, and the class action may be dismissed. The 10-Q's silence is also defensible if the suit was filed after the July 28 filing date. What survives those caveats is that the company has answered the allegation only by denial, in a July 9, 2026 8-K and one answer on the Q2 call where the chief executive said that is all Bloom will say (C27), and has named no supplier or country of origin in any document.
- 2026-09-24editedCaption and court added from the lead-plaintiff notices (S19); counter-evidence corrected, because the company has answered the allegation by denial in a July 9, 2026 8-K and on the Q2 call (C27), which the original pass recorded as unanswered.
- 2026-10-22editedQ3 2026 10-Q: the suit is now disclosed, lead plaintiff and counsel appointed October 15; no ruling on a motion to dismiss yet.
Concentrated selling after a large twelve-month move is what any rational holder of appreciated equity compensation does. The Form 4s now answer the plan question the first pass could not: of the nine open-market sales from August 13 to September 16, 2026, seven, carrying 92.2% of the shares sold, were marked as made under a Rule 10b5-1 plan, including Immelt's and Chambers'. The two that were not are small - the Chief Accounting Officer's 2,295 shares on September 16 and the Chief Legal Officer's 2,895 shares on August 14 - so the pattern is mostly scheduled selling, not a run of discretionary judgements.
- 2026-09-24editedRefreshed at the 2026-09-24 review: four more open-market sales and no purchases since the thesis, and the Rule 10b5-1 box now read on each Form 4 (S20) rather than left unknown.
- 2026-10-22editedInsider read refreshed at the Q3 print: one more planned sale on October 1, no purchases.
The statement names no supplier, no country of origin and no tonnage, so it cannot be checked against the short report's specific allegation that Chinese-origin scandium reaches Bloom through intermediaries in Thailand, Japan and South Korea (C11); 'not dependent on China' is compatible with some Chinese-origin supply. The US Geological Survey names China as the leading producer of a market of about 40 tons a year (C10). And with the chief executive declining to say more, the class action (C12) rather than a company filing is now the likeliest place the sourcing gets tested.
- 2026-10-22editedQ3 2026 10-Q and call read: the denial is repeated more narrowly and nothing further was said.
On the July 28, 2026 Q2 call the chief financial officer said that where a project is delayed the equipment can be redeployed to other projects and 'the financier is on the hook to take delivery of the equipment from Bloom', and the chief executive said 2026 guidance is 'not dependent on any single project' (S15); Oracle is reported to remain committed to the campus. So the delay may move where Bloom's servers are installed rather than when Bloom recognises the sale. What this ledger cannot establish is whether Project Jupiter sits behind the single customer that was 73% of first-half revenue (C06), because that customer is unnamed.
- 2026-10-22editedQ3 2026 release, 10-Q and call read: none restates the figure or the delay in this claim; Oracle appears only for the warrant and the Utah site, so it rolls to the annual report.
The quarter's cash use is working capital rather than weaker earnings: contract assets (revenue recognised before it is billed) stood at $792.8 million against $178.9 million at December 31, 2025, while deferred revenue and customer deposits rose $610.6 million over nine months. Management calls the business self-funding and expects positive free cash flow for the full year, which is a forecast, not a filed result.
What we still do not know
- Who the customers behind 69% of September-quarter sales are, and why the Q3 10-Q's nine-month figures do not match the 73% in the Q2 one.
- Where Bloom's scandium comes from. It says not China, and the 10-Q names no supplier.
The evidence under this - 2 claims
- Concentration this extreme can be a sign of winning a very large customer rather than of fragility, and the customer is described as not a related party, which is an improvement on the FY2025 structure. But the direction is the wrong way: a year earlier the largest four customers were 30%, 18%, 15% and 11%, so the business became dramatically more concentrated as it grew, and the buyer whose orders are three-quarters of revenue is disclosed to the market only as a percentage.
- The largest customer is 43% of nine-month revenue in the Q3 10-Q, below the 73% single customer of the Q2 10-Q, so the worst reading of the first half is not repeated; but two customers at 69% of the September quarter, one of them the related party, is still a business that depends on two relationships, and the 10-Q's customer definition (the contractual counterparty, possibly a financing affiliate) means the filing does not say who the end buyers are.
- 2026-10-22editedQ3 2026 10-Q: largest customer 43% and second 26% of the quarter; the 73% six-month figure is not repeated and does not reconcile with the nine-month one.
