Bloom is a genuinely inflecting business with real operating leverage and a product the AI build-out needs. It is also a company whose headline backlog figure does not appear anywhere in its audited financial statements — and that number is the anchor under every valuation model written on it, including ours.
The author holds no position in BE as at the date of publication, and held none while the research was produced. No position will be taken for 72 hours after publication.
The author has received no compensation from Bloom Energy or any affiliate, has no relationship with Hunterbrook Capital or any other party publishing on this company, holds no position exceeding 0.5% of the issuer’s share capital, and has no business relationship with the company. This report was not shown to the issuer before publication.
The backlog is not a GAAP number. Bloom does not disclose it in its filings. The roughly $20bn figure that anchors every model on this company is unaudited and undefined, and the company has not reconciled it to its binding obligations.
There is a live supply-chain question. Scandium is not a peripheral input — it is what makes the cell work. Management’s repeated denials of China dependence have been publicly disputed and not yet resolved with evidence.
Today’s price is a stack of best cases. Reaching $180.70 requires the bull growth path and the bull margin path and a discount rate well below what the stock’s own volatility implies. Each alone is defensible. All at once is not a valuation.
Bloom and the financial press routinely reference a backlog of approximately $20bn. Hunterbrook Capital, in a report published on 8 July 2026, asserts that audited binding remaining performance obligations are roughly $492m. The bar below is drawn to scale.
Some gap is entirely legitimate. Under ASC 606, remaining performance obligations properly exclude contracts carrying cancellation rights, variable consideration, and certain framework agreements. Firms across this sector show a gap. A gap of roughly forty times is not ordinary, and it has not been explained.
We are not asserting the backlog is fabricated. We are making a narrower and more defensible point: the number is unaudited, undefined, absent from the filings, and load-bearing for every valuation on the street. That is a disclosure problem regardless of who turns out to be right.
Bloom now generates cash, so we value it on unlevered free cash flow. The inputs below are deliberately generous: the base case assumes revenue compounds at 38% for five years and terminal EBITDA margin reaches 23%, both well above anything the company has yet delivered over a full cycle. Move them and see what your own view is worth.
Terminal value is 75% of enterprise value. On a company with five years of operating history at this scale, treat that as a warning label.
Cash tax held at 10%, reflecting the loss carryforward position. Net debt held at $0.5bn. Dilution is a slider rather than a constant because share count rose materially over the trailing year. All three are derived estimates and should be reconciled against the Q2 10-Q.
Our published ranges are bear $30–55, base $95–125, bull $185–240. At $180.70 the stock trades at the bottom edge of the bull case — meaning the market is already paying for the outcome in which everything works.
Reaching $180.70 requires simultaneously the bull growth path, the bull margin path, a discount rate roughly 250 basis points below what the stock’s own volatility implies, and a generous exit multiple. Each of those assumptions is individually defensible. All four at once is not a valuation — it is a stack of best cases.
Bloom makes solid-oxide fuel cells: modular units that convert natural gas, biogas or hydrogen into electricity on site, without combustion. The pitch to a data centre operator is simple and currently very powerful — bypass the grid entirely, get always-on power at your own facility, and stop waiting years for an interconnection queue.
AI capital expenditure turned a niche industrial product into a supply-chain bottleneck almost overnight, and the financial inflection is genuine. This is not a story stock with no revenue; it is a real business with real operating leverage that arrived suddenly.
Two structural caveats belong next to that. The product runs primarily on natural gas, which makes it vulnerable to carbon-intensity policy in California and the EU. And the window in which Bloom is the only viable grid-bypass option is finite — hyperscalers have also committed to small modular reactors, grid-scale storage and renewables. That window is probably two to five years wide, which is shorter than the terminal value in any DCF written on this company.
