--- title: "BE: Blowout results vs. stock plunge — is beta fear behind the AI power frenzy?" type: "Topics" locale: "en" url: "https://longbridge.com/en/dolphin/post/44511717.md" description: "Deliveries rose and guidance was raised again, but growth is now cash-burning" datetime: "2026-10-23T03:41:23.000Z" locales: - [en](https://longbridge.com/en/dolphin/post/44511717.md) - [zh-CN](https://longbridge.com/zh-CN/dolphin/post/44511717.md) - [zh-HK](https://longbridge.com/zh-HK/dolphin/post/44511717.md) author: "[Dolphin Research](https://longbridge.com/en/dolphin.md)" generator: "portal-rs" --- # BE: Blowout results vs. stock plunge — is beta fear behind the AI power frenzy? **After hours on Oct 22, 2026, Bloom Energy reported Q3 2026 results. Despite concerns that Q2 revenue was overly reliant on a single customer (Nebius) and that Oracle's Project Jupiter gas supply delay could slow deliveries, Bloom delivered a clear beat and raised full-year guidance for the third time this year. Details below:** **1) Revenue beat, deliveries accelerated post mega-order:** Q3 total revenue was $1.24bn, up 138% YoY and ~15% above consensus. Product revenue (direct SOFC sales) reached $1.07bn, up 179% YoY. Assuming ASP of $3,000/kW, shipments were ~356MW, up ~16% QoQ. Q2 saw ~70% of revenue from one customer, raising fears of a Q3 air pocket, yet other data center clients' sites came online and deliveries more than filled the gap. Even without Jupiter contribution, revenue held up. Bloom delivered by leveraging its 'sign-to-ship' capability. Management cited examples: the first Oracle project in Utah reached power in under two months, and the Nebius project in New Jersey secured planning in three months, with the first block usable two months post-approval. For power-constrained data centers, speed itself drives orders, and once sites are ready, revenue is recognized quickly. **2) GPM edged up, pricing and cost-down offset scale-up costs:** GAAP GPM was 33.8%, up 40bps QoQ. This was mainly driven by higher product margins, which offset weaker margins in services and installation (down 480bps). **a. Product margin continued to expand:** Product GPM rose 140bps QoQ to ~37.9% (vs. street at 37.3%). Pricing stayed firm, and materials/manufacturing cost-downs flowed through, offsetting short-term cost upticks from new line ramp and expedited shipments. On pricing, management reiterated that Bloom sells 'time to power' value, not a commodity. With gas turbine prices doubling in recent years and lead times still elevated, SOFC pricing remains sticky. **b. Services margin dipped slightly:** Services GPM eased from 18.7% in Q2 to 17.8% this quarter (street at 18.0%). It still marked the sixth straight quarter of double-digit services margin. Services revenue grew with installed base, but higher product performance guarantees (up ~$8.3mn YoY) weighed on services revenue and margin. Longer stack life remains key to services margin gains, as replacement cycles improved from nine months in 2018 to ~five years, with durability issues unlikely to surface until 2028–2029. **3) Operating leverage surged, single-quarter profit exceeded last full year:** Revenue more than doubled while Opex rose 33%. R&D and infrastructure stayed largely fixed, and SG&A and services ops scaled via automation, with incremental costs mainly from compensation, including stock-based awards. Automation underpins cost control. The stack line employed ~600 people at 100MW annual output and still ~600 at 1GW, and management expects a similar headcount at 5GW. This implies limited Opex growth with future capacity expansion and room for margin improvement. OP growth hence outpaced revenue. OP reached $230mn vs. initial street est. of $170mn, and OPM rose another 100bps QoQ to 18.2%. **4) Cash flow turned negative, growth is now cash-consuming:** Operating cash flow turned slightly negative at -$20mn. With capex of -$150mn, FCF was about -$180mn, turning negative again. Working capital drove the swing. Product revenue is recognized on delivery and transfer of control, but