Revenue is up 138%, operating margin is 23%, and guidance has been raised three times — yet operating cash flow turned negative and the stock fell 11% on the print. The market is pricing one question: is the ~$20B backlog an annuity of AI power demand, or a set of options on data centers that can slip? Six analyst lenses, three scenarios, four time horizons.
Gray line = month-end closes with the dated June 2026 high and December 2025 low; colored paths = synthesized scenario midpoints forward, probability-weighted (bear 30% · base 45% · bull 25%, bear-tilted because of project-slippage and AI-sentiment risk). Log price axis, mid-year marks. Clay open circle = Street 12-month consensus ≈ $285 (range $176–$351, 29 analysts, Buy). Month-end closes are from a public price-history table, so intramonth swings are smoothed except the two dated markers.
Those probabilities are a judgment call — so make them yours. Drag to set how likely the bear and bull cases are (base takes the remainder); the blended target, the dotted line on the chart, and the prob-weighted row of the scenario cards update live.
The same Q3 print supports targets from $86 to $330 depending on which framework you trust. Each lens is a synthesized perspective with its own 12-month target; the panel mean is $203 against a Street consensus near $285.
Three raises in 2026, operating income in Q3 alone above all of 2025, and Morgan Stanley now models 2.1 GW shipped in 2027 and 3.7 GW in 2028. Capacity is a dial, and demand is turning it up.
A $57B market cap against $42M of year-to-date free cash flow is not a value stock. Capex has already run four times all of 2025 and contract assets are absorbing cash. Pay for cash, not for backlog.
Oracle and AEP timelines have already moved, the backlog has not been updated since February, and short sellers question its quality. Orders that can slip can also be cancelled.
Sub-90-day deployment, on-site solid-oxide units with carbon-capture readiness, and a 700 MW AEP award for Google make Bloom the one supplier that gets power on when the grid cannot.
Below the 200-day average, 45% off the June high, and a 15%-a-day-type volatility profile (beta near 3.8). A re-test of the pre-print $218 close is the base technical case.
Surging Treasury yields and a faltering AI-capex tape hit long-duration power names first. Fundamentals are intact, but the multiple belongs to the rate cycle for now.
What the sell-side expects over the next year. Bars are sorted low to high; the dashed line is today’s $193.30 — every target shown sits above it, even after Friday’s cuts.
Post-print moves (Oct 23): Morgan Stanley raised to $335, JPMorgan cut to $292, BTIG to $250, Jefferies to $242, Roth to $230. RBC ($335), UBS ($350, Oct 7) and Mizuho ($351, Sep 14) are the latest levels found and may predate the print. Consensus differs by aggregator ($285 to $292); no Sell-rated firm was found among those sampled.
Every price is next-twelve-month EPS at that date times an exit multiple. Returns are vs. today’s $193.30. These are illustrative frameworks, not forecasts; five-year outcomes hinge on whether backlog converts to cash.
The income statement is the voltage; the cash statement is the current. The bull and bear theses both live in the gap between the sky bar and the olive bar.
Revenue: $1.33B / $1.47B / $2.02B, then the $4.4–4.6B guide midpoint ($4.5B) for 2026E. FCF swung from −$456M (2023) to +$33M and +$57M, and is +$42M year-to-date after a −$181M Q3. Capex ($84M / $59M / $57M) is $231M through Q3, four times all of 2025; the 2026E capex ($0.38B) and FCF ($0.15B) bars are desk estimates (faded) and total debt is held at roughly $3.0B. Debt includes leases; with $2.5B of cash, net debt is only about $0.5B, funded largely by the 2025 convertible issuance. The Q3 cash gap is mostly contract assets (−$430M).
Each target is an EPS estimate times an exit multiple. This is the earnings ladder the scenarios are built on.
Non-GAAP EPS. 2026E $3.27 sits just above the raised $3.10–$3.40 guide; 2027E $5.15 and 2028E $8.17 are Street averages (TIKR, Oct 23, 2026). 2029E–2032E are desk extrapolations beyond Street coverage, fading growth from ~27% to ~11%. At $193.30 the stock is about 37x 2027E and 24x 2028E; GAAP EPS ($0.78 in Q3) is lower, and trailing P/E on GAAP is above 130x.
Q3 FY26, year-over-year — read these against a stock that just lost 11% on the print.
Bloom reports no AI-segment split, so the frontier color marks the profit lines that show operating leverage: operating expenses grew 26% while revenue grew 138%. Service revenue ($67.6M, +15%) and installation ($84.9M, +29%) are the slow core; product revenue crossed $1B for the first time. Profit growth is off a small year-ago base ($0.15 EPS, $46M operating income).
The valuation argument compresses into one disagreement: does Bloom’s backlog convert into cash on schedule, or does it slip like Oracle and AEP?
Where each risk sits, not just how big it is. The hot upper-right corner is likely and high-impact; the project-specific risks cluster one row down in “possible.”
A stock with a beta near 4 trades on AI-capex sentiment; 78x to 40x forward P/E in three weeks shows how fast the multiple can move.
Oracle reportedly issued a force-majeure notice on a New Mexico site; AEP moved its target from 2028 to no later than 2030.
The roughly $20B backlog was last updated in February and is partly backstopped by customers who can walk away.
Oracle has run negative free cash flow; a financing squeeze at one anchor customer could cancel gigawatts of orders at once.
Q3 contract assets rose about $430M, turning operating cash flow to −$27M; management calls it not a trend.
Fremont is being expanded toward 5 GW a year; if demand dials down, fixed cost lands on margins.
Turbines, engines and batteries compete for the same on-site megawatts; GE Vernova trades near the same forward multiple.
New lines and expedited deliveries held gross margin near 34% despite higher volume; BTIG flagged near-term drag.
A class action over China-linked scandium sourcing disclosures was filed after a July 2026 short-seller report; allegations are unproven.
Hover the dotted terms in the metrics, or scan the desk’s working definitions here.