--- title: "BE (Trans): Raises FY guidance across the board" type: "Topics" locale: "en" url: "https://longbridge.com/en/topics/43010158.md" description: "Below is Dolphin Research's summarized transcript of $Bloom Energy(BE.US) FY26 Q2 earnings call. Section I: Core financial highlights.1) Full-year guidance raised across the board. Revenue guidance lifted to $3.9–4.2bn; the midpoint implies 100% growth vs. 2025 revenue slightly above $2.0bn. Full-year non-GAAP OP guidance raised to $0.8–0.9bn, implying ~21% OPM..." datetime: "2026-07-28T22:40:04.000Z" locales: - [en](https://longbridge.com/en/topics/43010158.md) - [zh-CN](https://longbridge.com/zh-CN/topics/43010158.md) - [zh-HK](https://longbridge.com/zh-HK/topics/43010158.md) author: "[Dolphin Research](https://longbridge.com/en/news/dolphin.md)" --- # BE (Trans): Raises FY guidance across the board **Below is Dolphin Research's Trans of** $Bloom Energy(BE.US) **FY26 Q2 earnings call** **I. Key Financial Highlights** 1\. **Full-year guidance raised across the board**: FY26 revenue raised to $3.9–4.2bn, with the midpoint implying ~100% growth vs. a little over $2.0bn in 2025. Non-GAAP OP raised to $0.8–0.9bn, implying ~21% OPM, vs. the start-of-year $0.425–0.45bn (mid ~14% OPM). Non-GAAP diluted EPS guided to $2.55–2.85. Non-GAAP GPM held at ~34%. Guidance is built bottom-up on two layers: 1) backlog conversion (based on customer site-ready dates), and 2) conversion of current-year new orders. 2\. **First-ever $1bn+ quarter; record profitability** (non-GAAP): revenue $1.065bn (+166% YoY, +42% QoQ). Product revenue $0.935bn (+215% YoY, +43% QoQ), ~90% of total. OP $240mn (+737% YoY), OPM 22.5% (+1,536bps YoY). Adj. EBITDA $253mn (~24% margin). Non-GAAP diluted EPS $0.78; GAAP diluted EPS $0.62. 3\. **Gross margin and operating leverage**: total GPM 34.3% (+604bps YoY). Product GPM 37.2% (+193bps QoQ, +291bps YoY). Service GPM 22% (+977bps YoY), the fifth straight quarter with double-digit service margins. Revenue grew 166% YoY while Opex rose only 48%. Mgmt. stressed leverage is structural: R&D base and infrastructure are largely fixed, so each incremental GW adds minimal G&A; SG&A, service ops and supply chain scale via automation and analytics, tied to tech not headcount. The company will keep investing in R&D and management, but Opex growth should trail revenue materially. 4\. **Cash flow and liquidity**: OCF $226mn, up $439.5mn YoY on better profitability and working capital. FCF $175mn. Cash on hand $2.7bn at quarter-end. 5\. **Disclosure alignment and OCF baseline**: non-official metrics in prior supplemental materials have been aligned to formal guidance; FCF guidance is withdrawn. As a proxy, mgmt. noted the initial mid OP guide of $450mn mapped to ~$200mn OCF; OP was then raised to $675mn, now to a mid $850mn. vs. last update, the ~$175mn OP raise should convert 100% to OCF, implying a new OCF baseline of $375mn+. **II. Call Details** **2.1 Management Highlights** 1\. **AI data centers: standard in under a year** a. It took until 2022 to hit the first $1bn revenue year, then three years to double that. Now guidance implies another doubling within a year, and the company achieved its first $1bn+ quarter. b. On last year’s Q3 call, mgmt. announced Oracle as the first direct hyperscaler and delivered power in 55 days. **All major US hyperscalers, plus 10+ US neo cloud players, AI labs and third-party DC operators, have since validated and approved Bloom’s power solution.** c. The cycle from first contact to first order has compressed materially, enabling sign-to-ship-to-revenue within the same fiscal year. This drives in-period results, while replenishing and diversifying the backlog. d. Several customers have switched from alternative solutions to Bloom this year. Multiple major wins are not in the year-end backlog, and the company will ship partial systems to them this year. **Customers are shifting to longer-dated orders, pushing backlog growth ahead of revenue growth.** e. Entry points vary, including late-stage requests to take over as the primary on-site power solution, but the outcome is consistent: once customers see capability, execution and full-lifecycle value, discussions expand from single projects to portfolios. The model does not rely on any one customer or project; copy-exact, lego-like servers can be redeployed across sites, making projects fungible. Diversity, fungibility and flexibility position Bloom to adapt to a fast-moving AI landscape, underpinning visibility and confidence into 2026+ growth. 2\. **Commercial & Industrial (C&I)** a. Steady growth; Bloom is a standard solution for on-site power at hospitals, factories, telcos, university campuses and retail. b. It took nearly a decade to be accepted as standard in C&I, in contrast to sub-1-year standardization in AI data centers. 