Coverage update · AI power · Bloom Energy · Accumulate, target about $198

We Cut Bloom’s Target and Raised the Rating, Because the Price Fell Faster

Posted November 16, 2026Author Brandon LeonData as of November 16 closeRead 13 min

The verdict

We move Bloom Energy to ACCUMULATE from Hold and lower our 12- to 18-month price target to about $198 from $235, 11.8% above the November 16 close of $177.28: the 2027 estimate held at $5.32 while the multiple fell to 33.3x, below the band our base case pays, so the price has now fallen further than our number.

The October 23 note put the $235 target under review and promised the re-struck figure within two weeks, with the weights shown. We built it on October 26, wrote down the rule that would set its weights before the hyperscalers reported, and then missed our own date: the fill did not run on October 30, nothing published, and by the time we came back to it the shares had fallen through the rating line and NTT DATA had signed what Bloom calls its largest international project to date. The promise was dated and we missed it, so it stands on our record as a miss.

The number is about $198. Bloom closed at $177.28 on November 16, down 18.6% since the October 22 report and down 36.1% from the $277.22 at which we initiated, against an S&P 500 fund down 3.4% over the same ten weeks.1 At that price the shares trade at 33.3 times a 2027 estimate that has not moved since the report, against 56 times on September 8 and 38 times on October 26. The earnings held and the price of a dollar of them kept falling, so this re-strike takes the Street’s earnings whole and re-prices the multiple: the target falls 15.6% from $235, and all of it is the multiple.

1. The estimate held, an order arrived and the price fell

The Street has not moved since the report. The fiscal 2027 consensus is $5.32 of earnings on $7.09 billion of revenue from 27 analysts, against $5.31 on October 26 and $4.90 a month ago, before the report; fiscal 2026 is $3.27 against guidance of $3.10 to $3.40, and the 2027 range still runs from $2.96 to $7.49.2 The target changes since the report have gone both ways, with J.P. Morgan cutting to $292, BTIG to $250 and Jefferies to $242 on October 23 while Morgan Stanley raised to $335, and Barclays moving to Overweight on November 3 with a target cut to $289; one aggregator puts the mean target at $284 from 30 analysts, 60.2% above the price and well above ours.3

Chart 1: The estimate stopped moving

Both estimates moved once, at the October 22 report, and have held since. The price fell 19% after it.

Consensus EPS: FY2026 (navy) and FY2027 (green), $ per share1234562.714.8990 days ago2.694.9560 days ago2.714.9030 days ago3.275.327 days ago3.275.32Nov 16Q3 report, Oct 22Both estimates moved once, at the report, and not since: FY2027 $4.90 to $5.32, FY2026 $2.71 to $3.27.FY2027 revenue consensus $7.09B (27 analysts), 56% above FY2026. Yahoo Finance estimate trend read November 16, 2026; the Street's figures, not ours.

Consensus EPS per the Yahoo Finance estimate trend, read November 16, 2026. These are the Street’s figures, used because the company gives no 2027 outlook.2

The news was an agreement with NTT DATA. On November 5 NTT DATA said it would deploy “900 megawatts (MW) of onsite power” from Bloom across its global data centers “as primary power,” with “initial shipments starting in 2027,” and the two companies said the relationship is “designed to expand across NTT DATA’s Global Data Center portfolio and reach gigawatt scale.”4 Bloom called it “its largest international project to date.” Neither company gave a price, the countries or a split by site, the release calls it only “an agreement,” and Bloom filed no 8-K for it. For scale, AEP paid about $2.65 billion in January for most of a 900-megawatt fuel-cell option, so an order of NTT’s size is plausibly worth billions over several years, but that is our inference from someone else’s contract, and we put no number from it into the tree.5

The same day Google said its first Wyoming data center, a $17 billion project, will draw on a 702-megawatt fuel-cell plant that AEP announced on October 14 and will run under a 20-year contract, and that “the fuel cells for this project will be made in America by Bloom Energy.”5 That names an end customer for capacity AEP had already bought, on our reading of AEP’s January filing, which does not name Bloom or the customer, so we count it as confirmation of an order already in the numbers rather than new volume. AEP expects the plant to be fully operating by the end of 2028.