The statement names no supplier, no country of origin and no tonnage, so it cannot be checked against the short report's specific allegation that Chinese-origin scandium reaches Bloom through intermediaries in Thailand, Japan and South Korea (C11); 'not dependent on China' is compatible with some Chinese-origin supply. The US Geological Survey names China as the leading producer of a market of about 40 tons a year (C10). And with the chief executive declining to say more, the class action (C12) rather than a company filing is now the likeliest place the sourcing gets tested.
- 2026-10-22editedQ3 2026 10-Q and call read: the denial is repeated more narrowly and nothing further was said.
What would change our view
- Unattractive→AttractiveThe revenue base becomes legible, at a multiple nearer its own history
It needs all of:
- ○The two largest customers fall well below their 69% of September-quarter sales, replaced by customers Bloom names.checked at the next results, due by 4 Feb
- ○Bloom reports the audited figure for revenue it is contracted to earn, instead of a tally of future orders that counts expected tax credits.checked at the next results, due by 4 Feb
- ○The shares fall to about 8 times sales: still double its old normal, but paying only for the turn that happened.about ≈$84.42 today, 70% below the last close
- Reopens the researchThe two overhangs clear
It needs:
- ○A filing documents a scandium supply outside China, and the securities class action is dismissed.watched as it happens
We would also be wrong if
- Margins above 33% and cash coming in hold for four straight quarters; the September quarter's cash flow was negative, so the count restarts.
- Insiders stop selling and buy shares with their own money.
What is coming
- 27 October 2026 Q3 2026 results and 10-Q: Whether one unnamed customer is again most of revenue, whether the full-year $3.9B to $4.2B guide holds, and whether the 10-Q says anything on scandium or the class action.
Every dated event ahead
- 27 October 2026 Q3 2026 results and 10-Q: Whether one unnamed customer is again most of revenue, whether the full-year $3.9B to $4.2B guide holds, and whether the 10-Q says anything on scandium or the class action.
- December 2026 Fremont reaches a 2 GW annual production run rate: Whether the doubling of factory capacity the 10-K promised for end-2026 arrives on time; the Oracle and Nebius deliveries need it.
- February 2027 Q4 2026 results and the FY2026 10-K: Full-year revenue against the guide, the year's customer concentration, the backlog definition, and whether 2027 guidance names new customers.
Since we wrote this
- 29 OctPrice moveShares fell 21%; price figures re-struck, view heldAbout 17.5 times sales now, from 27. Two customers still hold 69% of sales and scandium is unexplained, so a lower price alone does not change the view.
- 22 OctEarningsSales up 138% and guidance raised, but two buyers are 69%The single 73% buyer did not repeat, yet two customers still took 69% of sales, scandium went unmentioned and operating cash flow was minus $27M. View held.
- 6 OctThesis reviewThe bet's failure case is re-valued from the June reportOne customer was about 73% of first-half revenue, so the failure case now rests on what the rest bought, about $3.6B of value, not the 2025 revenue level.
- 5 OctThesis reviewThe bet now names what it ridesThe owner asked that every bet name the macro trend behind it, after BWXT's bet was missing its chief executive's case for nuclear over gas.
- 4 OctThesis reviewOwner's checklist: revenue streams, cost of growth, management, catalystsThe owner's checklist found the method silent on who runs the company, what growth costs, where the money comes from and what is dated ahead; this pass fills those four blocks from the proxy and filings.
- 26 SepThesis reviewPrice multiple rises to 27 times sales; a data note correctedThe shares rose again with no new results, so the price asks more of one unnamed customer. A note had called March-quarter figures June's; the June quarter was always read from the filing.
- 24 SepNewsBloom's July scandium denial found; biggest named site delayed. View heldThe denial names no supplier, so the risk stands. The shares rose 34.9% with no new results, which made the price objection stronger.
- 24 AugThesis writtenUnattractive, medium convictionWritten a month after the June quarter's profit turn, with one unnamed buyer at 73% of first-half sales and a fraud claim over scandium outstanding.
Also checked 6 times with nothing new to act on.