3\. **Friction removal I: capital** a. For a century, grid and plant costs were socialized over millions of ratepayers and amortized over decades; incremental load simply tapped surplus capacity with monthly payments. That surplus is gone; new DC loads require new infrastructure, heavy capex and long lead times, and ratepayers will not subsidize enterprise capacity. b. Islanded on-site power is faster, cheaper and more predictable, but needs capital budgets most end-customers lack, or a financing partner behind a PPA. Custom, deal-by-deal negotiations among financiers, developers, operators, OEMs and end-users are complex and slow. c. Brookfield anchors the financial shelf: set up at $5bn last fall and expanded 5x to $25bn in Jun. In parallel, Industrial Development Funding (IDF), together with Oaktree, MUFG Bank and Morgan Stanley, has committed a cumulative $2.6bn, with more partners lining up. Mgmt. emphasized capital of this quality and scale follows results, satisfied customers and financeable orders, not LOIs, MoUs or PR. GW-scale demand needs multi-bn-dollar capital, which Bloom has pre-arranged. 4\. **Friction removal II: community and permitting** a. Communities reject combustion but accept Bloom’s Energy Servers: no combustion, negligible air emissions vs. turbines and engines, negligible water use, friendly form factor, and quieter than AC units. b. No project is fully immune to NIMBY, but community acceptance is a real edge. Customers obtain air permits faster with Bloom than with combustion alternatives, and every month saved on permits brings power online a month earlier. 5\. **Friction removal III: speed, capacity and supply chain** a. Time to power is time to token revenue. Without power, AI chips are inventory, not intelligence. Grid operators offer timelines in years, leaving multi-bn-dollar compute clusters sitting in warehouses going stale. Traditional suppliers tout backlogs past 2029; mgmt. sees a 4-year backlog as an admission of supply constraints. Bloom delivers in months. b. **Since early this year, the company has been adding US manufacturing capacity via copy-exact increments and will continue to expand ahead of committed orders.** c. Repeatable speed relies on a resilient supply chain: broadly available raw materials, multiple qualified suppliers for each key input across multiple countries, inventory built ahead of ramps, and relationships built over two decades. No single supplier or country can dictate outcomes; every link is ready to scale with growth. 6\. **Biz. model and revenue recognition** a. Every deal starts with a contract with the end power user. **Two basic forms: customer CapEx purchase, or purchase of power/capacity without ownership,** including per-kWh PPA, capacity agreements or equipment leases priced per installed capacity. Economically, these are installment payments vs. asset ownership. All carry a committed COD. b. As most customers choose installments, Bloom brings in a financier (e.g., Brookfield) to buy and own the Energy Service Equipment from Bloom, then deliver to the end-customer under the original contract. Hence, the word 'customer' in disclosures can mean the financing counterparty in revenue/concentration, or the end-customer that created the demand. c. The recently announced IDF partnership is another example and contributed meaningfully this quarter: Nebbia signs the offtake, and IDF, as an independent third party, purchases energy services for specific sites and schedules, paying cash. d. Large-campus deliveries are lumpy: any quarter may be dominated by 1–2 customers, rotating as projects enter delivery windows, creating apparent concentration. This reflects delivery timing, not backlog composition. Backlog spans multiple hyperscalers, neo clouds, third-party DC operators and C&I customers, with payment protections appropriate to deal size. 7\. **Pricing, cost and services** a. Bloom sells value, not commodities. Customers pay for time to power and solution capability, whether tracking AI campus load curves or reserving headroom for carbon capture, and pricing reflects that. b. COGS continues to decline across materials, labor and manufacturing, even as capacity ramps and new capabilities are added. c. Service revenue is recognized net of guarantees, while service costs are expensed as incurred; fleet maintenance timing, especially stack replacements, drives quarterly volatility. Service GPM now exceeds 20%, driven by fleet performance, longer stack life and scale, and mgmt. believes this level is sustainable long term. d. Consolidated GPM will fluctuate with pricing mix by project, cost optimization vs. expedite trade-offs, and maintenance timing. When a '1-pt quarterly margin' conflicts with expediting for a multi-year customer, Bloom prioritizes the customer and long-term strategic value, consistent with the ~34% full-year GPM guide. 8\. **Operating philosophy and AI demand view** a. AI does not follow a traditional sales cycle: see demand, sign, ship when the customer is ready. Guidance level and shape stem from the same inputs: signed commitments and schedules, Bloom capacity, and the time-to-power pipeline. What used to be seasonality is now delivery timing tied to customer readiness. b. Mgmt. admits it cannot predict whether AI spend will keep compounding at the current pace (discussions suggest it not only continues but accelerates), but will not pretend to forecast it. The focus is on controllables: annual cost-downs (while others talk price-ups), sustained innovation and delivering on every commitment. c. Bloom benefits from both a rapidly expanding TAM and share gains, building durable advantages by earning the trust of customers and communities. **2.2 Q&A** **Q: Of the major US hyperscalers and 10+ operators who have validated and approved your solution, how many are live, how many are in backlog or near-term pipeline?