invoicing and cash collection follow contractual milestones, so faster revenue growth builds contract assets and receivables, which consumed ~$600mn of cash this quarter. Inventory also rose by ~$170mn to support Q4 deliveries and capacity ramp. Capex accelerated to ~$150mn in the quarter, with 9M already 4x the full year 2025 level. Management said this single-quarter outflow is 'not the start of a trend', and 9M FCF remains positive. Dolphin Research believes this is normal in a high-growth phase as long as collections remain on track. Period-end cash was about $2.5bn, roughly flat vs. the start of the year, indicating no financing pressure and capacity expansion funded by internal cash flows. **5) Full-year guidance raised for the third time this year:** Full-year revenue guide was lifted from $3.9–4.2bn to $4.4–4.6bn (street ~ $4.1bn). At the midpoint of $4.5bn, that implies ~122% YoY growth, while Non-GAAP OP was raised from $800–900mn to $960mn–$1.03bn, and GPM remains around 34%. Implied Q4 revenue is $1.35–1.55bn, pointing to continued QoQ growth and well above the prior street est. of $1.23bn. Management said guidance is based on site-readiness assessments for orders in hand by customer, and probability-weighted demand from in-pipeline customers requiring near-term power, with year-end delivery timing determining where results land in the range. The market had assumed Jupiter delays would cap further raises, yet the midpoint increased by about $450mn. This indicates other customers will fill the gap. However, the company did not provide 2027 guidance. The framework for backlog growth will be clarified on the Q4 call, and management gave no capacity figure or timeline for the Fremont expansion, noting only that it prepares Bloom for post-5GW and that expansion will track demand. On profit, 9M Non-GAAP OP was about $660mn. Guidance implies Q4 of $300–370mn, with Non-GAAP OPM at ~22%–24%, roughly in line with Q3, suggesting Bloom is not sacrificing margins to chase revenue. **Dolphin Research view:** Overall, with Q2 revenue heavily tied to a single mega-order and Jupiter gas delays, the market expected a 'smooth handover and unchanged full-year guide' for Q3. Bloom instead delivered across the board: revenue, GPM, and OP all beat, and both revenue and profit guides were raised for the third time this year, implying further QoQ growth in Q4. The caveat is cash flow. Working capital consumption plus faster capex turned FCF negative for the quarter, and the company still has not disclosed backlog and capacity details or a 2027 outlook. We have emphasized that BE's investment focus has fundamentally shifted. It has moved from 'tech validation' to 'capacity ramp validation'. Q3 progress reinforced this view. On demand, another hyperscaler (Google) adopted Bloom's SOFC via a utility PPA, and new use cases across the military, utilities, and Intl markets are opening up. On supply, Bloom leased another Fremont building to prepare for beyond 5GW, lifting the capacity ceiling from 5–6GW to 10–12GW. **Bloom's core advantages vs. alternatives were reaffirmed this quarter:** **1) Time to Power: monetizing speed** Nebius signed a 328MW order in May and shifted to Bloom due to air permitting hurdles for reciprocating engines. Bloom delivered ~260MW in the quarter, some by air freight with higher expedite costs, while product GPM still reached 38%, indicating customers pay premium for speed. For a 1GW AI data center, Bloom's solution entails several hundred million dollars of power and equipment premium. Yet bringing capacity online a year early can add $12–24bn of compute revenue, with short-term leases potentially higher. **2) Avoiding NIMBY, moving in front of the meter** Bloom's non-combustion tech has low emissions, low noise, and near-zero water use (about 1.4mn gallons over a 1GW lifecycle vs. ~25bn gallons for wet-cooled gas CCGT). This eases permitting and community support, and SOFC is now formally in utility resource plans. a. **Google/AEP**: AEP will invest ~ $4bn to build the 702MW Cheyenne fuel cell project