Sources 36
Sources - 36
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2026-02-09product and platform description, competition classes, customer concentration, backlog definition, the China supply-chain statement, SK ecoplant and Brookfield relationships, turbine supply-constraint assertionbacksC01C02C05C07C08C09C15C20C21C34C42C43
- S02primarycompany filing2026-07-28Q2 2026 revenue, margin, operating income, operating cash flow and per-share figures; raised FY2026 guidance; balance-sheet detail; the backlog-recognition risk factor; the absence of any China or scandium referencebacksC03C09C12C13C19C24C25
- S03primarycompany filing2026-07-28the H1 2026 single-customer 73% concentration disclosure, quarterly concentration detail, accounts-receivable concentration, share-count roll-forward and convertible-debt premiumbacksC06C12C13C14C34C38C39C42C43
- S04primarycompany filing2026-02-05the full-year 2025 results release filed alongside the annual report, read for the prior-year guidance and comparison baseread, but no claim rests on it
- S05secondaryvendor data2026-08-24Portfolio-management research pack (market tier: price, technicals, insiders, analysts, own-history valuation)content/BE/research-pack.jsonprice and technicals snapshot, volatility bucket and beta, insider Form 4 aggregation, analyst consensus and earnings-surprise history, own-history valuation time seriesbacksC14
- S06secondaryvendor data2026-08-24Portfolio-management research pack (vendor_fundamentals tier: statements, peer_valuation, segments)content/BE/research-pack.jsonthe filed annual and quarterly revenue and margin series from the filer's own XBRL, the peer_valuation unavailable result, the no_segment_axis reading, and the 2024 fiscal-year labelling defectbacksC04C16C24
- S07primaryindustry data2025-01US Geological Survey - Mineral Commodity Summaries, January 2025, Scandium chapter US Geological Surveythe authoritative independent measure of global scandium supply, China's position in it, and the absence of United States mine production - the external bound on the prize and the external test of the moat
In 2024, the global consumption of scandium oxide was estimated to be about 30 to 40 tons per year with a global capacity estimate of 80 tons per year. According to industry estimates, global production totaled 40 tons. Scandium was recovered from cobalt, nickel, titanium, and zirconium process streams. China was the leading producer.
backsC10 - S08tertiarynews2026-07coverage of the July 8, 2026 Hunterbrook short report, its scandium and backlog allegations, and the market reaction
Hunterbrook released a report titled "Bloom's Big Lie"
backsC11 - S09secondarynews2026-08the existence of the securities class action, its class period of February 27, 2025 to July 8, 2026, the alleged scandium and China misstatements, and the September 28, 2026 lead-plaintiff deadline
Bloom Energy Corporation (BE) Investors to a Securities Class Action Deadline on September 28, 2026
backsC11C12 - S10secondaryvendor data2026-08-24Portfolio-management filed-competitor roster (data/reference/competitors.json, BE entry)data/reference/competitors.jsonthe recorded cohort search: the classes Bloom names, the candidates chased, and the verdict on eachbacksC16
- S11secondaryindustry data2026independent tracking of behind-the-meter generation actually online, reused from this repo's Vistra and Constellation Energy theses
approximately 2 GW of behind-the-meter generation capacity is online across four projects, with the vast majority belonging to xAI
backsC22 - S12primaryindustry data2026EIA - Data center server energy use grows across the commercial building stock US Energy Information Administrationthe independent national demand base for the prize block, reused from this repo's Constellation Energy thesis
Data center server electricity use grows to 22%-33% of commercial building electricity use by 2050 across our cases.
backsC23 - S13secondaryvendor data2026-08-24Portfolio-management theses for Constellation Energy (CEG) and Vistra (VST), machine.demand blockscontent/CEG/thesis.json and content/VST/thesis.jsonthe named, quantified long-term data-center power agreements the merchant generators have signed, each traced in those theses to the sellers' own 8-K filings - the competing option Bloom's own Item 1 does not listbacksC26
- S14primarycompany filing2026-07-09Bloom's own filed response to the short report: it rejects the report's conclusions on scandium oxide and states its supply is sufficient for current demand and backlog and not dependent on China, with visibility to 25GW of fuel cells a yearbacksC09C27
- S15primarytranscript2026-07-28Bloom Energy Q2 2026 earnings call transcript (FMP), held 2026-07-28, question-and-answer sectionthe chief executive's answer on scandium (David Arcaro's question), the chief financial officer's answer on project delays ('the financier is on the hook to take delivery of the equipment from Bloom') and the statement that 2026 guidance is 'not dependent on any single project'; Oracle described as 'our first direct hyperscaler customer'
And we are not dependent on China. Those are the statements we make. That's all we are going to say about it. Everything else is proprietary to the company.