** A: We let customers speak to their own deployments. All three categories you cited are represented — customers already in operation; those who have ordered, we have shipped, power pending with sites under construction; and those with definitive agreements signed. **We will not break out the three, but your premise is correct: it includes all major US hyperscalers and 10+ neo clouds and their ecosystems, including third-party DC partners.** Nine months ago we announced our first direct hyperscaler and said we aimed to replicate our C&I journey — become the standard. It took a decade in C&I; I did not expect we would be the standard in nine months, which speaks to our value proposition. This is not a faster horse; it is the car, and it is irreversible. **Q: Is capacity expansion timing or scale accelerating vs. last quarter?** A: Our capacity planning is anchored to the commercial pipeline and booked orders, so we know when customers need product and when units must go live. Depending on which report you read, ~30–40GW of new AI DC capacity will come online by 2027, at varying development stages and all Greenfield. We have a mature algorithm to assess which projects land when. Our units are fungible, unlike others. Even once loaded, we can redirect to another site, given system redeployability. This underpins our confidence that capacity will not be a customer bottleneck. **Based on what we see today, capacity will not be a constraint.** **Q: What do hyperscalers focus on, and how do you assure them Bloom will not be a delivery bottleneck?** A: These are highly sophisticated buyers. Their diligence goes beyond product, track record or economics. They want to understand progress on committed orders and our headroom for incremental orders. Most are not signing one-offs; they are forging strategic partnerships. Under NDAs, they share capacity roadmaps and ask if we can meet them; we walk through deep, itemized plans to prove we can scale. Only after this do we count as 'validated and approved' — a rigorous process repeated with each customer. **Q: Do financiers like IDF and Brookfield run similar validations?** A: Ultimately, financial partners take ownership of the equipment. They care not only about commissioning but lifecycle performance and delivery continuity. So they add layers of diligence: our delivery capability, availability commitments, and our ability to maintain assets over their model lives to realize returns. That is two extra layers on top of the customer process. Also, they did not jump straight to $20bn. They first committed $5bn, watched execution, probed our limits, evaluated delivery performance, and interviewed multiple customers — some with 15+ years of history — to assess outcomes and satisfaction. Satisfied customers were critical, and on that basis they scaled up. **Q: With Brookfield’s shelf up by $20bn, how should we think about the draw timeline?** A: Think of it as a financial shelf. It is stocked and ready, and the pace depends on utilization. This mirrors becoming the AI standard in under nine months. When they put in $5bn, I would not have predicted a $20bn top-up so quickly. It underscores the acceleration in AI and our business. **Q: Media flagged delays at large projects. What financial exposure do you have to project slippage, and what protections or alternatives exist in contracts?** A: We do not comment on specific projects. Stepping back, our MSAs provide flexibility to deploy copy-exact equipment across different customer projects. Our contracts have strong protections, mirrored by our financiers. If a project slips, end-customers can redeploy equipment to other projects; financiers remain obligated to take delivery from Bloom. Importantly, our raised FY26 revenue guide does not rely on any single project. Our algorithms assume some pushouts, some pull-ins, and some net-new in-year demand — as noted up front. Construction always entails delays; that is modeled as baseline. Hence, the FY26 guide has no single-project dependency. **Q: Market worries about scandium supply. What are your usage, supply visibility and inventory?** A: We published a detailed blog and filed an 8-K. Three takeaways: **first, economically recoverable scandium on earth is sufficient to power the planet; second, based on active workstreams, we have supply visibility for ~25GW of deployments; third, we do not rely on China.** That is all we will say; the rest is proprietary. **Q: With guidance sharply higher and $2.7bn cash, why withdraw FCF guidance? How are you thinking about capital allocation?** A: Context first: prior supplemental metrics that were not formal guidance have been aligned to what we formally guide. On cash, OP converts strongly to OCF. At the start of the year, a mid OP of $450mn mapped to ~$200mn OCF; later raised to $675mn, now mid $850mn. vs. last guide, the ~$175mn OP uplift should convert 100% to OCF, implying $375mn+ as a new baseline. We are not formally guiding OCF; this is to build confidence in conversion. **Q: On long-term AI demand and hyperscaler returns, are inference and reasoning already showing up, or is it still mainly training and time to power?