in Wyoming using Bloom paired with CCS, with a 20-year PPA with Google and COD by end-2028. This is the second top-tier cloud backstop via PPA after Oracle. Management said all hyperscalers are asking about CCS. Bloom can deliver turnkey solutions with a 10%–15% cost uplift, with CSPs willing to pay a low-carbon premium, and views CCS as a trillion-dollar-plus opportunity. b. **Ameren**: Ameren Missouri's 20-year IRP includes ~500MW of fuel cells by 2030, and currently only Bloom can supply at that scale. On the call, management said the utility channel will be a 'huge growth business'. Utilities can add Bloom servers to rate base and deploy them in front of the meter as T&D assets under IRPs. This shifts Bloom from 'negotiating customer by customer' to utility-planned batch procurement with larger orders and more predictable cadence, and management expects major orders. **3) Native 800V DC compatibility** Bloom natively outputs 800V DC. For a 1GW data center, direct DC supply can lower 5-year TCO by ~9% (~$5.5bn), based on company estimates. Nvidia requires 800V DC from Rubin Ultra in 2027, giving Bloom a first-mover edge, with solid-state transformers (SST) the main competitive threat. **Orders: steady growth, but dollar amount still undisclosed** The company said orders in hand are growing faster than revenue. It did not give a specific guide, but the order mix keeps improving, with broader customer coverage. a. From one cloud to many: After Oracle (MSA up to 2.8GW, 1.2GW signed), Google became the second top-tier adopter via AEP's PPA. Management said all major U.S. hyperscalers, plus a dozen-plus neoclouds, AI labs, and third-party DC operators, have validated its solution. b. Existing customers are reordering: AEP exercised from a 100MW framework to 900MW, and MiTAC continues to expand orders. c. Yet quarterly concentration remains high. Management noted large campus deliveries cause quarterly lumpiness. Concentration reflects delivery timing rather than order structure, and any given quarter can be dominated by one or two customers. Two customers contributed ~43% and ~26% of Q3 revenue (the latter an affiliate), totaling ~69%. This is slightly less concentrated than Q2's ~73% single-customer share, but still high. Given ongoing reorders by existing clients, Dolphin Research views this risk as manageable, though quarterly revenue volatility will be normal. As AI DC power demand stays strong and CSP capex remains robust, focus has shifted from 'is demand real' to 'can supply keep up and can deliveries be on time'. **Supply-side risks: three concerns are easing** **a. Oracle Jupiter: affects revenue cadence, not orders** Oracle's ~2.45GW hyperscale AI DC (primarily for OpenAI) delayed gas supply from Aug 2026 to Feb 1, 2027 due to pipeline rerouting. This pertains to lease rent obligations, allowing rent start to slip if COD is delayed, but does not change Bloom's order contract. Bloom shipments hinge on New Mexico air permits (decision by Nov 23). Commissioning uses little gas and can be met by trucked LNG, and Jupiter's planned ~$400mn 2026 revenue will likely shift into 2027. Yet full-year guidance rose, indicating other customers will backfill. Even if the site slips again, the MSA is not site-bound and equipment can be reassigned to other Oracle DCs. **b. Project Tembo (ex-Jade): concerns largely resolved** Last quarter, the market worried that partner Crusoe's exit would stall the project. In Q3 the picture cleared as Google replaced Crusoe as lead, Bloom's counterparty remains AEP with a 20-year PPA with Google, and the ~700MW site is Bloom's largest single-site project. **c. Supply chain and capacity:** Management did not detail constraints near 5GW, but raw materials are sufficient for current demand and orders. Core components such as hot boxes have multiple external suppliers like MTAR and Kaori expanding in parallel, supporting ~2.5GW annualized deliveries by 2028. On capacity, Bloom leased another Fremont building similar in size to the current plant. This lifts the capacity ceiling from 5–6GW