Bloom Energy Lifts 2026 Revenue Growth Guidance to About 122% and Moves to Expand Fremont Beyond 5 GW, but Cash Flow Turns Negative as Working Capital Balloons
Third quarter 2026 earnings call, October 22, 2026: record $1.24 billion quarter, third guidance raise this year
Bloom Energy reported third quarter revenue of $1.24 billion, up 138% year-over-year and 16% sequentially, and raised its full year outlook for the third time in 2026. Management now guides to revenue of $4.4 billion to $4.6 billion, or roughly 122% growth at the midpoint. On the first quarter call, the company had been expecting about 60% growth. Non-GAAP operating income guidance rose to $960 million to $1.03 billion, a 22% margin at the midpoint versus 14% in the February outlook. Non-GAAP EPS guidance is now $3.10 to $3.40. Third quarter non-GAAP EPS was $0.94, against $0.15 a year ago.
Operating leverage is now visible, but gross margin guidance did not move
Operating income of $286 million rose more than 500% year-over-year and, in this one quarter, exceeded the whole of 2025. Operating margin reached 23.1% from 8.9% a year ago, as revenue grew 138% while operating expenses grew 26% to $141 million. Adjusted EBITDA was $299 million, or 24% of revenue. Year-to-date revenue of $3.05 billion is already 50% above full-year 2025.
Gross margin is more mixed. Product gross margin was 38.4%, up 250 basis points year-over-year and 120 basis points sequentially, despite the company ramping new production lines and expediting deliveries. Services margin was 21.4%, the second consecutive quarter above 20%. Total gross margin of 34.5% was held back by the mix effect of installation revenue, which jumped 66% sequentially to $84.9 million. Despite the higher revenue outlook, the company kept its full year non-GAAP gross margin guidance at about 34%, up from 30% in 2025 and the original 32% target. CFO Simon Edwards, six months into the role, attributed the unchanged outlook to start-up costs and expedite costs. He said engineering, sourcing and manufacturing cost-reduction programs "will take some time to flow through to COGS." He declined to guide 2027 but pointed to scale benefits as lines mature and utilization rises. The read-through is that near-term margin expansion is being traded for speed of delivery.
Cash flow turned negative, with working capital absorbing the growth
The weakest part of the quarter was cash. Operating cash flow was a use of $27 million and free cash flow a use of $181 million, after capital expenditures of $153 million. Year-to-date free cash flow remains positive at $42 million. Management attributed the swing to inventory, contract assets and receivables, as revenue is recognized when control transfers while billing follows contract milestones that can fall in a later quarter. Edwards said the quarter was "not the beginning of a trend" and that the company views itself as self-funding, which "reduces our dependence on external capital." Cash ended the period at $2.5 billion, roughly flat versus the start of the year despite year-to-date capex of $231 million, about 4x all of 2025. In response to Oppenheimer's question on contract assets, Edwards said the balance may keep growing but that the priority is that it turns into billings and collections as milestones are reached. Investors should watch whether fourth quarter conversion supports that claim, since the company also said it will keep investing in working capital.
Fremont expansion signals planning beyond the 5 GW milestone
In response to JPMorgan, CEO K.R. Sridhar confirmed the company is expanding its real estate footprint in Fremont to prepare for what comes after the 5 GW annual capacity milestone that the current footprint supports. He gave no timeline and no capacity target, and did not endorse the analyst's suggestion of 10 GW. Capacity, he said, will not be added in a step function: "We will never go past our headlights, but we will never be slow enough where we are the bottleneck to our customer getting the power." Pressed by Baird on the pace of quarterly additions, he said the company has not given quarterly numbers and that it is "all dialing up."
On automation, Sridhar cited the factory's general manager: when stack production was 100 MW a year, about 600 employees ran the line around the clock, and at 1 GW the same 600 employees will suffice, with the same headcount expected at 5 GW. This labor productivity is the cost-reduction argument management is leaning on.
