As the global artificial intelligence infrastructure build-out accelerates, Bloom Energy, the foundational energy provider and so-called "AI energy stock," has delivered its first-ever quarterly revenue exceeding $10 billion, smashing expectations. The company has also significantly raised its full-year guidance for 2026.

On July 28, Bloom Energy reported its second-quarter 2026 financial results and held a conference call. The figures exceeded Wall Street estimates across the board, prompting a substantial upward revision to its annual guidance. The company's shares rose over 10% in after-hours trading.

Financial data reveals that Bloom Energy's total revenue for the second quarter hit a record $1.065 billion, a 166% year-over-year increase that dramatically surpassed the consensus estimate of $826 million. Non-GAAP diluted earnings per share (EPS) came in at $0.78, far exceeding the analyst forecast of $0.41.

Driven by a robust conversion of its order backlog, Bloom Energy has raised its full-year 2026 revenue guidance from $3.4-$3.8 billion to a new range of $3.9-$4.2 billion. The full-year adjusted operating profit guidance has been doubled from an initial $425-$450 million to a new range of $800-$900 million.

During the earnings call, Chairman and CEO KR Sridhar set the tone for the company's accelerating growth with a timeline: "It took Bloom 21 years to achieve its first billion-dollar annual revenue in 2022. It took another three years to double that figure. Now, we anticipate doubling it again in just one year—and we just completed our first billion-dollar quarter."

Full-Year Guidance Boosted: Midpoint Suggests Annual Revenue Doubling

The most market-moving signal from the report was the substantial upward revision to the full-year 2026 outlook. Bloom Energy now projects full-year revenue of $3.9-$4.2 billion, with a midpoint of approximately $4.05 billion. This implies a near 100% increase compared to the roughly $2 billion in revenue generated in 2025.

Full-year non-GAAP diluted EPS guidance has been raised from $1.85-$2.25 to $2.55-$2.85. The non-GAAP operating profit forecast has been elevated from $425-$450 million to a range of $800-$900 million.

CFO Simon Edwards emphasized that the guidance is built on two layers: the conversion of signed contracts from the order backlog and the absorption of reserved capacity by new orders secured within the year. "The demand environment remains robust. We are not constrained by capacity, nor are we hampered by supply chain issues," he stated.

Notably, the company removed its previously disclosed free cash flow guidance from this report. Edwards explained this was merely an alignment of formal guidance metrics, offering an informal estimate for reassurance. Based on a midpoint operating profit of about $850 million, he projected operating cash flow conversion to be near 100%, implying an annual baseline of over $375 million. Operating cash flow for the quarter was $226 million, and free cash flow reached $175 million, with the company holding roughly $2.7 billion in cash and equivalents at the end of the period.

The earnings quality for the quarter was also noteworthy. Gross margin stood at 34.3%, an improvement of 604 basis points year-over-year. Product gross margin was 37.2%, up 291 basis points from last year. Service gross margin reached 22%, a 977 basis point year-over-year increase, marking the fifth consecutive quarter of double-digit service margins. Operating profit was $240 million, a 737% year-over-year increase, with an operating margin of 22.5%. Adjusted EBITDA was $253 million, representing about 24% of revenue.

Edwards highlighted the structural nature of the operational leverage: "Revenue grew 166%, while operating expenses increased by only 48%. The mechanism behind this will persist. Our R&D and G&A infrastructure are largely fixed against a rapidly growing revenue base. Each additional gigawatt of delivery incurs near-zero incremental administrative expense."

In response to a question about the order backlog, CEO Sridhar revealed a key signal: "Our backlog is growing faster than our revenue. Let me say that again—backlog is growing faster than revenue." He also noted that several major customers this year have placed orders and received shipments within the same year, without being included in the year-end backlog statistics.

Disrupting the Traditional Grid: 'Chips Without Power Are Just Inventory'

During the call, CEO Sridhar delivered his most impactful statement regarding the company's position in the AI data center market: "Today, all major US hyperscale cloud providers, and over a dozen emerging US cloud providers, AI labs, and colocation data center operators, have validated and approved our power solution for their AI factories."

He placed this achievement in a historical context: "Nine months ago, we announced our first direct hyperscale cloud customer (Oracle) and delivered power to their data center in 55 days. It took us nearly a decade to become an accepted standard in the commercial and industrial market. In the AI data center market, we became the standard in under a year."

When asked by an analyst about the competitive moat, Bloom's management repeatedly stressed a core advantage: Time to Power. In the AI arms race, computing giants cannot afford the multi-year grid expansion timelines of the traditional electric grid. "We deliver power at AI speed, enabling our customers to grow. This makes us a critically important strategic partner," Sridhar said, describing the current pain point with a powerful metaphor: "Bloom is increasingly seen as the solution to eliminate the friction of AI power availability. Chips without power are inventory, not intelligence."