backsC27C35 - S16primaryother2026-07-02Oracle Corporation, comments to the New Mexico Environment Department urging approval of the updated Project Jupiter air permit application Oracle Corporationthe counterparty's own statement that Project Jupiter, where Oracle is the tenant, will be powered by Bloom fuel cells running on natural gas, under an air permit application then pending
The project will now use Bloom Energy fuel cells to power the AI data center campus, replacing the previously proposed gas turbines and diesel generators with a lower-impact fuel cell microgrid.
backsC28 - S17tertiarynews2026-09-24coverage of Bloomberg's report that Oracle sent a force majeure notice to Project Jupiter's developer, and of the gas-pipeline delay behind it
The pipeline has been delayed from its original August 2026 target to February 2027, following repeated denials by the New Mexico State Land Office for right-of-way permits across state trust lands.
backsC28 - S18secondarynews2026-08the pipeline's revised in-service date as filed with FERC by Transwestern, and the land commissioner's two refusals of the route; FERC itself refuses this repository (403), so the filing was not read
Energy Transfer's Transwestern unit told federal regulators on Friday that the Green Chile Project now has an in-service date of 1 February 2027, against a previous target of 15 August.
backsC28 - S19secondarynews2026-09-21the case caption, court and docket number, and the unchanged class period and lead-plaintiff deadline
Captioned Nevins v. Bloom Energy Corporation, No. 26-cv-07944 (N.D. Cal.), the Bloom Energy class action lawsuit charges Bloom Energy and certain of Bloom Energy's executives with violations of the Securities Exchange Act of 1934.
backsC12 - S20primarycompany filing2026-09-18the transactions since the thesis, with the Rule 10b5-1 box as filed on each sale (read by scripts/insider_fetch.py)backsC17
- S21secondaryvendor data2026-09-24Portfolio-management research pack, refreshed at the 2026-09-24 review (market tier at the 2026-09-23 close)content/BE/research-pack.jsonprice, technicals, volatility bucket, beta, trailing multiples against Bloom's own history, insider roster and analyst consensus as of the 2026-09-23 closebacksC17C19
- S22secondaryvendor data2026-09-26Portfolio-management research pack, refreshed at the 2026-09-26 review (market tier at the 2026-09-25 close)content/BE/research-pack.jsonprice, technicals, volatility bucket, trailing multiples against Bloom's own history (the vendor's trailing twelve months now running through the June 2026 quarter) as of the 2026-09-25 close; the filed quarterly series, which still ends at the March 2026 quarterbacksC15C18
- S23primarygovernment data2024-12-20US Department of Energy press release - DOE Releases New Report Evaluating Increase in Electricity Demand from Data Centers (LBNL 2024 United States Data Center Energy Usage Report) US Department of Energy / Lawrence Berkeley National Laboratorythe government laboratory's range for US data-centre electricity use in 2028, the independent bound on how much new load there is to power at all; read through ir_fetch
The report finds that data centers consumed about 4.4% of total U.S. electricity in 2023 and are expected to consume approximately 6.7 to 12% of total U.S. electricity by 2028. The report indicates that total data center electricity usage climbed from 58 TWh in 2014 to 176 TWh in 2023 and estimates an increase between 325 to 580 TWh by 2028.
backsC29 - S24primaryindustry data2026-06-25SemiAnalysis - US Grid Constraints: Towards 40GW+ of Behind-The-Meter Datacenter by 2028? SemiAnalysisan independent research firm's forecast of US data-centre additions (+21GW in 2026 to +84GW by 2030), the behind-the-meter share from 2028 and a BTM equipment market above 50GW a year by 2029, naming Bloom among the new entrants that overcame turbine constraints; the free preview of a paid report, read through ir_fetch
Our research suggests that BTM will power well over half of new US datacenters in 2028+ , and the Total Addressable Market (TAM) for DC BTM equipment to cross 50GW/ year by 2029. ... Faster technologies like Fuel Cells and RICE are being aggressively secured directly by datacenter operators for onsite generation.