to 10–12GW by Dolphin Research estimates, and management said the expansion prepares for beyond 5GW, with ramp tied to demand and no ahead-of-curve spend. Automation keeps line headcount roughly flat from 100MW to 5GW, lowering marginal expansion cost. Dolphin Research believes supply and execution risks are manageable overall. Under a base case where CSP capex stays elevated, AI DC power shortages persist, and gas turbine core component bottlenecks keep lead times high, SOFC should retain its market window. **V. Valuation and stock implications** Last quarter, we valued BE at $53.1–60.7bn on a 5GW 2030 capacity case, implying ~12%–28% upside vs. then market cap. The stock rebounded from a late-Jul low of ~$167 to ~$296 on MiTAC order expansion, Fremont expansion, and S&P 500 inclusion, then pulled back with tech to ~$218 (market cap ~$64.2bn), slightly above our neutral upper bound. **Q3 developments are raising the valuation ceiling:** a. Profit delivery ahead of plan: full-year revenue guide raised for the third time, 2026 revenue midpoint up ~122%, and Non-GAAP OPM ~22%. b. Physical basis for expansion: a second Fremont building supports a 10–12GW ceiling, anchoring a path to 8–10GW shipments by 2030 in an optimistic case. c. More diversified demand: demand is extending from primarily Oracle behind-the-meter to more clouds like Google and front-of-the-meter utility demand like Ameren. Military and Intl markets are progressing, though some are still in planning or negotiation. Net, this print supports the stock near term. Upside near term is capped by two factors: market cap is already slightly above neutral valuation, and order transparency is limited, so some funds may wait for the Q4 call. **Midterm, Dolphin Research stays 'neutral to positive' on BE:** Fundamentals keep beating, but a higher valuation band needs more proof on orders and capacity. With AI DC power still constrained, if orders and ramp cadence are confirmed, BE could trend toward $93.3–115.0bn in the optimistic case, and we will watch backlog disclosure and 2027 guidance on the Q4 call. **Looking ahead, a higher BE valuation base hinges on:** If North American AI power shortages persist, the second Fremont plant starts on schedule and validates the post-5GW path, Jupiter air permits land, front-of-the-meter projects like Project Phoenix and Ameren convert to firm orders, and 800V DC adoption scales in 2027–28. But risks include significant CSP capex cuts, major revenue timing shifts for Jupiter and other large projects, and competitor low-cost tech (e.g., metal-supported SOFC, SST plus grid) proving substitutive. Under such scenarios, the stock could face sharp drawdowns from elevated levels. \ Risk disclosure and statement: [Dolphin Research disclaimer and general disclosure](https://support.longbridge.global/topics/misc/dolphin-disclaimer) Past articles: Deep dive: **<**[**AI Power Crunch: Is Bloom Energy the New Cure?**](https://longbridge.cn/en/topics/42800822?app_id=longbridge&utm_source=longbridge_app_share&share_track_id=52a4500f-65e7-4658-8231-9f9641ab1a9c&invite-code=4NOXYT&locale=zh-CN&community_badge=1&profile_following_followers_activities=1)**\>** Earnings review: [**Bloom Energy: Riding the AI Power Shortage Despite Delay Fears?**](https://longbridge.cn/en/dolphin/post/43019264?app_id=longbridge&utm_source=longbridge_app_share&share_track_id=995a207a-a553-454d-b2e4-90ebd8372246&invite-code=4NOXYT&locale=zh-CN&community_badge=1&profile_following_followers_activities=1) ### Related Stocks - [BE.US](https://longbridge.com/en/quote/BE.US.md) - [ORCL.US](https://longbridge.com/en/quote/ORCL.US.md) - [ORCL-D.US](https://longbridge.com/en/quote/ORCL-D.US.md) - [BEX.US](https://longbridge.com/en/quote/BEX.US.md) - [BEZ.US](https://longbridge.com/en/quote/BEZ.US.md) - [ORCX.US](https://longbridge.com/en/quote/ORCX.US.md) - [ORCU.US](https://longbridge.com/en/quote/ORCU.US.md) - [ORCS.US](https://longbridge.com/en/quote/ORCS.US.md) --- > **Disclaimer: This article is for reference only and does not constitute any investment advice.**