Addressing traditional equipment suppliers who boast about long-term order books, Sridhar offered a sharp critique: "Grid operators provide timelines stretching for years. Under such a timeline, a billion-dollar computing cluster becomes obsolete in a warehouse. Traditional suppliers celebrate backlogs extending to 2029 and beyond. We view a four-year backlog not as a trophy, but as a confession of constrained supply."

Bloom's ability to deliver power within months creates a disruptive advantage over the traditional grid and other power generation forms. Sridhar analogized, "This isn't a faster horse; it's a car. That's why this is happening and is irreversible." For executives, the core logic for customers paying a premium for Bloom is that powering an AI data center just one month early can generate hundreds of millions of dollars in additional token revenue for computing giants.

Regarding the competitive landscape, when asked about competition from gas turbines, reciprocating engines, and other fuel cell technologies, Sridhar stated that Bloom's current market share in the data center sector is "in the very high 90s." He emphasized, "Currently, no commercial supplier can offer the complete value proposition that Bloom provides—800-volt DC power, high reliability without overbuilding, deployability both in and out of cities, and faster air permits."

Looking ahead to the potential slowdown of AI investment, Sridhar maintained a neutral yet optimistic tone: "Regardless of the lab in China or the US, the cost of tokens will decrease, but total usage will explode. When that happens, you need more power, not less. If anything, current predictions for AI power demand are understated, not overstated."

The full transcript of Bloom Energy's second-quarter 2026 earnings conference call is as follows (translated with AI assistance).

Date of Conference: July 28, 2026
Company Name: Bloom Energy
Conference Description: 2026 Second Quarter Earnings Conference Call
Source: Bloom Energy

Opening Remarks

Operator: Good day and welcome to the Bloom Energy 2026 second-quarter earnings conference call. Please note that today's call is being recorded. I will now turn the conference over to Mr. Michael Tierney. Please go ahead.

Michael Tierney, Vice President, Investor Relations: Thank you, and good afternoon, everyone. Thank you for joining us for Bloom Energy's second-quarter 2026 earnings conference call. In conjunction with this call, we have filed our second-quarter 2026 earnings press release and supplemental financial information on Form 8-K with the SEC, and these materials are available on our investor relations website. We will be referencing these materials throughout the call. During this call, whether in our prepared remarks or in response to your questions, we may make forward-looking statements that represent our expectations regarding future events and future financial performance.

These statements pertain to our business results, products, markets, customers, strategy, financial condition, liquidity, and our 2026 full-year outlook. All such statements are based on our expectations, estimates, and assumptions and are forward-looking in nature. However, since they involve future events, they are subject to numerous known and unknown risks and uncertainties, which are described in our filings with the SEC, including our most recent 10-K and 10-Q reports. We undertake no obligation to revise any forward-looking statements made on today's call. During this call, and in our second-quarter 2026 earnings press release and supplemental financial information, we will reference both GAAP and non-GAAP financial measures. Non-GAAP financial measures are not prepared in accordance with U.S. GAAP and are provided only as a supplement to, not as a substitute for or superior to, measures of financial performance prepared in accordance with GAAP. A reconciliation of the GAAP and non-GAAP financial measures is included in the materials mentioned and is available on our investor relations website.

Joining me on the call today are KR Sridhar, Founder, Chairman, and Chief Executive Officer, and Simon Edwards, Chief Financial Officer. KR will begin with an overview of our progress, and Simon will then review the financial highlights for the quarter. Following the prepared remarks, we will open the call for questions.

I will now turn the call over to KR.

KR Sridhar, Founder, Chairman, and CEO: Good afternoon, everyone, and thank you for joining. It took Bloom 21 years to achieve its first billion-dollar annual revenue in 2022. It then took another three years to double that 2022 revenue. Now, we expect to double it again in just one year—and we have already delivered our first billion-dollar quarter. Bloom Energy's business is accelerating. We are scaling at an incredible pace while continuously improving our profitability.

We have successfully demonstrated the profitability of this flexible business model, which is operating exactly as designed. Throughout this journey, we have consistently delivered on our promise to customers: Bloom will not be your bottleneck. We deliver power at AI speed, enabling our customers to grow. This makes us a critically important strategic partner and builds customer loyalty. Demand is building upon itself. New customers are arriving faster than ever before. In the past, landing a major customer took years; that 'moat' once protected the industry's incumbents. Now, our technology's proven track record, combined with the increasingly urgent need for efficient, clean, reliable power, has dramatically shortened the time from first contact to first order. This is because we can sign, ship, and recognize revenue for orders within the same fiscal year. This demand is visible in our results today while also enriching and diversifying our order backlog. Notably, this year, several customers who had already pursued other solutions abandoned them to choose Bloom.