Carbon capture moves from concept to a named project
The most important new commercial datapoint is the AEP announcement of 700 MW of carbon capture and sequestration-enabled Bloom Energy Servers to supply Google with near net-zero electricity. Sridhar framed the technical argument directly: "Combustion technologies do not offer an affordable and scalable pathway to carbon capture today." Bloom's systems, he said, produce a concentrated CO2 stream that makes capture viable. He sees a market of more than $1 trillion for the technology.
In response to Bernstein, he said all hyperscale customers are asking about carbon capture optionality, and that almost all would soon at a minimum buy systems that are "carbon capture ready" so they can connect to a CO2 pipeline and sequestration well when available. He argued that hyperscalers would voluntarily accept a 10% to 15% power cost premium for zero-carbon power, and that Bloom can deliver within that range, all-in. These are management claims and not yet demonstrated economics. On monetization, Sridhar said the company is agnostic: oil and gas partners could bolt on capture and monetize credits, Bloom could work with a partner on an end-to-end solution, or the end customer could keep the credits and pay Bloom for equipment plus a margin. No specific revenue model was committed to.
Utility channel and the front-of-the-meter argument
Management gave a clearer explanation of its utility strategy. Sridhar argued that utilities risk losing customers and rate base when large loads go behind the meter, shifting costs to remaining ratepayers. His proposal is for utilities to procure Bloom servers as distribution assets, placed in front of the meter and rate-based, without waiting for a specific customer order, much as they procure under integrated resource plans. He said utilities are "finally getting the leeway they need from their regulators" and that he expects "significant wins" in this category. No contracts were announced, so this remains an expectation, not yet a revenue stream.
Time to power and the hyperscaler economics argument
Sridhar said planning board approvals for the Nebius site in New Jersey came within 3 months of contract signing, with the first power block ready two months after approvals. He said Bloom's equipment is ready for the full 300 MW site, but the pace of energization is now up to the customer, which builds out block by block. He did not disclose the first block's size. For Oracle in Utah, power was delivered in under 2 months. He cited an NVIDIA claim that each quarter of delay on a gigawatt data center costs $10 billion in missed revenue, arguing this "overwhelms every other value proposition" on the power block.
In response to UBS, he expanded on the total-cost framing, saying LCOE is the wrong yardstick. He walked from LCOE to delivered grid power, to rack power cost, and then to the cost of power per token, citing UPS, voltage conversion and inefficiencies from turbines and engines that cannot cycle up and down. On water, he estimated about 25,000 million gallons over the life of a 1 GW data center on wet-cooled combined cycle gas turbines, against 1.4 million gallons for Bloom. He also repeated that 800V DC architecture would save a 1 GW data center more than $5 billion in its first five years, and said he is in talks with all hyperscalers. He conceded that timing depends on when 800V DC server racks arrive, which is for customers to answer.
Broadening demand: C&I, military, Texas and international
Management cited five wins: a conditional U.S. Army award of long-term leases to partner Indelible for power generation using Bloom servers on Army installations; the AEP/Google project; first shipments into Texas this quarter, one of the lowest power-cost markets in the U.S.; growing utility engagement; and international interest, with local partners being engaged and announcements promised "soon." Sridhar noted that higher natural gas prices abroad magnify the benefit of Bloom's efficiency. A company survey of over 200 commercial and industrial businesses found 25% expect to be fully powered on-site by 2030 and 35% by 2035, although this is self-commissioned survey data. Project scale now spans 150 kW systems at retail sites to the 700 MW Wyoming project, with the largest operating sites in the 100 MW class.
Backlog and what remains undisclosed
Management again said backlog is growing faster than revenue but declined to quantify it, with Sridhar saying exact figures will come with guidance early next year. Guidance rests on backlog assessed customer by customer for site readiness, plus a probability-weighted view of pipeline. Edwards noted that "the timing of a small number of deliveries around year-end will determine where in the range we land," which is a fair warning on fourth quarter lumpiness given the wide $200 million range. Bloom was added to the S&P 500 during the quarter.