The shares fell 16.4% that day, to $167.52, while the Nasdaq-100 tracking fund fell 6.5% on fears that the midterm results would slow data-center construction, and Bloom’s own good news did not stop it.1 The fear had a basis: Ballotpedia counted voters favoring limits on data-center development in 40 of the 41 local ballot contests on November 3.6 The shares recovered to $209.47 by November 11 on a sector rebound, with no company announcement, and closed at $177.28 on November 16. A stock that moves $42 in four sessions on an unchanged estimate is telling us where the multiple sits, not what the business earns.

The October 22 10-Q facts that set the multiple in our October build still stand. Two customers were 43% and 26% of third-quarter revenue, the second a related party; related-party revenue, which runs through the Brookfield financing structures Bloom part-owns, was $321.6 million, about a quarter of the total; and contract assets, the revenue recognised ahead of billing, reached $858.6 million including the non-current portion, three and a half times the year-end balance.7 On the call management declined to frame 2027, and the chief financial officer called the working-capital build “not the beginning of a trend,” a claim the fourth quarter will test.8 Oracle, the anchor customer, closed at $119.09 on November 16, down from $126.06 on October 26, after its credit-default swap reached a record in mid-October; none of that changes the order, and it stays among the tests below.9

2. The multiple has fallen below our base band

Our base leg pays 34 to 40 times the 2027 estimate, which is what the market paid in the days after the report. It now pays 33.3 times, below the bottom of that band, after 32 times at the November 5 close. The power-equipment group trades at a median of 23.8 times calendar 2027, with GE Vernova at 29.4, Quanta 27.0, Eaton 20.5 and Vertiv 19.1, and the generators lower still at Constellation 20.7 and Vistra 12.8.10 Bloom’s premium to that median is now about 40%, from about half in late October and about double in August, which is still a premium for roughly 56% growth whose earnings are not yet converting to cash.

Chart 2: What the market pays for a dollar of 2027 earnings

Fifty-six times on September 8, thirty-eight on October 26 and 33 now, below the base band, on an estimate that went up.

Price divided by the FY2027 consensus EPS in force each day (times)30x35x40x45x50x55x60x65xSeptember base leg: 50-55xbase leg since October 26: 34-40xSep 8: 56xOct 6: 60xOct 26: 38xNov 5: 32xNov 16: 33.3xSepOctNovThe estimate rose 8% at the report and has held; the multiple fell from 56x to 33x, below the base leg's band.Consensus in force: $4.92 (Sept 1 pull) to Sept 30; $4.93 (Oct 16 page) to Oct 22; $5.31 from Oct 23; $5.32 from the Nov 9 reading. Closes, Yahoo Finance.

Daily closes divided by the fiscal 2027 consensus in force each day; the estimate jumped on October 23, which is why the line falls on a day the price rose.1

We leave the base band where the October build put it. Marking a base leg to the tape is how the September target went wrong, and the case for 34 to 40 times rested on the filing rather than the price: earnings outran cash and two customers were 69% of revenue, and nothing since has changed either fact. A provider beta of 3.78 implies a 24% cost of equity at a 5% risk-free rate, against the 13% our cash-flow model uses, and a discount rate like that is the bill for a multiple like this one.10 What has changed is that the price now sits below the band, which is the case the rating bands in section 4 were written for.

3. The cash-flow anchor has not moved, and the gap to it has narrowed

The discounted cash-flow re-run of October 26 stands: about $77 a share in the base case, $121 in the bull and $37 in the bear, $76 weighted, against $66 in the August model, with fiscal 2027 revenue at the consensus and the same fade, margin, tax, stock-compensation and discount-rate assumptions as the initiation.11 Nothing since has changed its inputs, because the revenue consensus is $7.09 billion against $7.05 billion then, and the November agreement ships from 2027 at an undisclosed price. The price is now 2.3 times the base value, against 2.6 times on October 26 and about 4.3 times in September, so the regime premium we described in August, about $213 a share at the initiation, is about $100 now. It has narrowed because the price fell, not because the cash got closer.

Chart 3: The DCF, August and October

The base value rose by a quarter in October. The price is 2.3 times it.