** A: **Both, and both are highly certain.** Time to power is critical. Framed differently for utilities-focused investors: a 1GW DC, depending on AI customer mix, can generate $12–24bn revenue per year. If we can connect power within a month — power being the key bottleneck — that unlocks $1–2bn of revenue that would otherwise be missed, with 40–50% GPM and 20–25% NPM. Of the 35–40GW slated to come online next year, how many will slip due to power delays? We are the time-to-power answer. On inference, if the nation struggles to build 'highways' (transmission), imagine upgrading the 'city streets' (distribution), which is where inference power sits. Bloom fits perfectly — you cannot place a gas turbine in midtown Manhattan. Both opportunities are very strong and span many years. **Q: Given overall supply-demand and new capacity (recip engines, CCGTs, etc.), where is competition, and how do customers choose?** A: With the current gap, every fast-to-power technology will find demand over the next few years. If engine and turbine makers expand, the market needs it; if Bloom expands, the market needs it too. Longer term, when a customer must choose among turbines, engines and fuel cells, the primary metric is not LCOE — for on-site power, that is a flawed construct — but total cost to token, i.e., the total cost from power to token revenue. Bloom delivers 800V DC, reliability without excessive redundancy, siting in or near cities due to negligible air emissions, and permits that clear. No commercial supplier today matches this end-to-end value proposition. We focus on customers, not competitors. **Q: Do model commoditization and China’s open-source models pose a threat or an opportunity?** A: Whether labs are in China or the US does not matter. As a tech optimist, token cost, token efficiency and token utility will keep improving. Costs fall, efficiency rises sharply, and productivity per token climbs. Prices per token will drop, but total token volume will soar — Jevons paradox. That means more power, not less. If anything, this accelerates trends. Current AI forecasts are likelier underestimates than overestimates. **Q: You had a fuel-cell vision 21 years ago. What is the 3–4 year product vision?** A: Clear: DC will dominate on-site power, whether data centers, EV fleet charging, large residential complexes or microgrids for neighborhoods. The world is moving DC. Generate DC on-site, use waste heat for heating and cooling, and pursue decarbonization, where Bloom’s carbon capture is best-in-class. We aim to deliver an integrated solution: \>90% fuel utilization, no air pollution, negligible water use. The same architecture will power hyperscale DCs and your neighborhood store or inference site. We want it appliance-like — plug in, get power. **Q: Will capacity adds face capex inflation like broader manufacturing as you scale past 2GW?** A: Thank you — this is a key point of differentiation. **Our factory payback is measured in months.** We are not an old-world utility. We leverage the technology stack that made consumer electronics and semis better, larger-scale, lower-cost and higher-value. That is our model. Others can wrestle with their issues; for us, if demand is there, we keep adding capacity, with payback in months. **Q: Over the next 12–36 months, how might other fuel-cell tech (e.g., MCFC) affect DC share?** A: They should speak to their own MWs and GWs; not for us to opine today. In DCs, our share is north of 90%. If competitors enter, competition is healthy — it makes us hungrier, faster and sharper. We will grow through it, and we welcome it. **Q: As DC engagement deepens, how are pricing and target margins evolving? How many projects replaced other pre-planned tech?** A: We do not price by LCOE; we are not a utility. We are a strategic partner delivering value, and customers should be willing to buy and share value accordingly. So this is not just cents per kWh; it is the incremental value of the solution. Capturing that value to lift margins is critical. Also, remember where we came from: seven years ago, investors worried only about service losses. We said product reliability would improve and we would reach 20% service GPM. This quarter, service GPM was 22%. Eight years post-IPO, service GPM has moved from -21% to +22%, up 43pts. And the first word in service margin is 'service' — we serve customers. **Into 2025, 80% of orders come from repeat customers,** which speaks louder than any metric about satisfaction. Combine this with our backlog, external demand, tight fit with the digital future, execution, and the trust we have built with communities and enterprises, and we have much to be grateful for. **Risk Disclosure and Disclaimer:**[**Dolphin Research Disclaimer and General Disclosure**](https://support.longbridge.global/topics/misc/dolphin-disclaimer) ### Related Stocks - [ORCL.US](https://longbridge.com/en/quote/ORCL.US.md) - [BE.US](https://longbridge.com/en/quote/BE.US.md) - [ORCL-D.US](https://longbridge.com/en/quote/ORCL-D.US.md) ## Comments (1) *1 comments available on the platform.*