backsC30 - S25secondaryanalyst2026-05-14RBC Capital Markets insight - Natural gas powers the data center boom RBC Capital Marketsa bank's tally, from Bloomberg and company reports, of announced, ordered and under-construction behind-the-meter gas capacity for US data centres and its split by technology, in which fuel cells are not a category; read through ir_fetch
As of 2026, data center developers have announced approximately 101 gigawatts of behind-the-meter natural gas generation capacity in the United States. While some of these projects may be speculative, over 57 gigawatts have already publicly disclosed equipment orders, and approximately 7 gigawatts are already under construction. ... The left vertical bar chart categorizes capacity by technology, showing Gas turbine as the dominant type at nearly 70 GW, followed by Engine at approximately 18 GW, Unknown at 13 GW, and Mixed use at 3 GW.
backsC31 - S26primarycompany filing2026-07-22the incumbent turbine maker's gas equipment backlog and slot reservations, its annual turbine output now and its expectation for year-end contractsbacksC32
- S27primarycompany filing2026-01-08American Electric Power Form 8-K, Item 8.01, date of report 2026-01-04 (fuel cell purchase agreement and Wyoming offtake) American Electric Powerthe buyer's own filing of its $2.65 billion unconditional purchase of most of its 900 MW option on Bloom's solid oxide fuel cells, the only filed price for a block of Bloom capacitybacksC33
- S28primarycompany filing2026-05-20the customer's own filing of its May 14, 2026 capacity agreement with Bloom: three ten-year phases, about 328 MW installed and 250 MW guaranteed, up to $2.6 billion of service feesbacksC33
- S29primarycompany release2026-04-13the size of the Oracle master agreement and the part of it contracted; no Bloom filing states either figure (the April 13 8-K covers only the warrant); read through ir_fetch
Under a master services agreement, Oracle intends to procure up to 2.8 gigawatts (GW) of Bloom’s fuel cell systems. As part of this agreement, an initial 1.2 GW of capacity has been contracted, with deployment underway and continuing into next year.
backsC33 - S30secondaryvendor data2026-10-01FMP key-metrics-ttm and analyst-estimates (annual) for BE, GEV, VRT, CAT and SIEGY, fetched 2026-10-01 through scripts/api/fmp.pytrailing enterprise value to sales for the listed makers of the equipment Bloom displaces or sits beside, the same enterprise value over consensus 2029 revenue, and Bloom's own consensus revenue path to 2029backsC36
- S31primarycompany filing2026-04-08Bloom Energy Corporation proxy statement (DEF 14A), filed April 8, 2026, for the 2026 annual meetingbeneficial ownership as of February 27, 2026, the executive officers and their start dates, the 2025 annual cash incentive and its 200% payout, the PSU metrics for 2025-2027 and the CEO's December 2024 equity package, the CFO appointment effective April 13, 2026backsC39C40C41
- S32primarycompany filing2026-10-04Bloom Energy XBRL company facts, 2020 to the four quarters to June 2026, frozen as content/BE/financials.jsoncontent/BE/financials.jsonrevenue, gross profit, research and development (ResearchAndDevelopmentExpense, the same line the income statement labels Research and development), sales and marketing, general and administrative, capital spending, operating income and diluted share counts per yearbacksC43
- S33secondaryvendor data2026-10-04Earnings calendar saved by the daily routine, content/calendar/2026-10-04.json (FMP earnings, per symbol)content/calendar/2026-10-04.jsonthe next print date, 2026-10-27, with the consensus revenue and earnings per share for itread, but no claim rests on it
- S34primarycompany filing2026-10-22Bloom Energy Reports Record Third Quarter 2026 Financial Results and Raises Full Year 2026 Guidance (8-K Ex-99.1) Bloom EnergyQ3 2026 revenue, margins, operating income, per-share figures, balance sheet and the raised 2026 guidance; the release does not mention China or scandiumbacksC04C09C13C44
- S35primarycompany filing2026-10-22the three- and nine-month customer concentration, receivables concentration, nine-month cash flow statement, share count, litigation note and supply-chain risk factor; the word scandium does not appearbacksC04C06C07C09C12C13C14C15C27C34C44
- S36primarytranscript2026-10-22Bloom Energy Q3 2026 earnings call transcript (FMP), held 2026-10-22the backlog and pipeline description behind the guidance, the Fremont expansion answer and the free-cash-flow answer; no analyst asked about scandium or China
Our customer pipeline continues to grow and our backlog is continuing to expand faster than our revenue.
backsC04C07C09C27C34