The cash-flow anchor: $ per share, August model (pale) and October re-run (solid)$50$100$150$200Bear$27$37Base$64$77Bull$114$121Weighted 25/45/30$66$76Nov 16 close $177, 2.3 times the base valueSame model as the initiation (13% base discount rate, 12.5% bull, 13.5% bear; stock compensation as a cost; converts as equity), FY2027 revenue at the consensus.The base value rose a quarter in October. The price is still 2.3 times it, from 2.6 times on October 26.

Per-share values from our model; the August figures per the initiation, the October figures per the October 26 re-run. Weights 25% bull, 45% base, 30% bear in both.11

4. The rule set the weights, and the band sets the rating

On October 26, before the hyperscalers reported, we wrote down a rule for reading them: count the 2027 capital-spending cuts across Microsoft, Alphabet and Amazon, with Meta’s October 21 report counted as neither, where a cut is guiding 2027 below an earlier stated figure or walking back “increase” to “flat,” “moderate,” “discipline” or “digestion,” and set the weights by the count.12 The count was no cuts, and two of the three framed 2027 as a significant increase: Alphabet said it expects capital spending “to increase significantly in 2027,” and Amazon that it expects “to spend meaningfully more in 2027,” while Microsoft said its fiscal 2027 second-half spending would “accelerate” without a calendar figure.12 That is the first row of the rule, as written:

Capex-week result (rule of October 26)WeightsTarget
No cuts, and two or more frame 2027 as a significant increase (applies)30 / 45 / 25about $198
No cuts, neutral framing25 / 45 / 30about $189
Exactly one cut20 / 45 / 35about $181
Two or more cuts15 / 40 / 45about $167

We considered moving the weights again for November 5 and did not. The NTT DATA agreement is evidence for the bull leg, which asks for a large order beyond Oracle, and the ballot results are evidence for the bear, because a site that cannot be built is an order that cannot ship; neither changes a 2027 number yet, since the agreement carries no price and ships from 2027 and the ballots bear on where data centers get built rather than on orders already placed, and moving weights by judgment after a rule has set them is the habit the rule exists to stop. Both become tests in section 6.

CasePWhat has to happenFY2027 EPS and multiple12–18 month range
Bull30%The fourth-quarter call gives a 2027 framework at or above the consensus; contract assets turn to cash; Fremont reaches 2 gigawatts and NTT DATA’s first shipments land on schedule; the multiple goes back toward the August premium$6.20 × 42–48×$260–298
Base45%Fiscal 2027 lands near the $5.32 consensus; cash conversion mixed; the multiple settles in the band the market paid after the report$5.32 × 34–40×$181–213
Bear25%One of: a 2027 framework implying growth under 30%; an Oracle slip or cut under credit stress; a conversion gap that forces an equity-linked raise; named sites stalled by local opposition. No premium to the equipment peers$4.00 × 24–28×$96–112

Weighted, 0.30 × $279 + 0.45 × $196.8 + 0.25 × $104 = $198.28, about $198, 15.6% below the published $235 and 11.8% above the November 16 close.

Our rating bands put a twelve-to-eighteen-month value from 10% below to 5% above the price at Hold, from 5% to 15% above at Accumulate, more than 15% above at Buy and more than 10% below at Reduce.13 At $177.28 the target sits 11.8% above the price, so the rating moves to Accumulate from Hold, by the band and without a vote. The October rule named only Hold and Reduce because Bloom then traded above every row’s target; the price has since fallen through the base range, and on this target the rating would be Accumulate at closes from about $172 to $189, Hold from $189 to $220, and Buy below $172.

Chart 4: The tree, struck twice

Every leg is built on a higher earnings number and a lower multiple. The whole move is the multiple.

The tree, struck twice: September 9 on a $4.92 estimate and Nov 16 on $5.32, $ per share$100$150$200$250$300$350September 9 (price $277)25%290-34045%245-27030%125-150$235Nov 16 (price $177)30%260-29845%181-21325%96-112about $198Green bull, navy base, rust bear; gold = weighted midpoint; dotted = the close; weights 30/45/25 per the capex-week rule. All estimates.