Backlog is not a GAAP concept, and Bloom does not formally disclose it in its filings. That alone would be unremarkable — many companies discuss backlog in commentary. What makes it material here is that the figure is doing structural work: it is the primary justification offered for the multiple.
| Position | Detail |
|---|---|
| The claim | Approximately $20bn of backlog, referenced by the company and repeated widely in press coverage. |
| The challenge | Hunterbrook Capital, 8 July 2026, asserts audited binding remaining performance obligations of roughly $492m — a gap of about forty times. |
| The company’s response | An 8-K on 9 July reaffirmed the integrity of the audited financial statements, but did not reconcile backlog to RPO. |
| On the Q2 call | Management asserted that backlog is growing faster than revenue, without quantifying either figure. |
| What would resolve it | One reconciliation table in the 10-Q, separating firm purchase orders, master agreements with committed commercial operation dates, and framework capacity. The company has had three weeks to produce one. |
We take no view on the allegations themselves and have not independently verified either figure. Our point is narrower and does not depend on who is right: a number that is unaudited, undefined and absent from the filings should not be carrying a valuation. A company that can reconcile it in one table, and has not, is choosing the ambiguity.
Scandium oxide is the dopant in Bloom’s ceramic electrolyte. It is not a peripheral component — it is the material that makes the cell commercially viable. China accounts for roughly 80–90% of refined scandium chemical production and effectively all metallised scandium for advanced applications, and added scandium to its national-security export licence list in April 2025.
Management has stated on at least five occasions since February 2025 that Bloom has no China supply chain and is not dependent on China for scandium. Hunterbrook alleges four China-linked trade routes reaching Bloom indirectly, and cites a representative of a Chinese scandium producer identifying Bloom as its largest customer. It further models that a 5 GW production target would require roughly 220 tonnes of scandium oxide annually, against estimated global production of roughly 240 tonnes.
Again we take no view on the allegations and have verified none of them independently. But the arithmetic is checkable and it is uncomfortable regardless of sourcing: if the capacity target is real, Bloom needs close to the entire world’s annual scandium output. Either the target moves, the supply expands, or the input changes. That is a question about physics and industrial capacity, not about anyone’s credibility.
Five disclosures would move this materially, and four of them are in the same document.
The remaining performance obligations line in the Q2 10-Q is the single most decision-relevant number in the filing. A reconciliation to backlog resolves the largest open question on the name.
Whether Q3 related-party revenue re-concentrates the way Q1’s did. Roughly half of Q1 revenue came from a single related-party relationship.
$306.5m at 30 June against zero at 31 December. Its size, its amortisation schedule, and where it lands in the P&L all matter to reported margin.
Grew materially faster than revenue. On a product with a multi-decade service life that is worth watching rather than assuming benign.
On backlog, on manufacturing capacity, or on scandium sourcing. Management has stopped quantifying. Continued silence is itself information.
A clean RPO reconciliation, a named second anchor customer, or documented non-China scandium supply at scale would each remove a discount we are currently applying.
We rate BE overvalued. That is a statement about price against our modelled range, not a claim that the business is unsound — the operating inflection here is real and we say so throughout. But at $180.70 against a $95–125 base case, the shares already pay for an outcome in which every open question resolves favourably.
Bloom builds refrigerator-sized boxes that make electricity from natural gas without burning it. You put them next to a building and it makes its own power instead of drawing from the grid. AI data centres need enormous amounts of electricity and cannot wait years for grid connections, so demand has exploded.
Backlog is meant to be the value of orders a company has won but not yet delivered. Investors use it to guess future revenue. The catch is that there is no legal definition of the word, and companies are not required to publish it or have it audited. Bloom talks about roughly $20 billion. The figure that is audited — the one covering firm, binding obligations — has been reported as roughly $492 million by a research firm that published on the company. Those two numbers are forty times apart.
Not necessarily. Some gap is normal, because backlog often includes agreements a customer could still walk away from. But a gap this large should have an explanation, and the company has not given one despite having several weeks to do so.
This is a real business that is genuinely growing fast. The question is not whether it works — it is whether the share price already assumes everything goes right. Our arithmetic says a fair price is somewhere around $95 to $125. It currently trades near $181. You would be buying on the assumption that the optimistic case is the correct one.