September ranges per the initiation; the current ranges per the table above, with the weights the October 26 rule assigned.13

5. Scoring the September view, and our missed date

The $235 target published at $277.22 was reached from above inside six weeks, so the direction and the level were right, and the price has since fallen a further 24.6% below it. The reasoning was half right. The initiation said the multiple would hold through the doubling year, and it has fallen from 56 to 33 times while the doubling arrived on schedule. The October 23 decision not to re-strike on an after-hours print was right, and the October 26 decision to mark the target to a lower multiple was right in direction, since the market kept cutting. Missing our own publication date was a process failure of the kind this research has named before, and it is recorded as one.

Chart 5: Since the initiation

Reached the target from above, fell through the base range, and now sits inside the Accumulate band for the new target.

BE since August: daily closes against the targets and the rating bands ($)150175200225250275300Accumulate on a $198 target: closes from $172 to $189target $235 (Sept 9 to this note)target about $198 (this note)Sep 8 initiation $277Oct 6 peak $296Q3 report $218Nov 5: NTT DATA agreement; QQQ -6.5%; $168AugSepOctNovThe $235 target was reached from above inside six weeks; the price has since fallen through the base range into the Accumulate band.Daily closes, Yahoo Finance; last Nov 16 $177.28 (rust). A target 5% to 15% above the price is Accumulate; 10% below to 5% above is Hold.

Daily closes from August 3 against the targets as published and the band in which the new target makes the rating Accumulate.1

For readers who hold the shares the record is simpler and worse: a Hold at $277 on a position we owned is down 36.1% while the index fund is down 3.4%. This note does not improve that score. It makes the next one gradable.

6. What has to be true from here

The thesis-break conditions from September are unchanged and held as written: a Fremont ramp slipping two or more quarters, the Oracle order materially cut, a repeal of the 48E credit or disqualifying supply-chain guidance, the scandium allegation substantiated, or a first 2027 framework implying growth under 30%. With the consensus at 56% growth for 2027, a 40% framework would be about a 10% cut to the Street’s revenue and 30% about a 17% cut, so the bars stand and this is how to read them. Two tests are new.

TestSupports the targetBreaks it
Cash conversion, the number this note hangs onFourth-quarter operating cash flow positive and contract assets below the September 30 balanceContract assets higher again with operating cash flow negative: the CFO’s one-quarter claim fails and the bear leg governs
The 2027 framework, at the fourth-quarter reportAt or above the 56% consensus: the base case. Between 40% and 56%: an estimate cut, base toward bearUnder 30%: the September kill as written
The 2027 consensus, now $5.32Holds or rises after the reportA cut of 10% or more within 30 days of a report
The NTT DATA agreement (new)A value, a schedule or a backlog line for it in the 10-K or the fourth-quarter report, and first shipments in 2027 as statedResized, pushed past 2027, or described as non-binding
Siting (new)The named Bloom projects, Oracle’s and the AEP Wyoming plant among them, keep their announced schedulesA named Bloom site delayed two or more quarters from its announced schedule, or cancelled, because of local opposition or a permitting loss
Oracle’s creditRating held; financing on ordinary terms; the 1.2 gigawatts delivered on scheduleA downgrade to junk or financing on visibly worse terms moves the weights toward the bear; a cut or delay of the order itself is the September kill
ConcentrationRelated-party revenue below a quarter of the total; the largest customer below 43%Both higher: the multiple we pay falls to the bear leg’s

The levels, restated. The September zones and bands, set around a $220 to $235 target, are retired with it. On this target the rating is Accumulate while the price sits roughly 5% to 15% below it and returns to Hold above about $189; above about $213, the top of the base range, the price begins to pay for the bull case.

This note changes a rating and a target, so under our trading policy it opens a no-trade window on Bloom from ten days before publication to five days after it.

What makes this note wrong

The base multiple is a judgment we have now set twice in ten weeks, once too high. The rating rests on one day’s close in a stock that has moved between $167 and $209 in five sessions, so it can be right on the arithmetic and stale within a week, and we say so rather than smooth it. The NTT DATA agreement carries no price, and our view that it moves no 2027 number would be wrong if it proves large and early. The DCF is our model with our fade schedule, and a reader who believes the power shortage runs past 2030 should put little weight on it. And if the November ballot results turn into delays at named Bloom sites, the bear weight here is too low.

What it means: Bloom is earning what the Street expected after the report and has signed a 900-megawatt agreement since, and its shares are down 13.1% from our October build because the market keeps paying less for each dollar of that profit. We cut the target to about $198 from $235, all of it the multiple, and move the rating to Accumulate, because at $177.28 the price has fallen further than our number; the tests above say what would change that, starting with whether the cash shows up in the fourth quarter.

Sources

  1. Prices are daily closes from the Yahoo Finance chart API, pulled November 16, 2026 after the close: BE $277.22 on September 8, $295.78 on October 6, $217.77 on October 22, $203.92 on October 26, $200.40 on November 4, $167.52 on November 5, $209.47 on November 11 and $177.28 on November 16; SPY $765.96 on September 8 and $739.95 on November 16; QQQ $691.95 on November 4 and $647.05 on November 5. The November 5 context (post-midterm fears about data-center construction; the Nasdaq-100’s worst day since 2020) is per MarketBeat and BigGo Finance, secondary; the November 11 rebound per The Motley Fool and TradingKey, secondary, which found no company announcement. ↩
  2. Consensus per the Yahoo Finance estimate trend read November 16, 2026: fiscal 2026 EPS $3.27 (23 analysts; $2.71 thirty days earlier); fiscal 2027 EPS $5.32 (27 analysts; range $2.96 to $7.49; $4.90 thirty days earlier); fiscal 2027 revenue $7.09 billion, 56% above fiscal 2026. The October 26 figures ($5.31, $7.05 billion) are our read of the same page that day. ↩
  3. Rating and target actions per the stockanalysis.com ratings page, TIKR and Stocktwits, all secondary: October 23, J.P. Morgan $292 from $314 (Overweight), BTIG $250 from $295 (Buy), Jefferies $242 (Hold; the prior target differs by source), Morgan Stanley $335 from $310 (Overweight); November 3, Barclays to Overweight from Equal Weight with a target of $289 from $316; November 9, RBC Outperform, $335 reiterated. Mean target $284 from 30 analysts per TIKR, November 10. ↩
  4. Bloom Energy release, London and San Jose, November 5, 2026, and the same text on NTT DATA’s site. Bloom’s only 8-K since October 22 is the third-quarter filing of that date (EDGAR). ↩
  5. AEP Form 8-K of January 8, 2026: a November 2024 agreement for “100 MWs… with an option to purchase an additional 900 MWs,” the purchase on January 4, 2026 of “a substantial portion of its option” for “approximately $2.65 billion,” the facility “expected to be located near Cheyenne, Wyoming”; the filing names neither the supplier nor the customer. AEP, October 14, 2026: a “702-megawatt fuel cell generation project near Cheyenne, Wyoming” that “uses advanced fuel cell technology developed by Bloom Energy,” about $4 billion, a 20-year agreement with Google, fully operating by the end of 2028. Google, November 5, 2026. Linking the January purchase to the 702-megawatt plant is our inference. ↩
  6. Ballotpedia, November 6, 2026: “Voters favored limiting data center development in 40 of the 41 local ballot measure contests on Nov. 3.” ↩
  7. Bloom Energy Form 10-Q for the quarter ended September 30, 2026, filed October 22, 2026 (EDGAR): two customers “accounted for approximately 43% and 26% of our total revenue” for the three months, the second a related party; related-party revenue $321.6 million for the quarter; contract assets $792.8 million current and $65.8 million non-current at September 30 against $241.2 million at December 31; product and installation performance obligations $661.2 million, “We expect to recognize the associated revenue within the next 1 to 2 years.” Two directors sold shares on November 2 and 3 under 10b5-1 plans adopted in February and July (Form 4, EDGAR); we read nothing into pre-arranged sales. ↩
  8. Third-quarter 2026 earnings call, October 22, 2026, per third-party transcripts (The Motley Fool, stockanalysis): the chief financial officer, “that’s not the beginning of a trend” and “those differences even out over time”; the chief executive, “We are not going to provide a timeline.” ↩
  9. Oracle: an order of up to 2.8 gigawatts with an initial 1.2 gigawatts contracted for 2026 and 2027 per Bloomberg, April 13, 2026; the five-year credit-default swap at a record 288 basis points on October 15 per S&P Global Market Intelligence and LSEG data as reported by CNBC, October 19; S&P rating BBB− after a July downgrade. Secondary, except the closes (Yahoo Finance). ↩
  10. Comparables at calendar-2027 consensus earnings and November 16 closes, per Yahoo Finance estimate and price data: GEV 29.4, PWR 27.0, ETN 20.5, VRT 19.1, CEG 20.7, VST 12.8; the equipment median is of the first four. Beta 3.78 per the provider; the 24% cost of equity is 5% plus beta times a 5% premium, our convention. The August comparison (about twice the equipment group) is per the initiation, and late October per our October 26 read (median 25.6). ↩
  11. Our model, re-run October 26, 2026: fiscal 2026 revenue $4.5 billion (the guidance midpoint), fiscal 2027 $7.05 billion (consensus), then growth of 35%, 22%, 14% and 9% to fiscal 2031 and a fade to 4% by fiscal 2036; operating margin after stock compensation from 19% to 26%; $2.5 billion of cash; 324.3 million diluted shares; 13% discount rate and 3% terminal growth in the base case; the bull at 12.5% on 73% fiscal 2027 growth, the bear at 13.5% on 25%. The August model’s inputs are in the September 9 initiation, section 5. ↩
  12. The rule was written on October 26 and held in our internal negative-results ledger, which records each published thesis’s break conditions before the data arrives and grades them afterwards; it is private and its results are reported in the notes they concern. The count was read October 30 from Alphabet’s October 28 release and call, Amazon’s October 29 call and Microsoft’s October 28 call, quotations per third-party transcripts; Meta’s October 21 report gave no 2027 figure. ↩
  13. The rating bands are our standing definitions; the Hold band was first published in section 11 of our November 12 Nvidia report. The September tree per the initiation: bull $290 to 340 at 25%, base $245 to 270 at 45%, bear $125 to 150 at 30%, weighted $235.88. ↩

Next scheduled events: Oracle’s fiscal second-quarter report in mid-December (date unconfirmed); Bloom’s fourth-quarter and fiscal 2026 results, expected in late February 2027 by the company’s pattern (date not announced), which carry the first 2027 framework and the contract-asset turn; the fiscal 2026 10-K, which should restate the capacity language the June 10-Q dropped and may carry the NTT DATA agreement.

Methodology: company figures are quoted from Bloom’s own filings in dollars as reported; call quotations are from third-party transcripts and labelled as such. The target is a probability-weighted average of three twelve-to-eighteen-month scenario ranges, each a multiple applied to a stated fiscal 2027 earnings figure; the weights here were set by a rule written before the data, and the ranges are estimates. The discounted cash-flow figures are from our own model and are shown as a cross-check, not as the target. Tests are graded against their published wording.

Nothing on this page is investment advice. We work in scenarios and trip-wires, not predictions. See disclaimer.

Disclosures

Ownership. As of publication I hold 10 shares of Bloom Energy and no options on it, and I hold Alphabet class C shares. I hold no position in NTT DATA, AEP, Oracle, GE Vernova, Quanta Services, Eaton, Vertiv, Constellation Energy or Vistra, or in the SPY or QQQ funds.

AI assistance and conflicts. AI tools helped draft this note and compute its figures; the AI is Claude, made by Anthropic, whose product I use as a customer. Alphabet, whose Wyoming data center appears in section 1, sells Google Cloud TPU capacity to Anthropic (Google Cloud, October 23, 2025; April 6, 2026). We found no disclosed relationship between Bloom Energy and Anthropic.

Compensation and certification. bpleon research received no compensation from Bloom Energy or from any party with an interest in it. The views in this note are my own, and no compensation is tied to them.

Disclaimer. This note is research and opinion, published to all readers alike, and is not investment advice for any person or an offer to buy or sell any security. Prices are as of the 4:00 PM ET close on November 16, 2026 unless a date is given.