Once a customer joins, they see the overall economic value of our solution and place repeat orders. As of today, we have signed several significant customers that were not part of the backlog we reported at the end of last year, and we will ship systems to them this year. These customers are now placing longer-term orders, which is causing our backlog to grow faster than our revenue. Let me repeat that: the rate of backlog growth is exceeding the rate of revenue growth.

Bloom Energy has become the industry standard for on-site power, just as we predicted during our third-quarter 2025 earnings call. At that time, we had just announced our first direct hyperscale cloud customer, Oracle, and delivered power to their data center in 55 days. Today, all major U.S. hyperscale cloud providers, along with over a dozen emerging U.S. cloud providers, AI labs, and colocation data center operators, have validated and approved our power solution for their AI factories. Our commercial and industrial business is also growing.

We are the standard solution for on-site power for hospitals, factories, telecom operators, university campuses, and retail stores. However, it took us nearly a decade to become an accepted solution in these verticals. In contrast, we have become the standard in the AI data center market in less than a year. I founded Bloom with a firm belief that on-site power is critical to powering the world and fueling the digital transformation. We built this company to provide the best on-site power solution, removing all friction for our customers.

Our solution is clean, reliable, fast, and affordable. Customers do not need to compromise or trade off between these characteristics. Over time, we are continuously reducing friction, turning headwinds into tailwinds. Let's take a moment to discuss four friction points: capital, community, permitting, and speed.

First, capital. For a century, the cost of power plants and the grid was spread across millions of users and amortized over decades. New demand simply connected to existing spare capacity and paid a monthly bill. But today, grid spare capacity is exhausted. New data center loads require new infrastructure, involving significant capital, long construction cycles, and regular power users will not pay for or subsidize the capacity needs of corporate customers. For these customers, a faster, more economical, and predictable path is a self-contained on-site power system. However, this path requires either capital budgets most end customers lack or the introduction of financing partners through Power Purchase Agreements (PPAs). Negotiating bespoke terms between financiers, developers, operators, equipment manufacturers, and end-users is both complex and time-consuming.

Therefore, we eliminate this friction by providing project capital through our strong financial partners. Brookfield is a central pillar of this financing framework. We established this partnership last fall with an initial commitment of $5 billion. Nine months later, in June of this year, Brookfield expanded its commitment five-fold to $25 billion. One of the world's largest and most experienced infrastructure investors evaluated our technology, track record, and project pipeline, supported us with $5 billion, witnessed our execution, and then increased its support by 500%. Capital of this magnitude and quality does not follow letters of intent, memoranda, or press releases; it follows performance, satisfied customers, and solid, financeable orders. Brookfield is not alone. This quarter, Industrial Development Funding, which has previously financed Bloom projects, partnered with Oaktree, MUFG Bank, and Morgan Stanley to provide financing for Bloom projects, with a cumulative commitment of $2.6 billion, and more financing partners are waiting in the wings. Gigawatt-scale demand requires gigawatt-scale capital, and we have arranged for it in advance.

Next friction point: community. Communities have realized they cannot live near combustion-based equipment. However, they have also realized they are perfectly comfortable having a Bloom Energy server nearby—no combustion, negligible air pollution compared to gas turbines and engines, minimal water usage, an attractive appearance, and noise levels quieter than air conditioning units. No construction project is entirely immune to NIMBYism, but community acceptance of Bloom is becoming a true competitive advantage. Our customers obtain air permits faster using our technology than with combustion alternatives. Every month saved in permitting means a month closer to having power available, which brings us to our next topic—speed.

Because the availability of power is fundamentally the time when computing power generates revenue. Bloom is increasingly seen as the solution to eliminate the friction of power availability in AI. Chips without power are inventory, not intelligence. Grid operators provide timelines measured in years, and during that time, a multi-billion dollar computing cluster can become obsolete in a warehouse. Traditional suppliers celebrate backlogs extending to 2029 and beyond. We view a four-year backlog not as a trophy, but as a confession of constrained supply. In contrast, Bloom can meet our customers' urgent timelines, delivering power in months.

Since the beginning of this year, we have been consistently expanding our U.S. manufacturing capacity on an incremental 'copy-exact' basis and will continue to do so against committed orders. Our speed is reliable because we have deliberately built a resilient supply chain using widely available materials, with multiple certified suppliers in multiple countries for every key input. We carry inventory ahead of ramps, and our partnerships have been forged over two decades. No single supplier and no single country can dictate our fate. Every link in our supply chain is ready to scale with our growth.

When the business value is measured in months of AI computing power, the fastest, most reliable path to power delivery is the winner. We are on that path, and no competitor is close. None of this is accidental. Years ago, we embedded scalability into our business model for a simple reason: if our thesis about on-site power demand is correct, the company must be capable of rapid expansion. You are now seeing this borne out in practice.

Capital, community, permitting, speed—eliminate these four frictions, and the market delivers its verdict: the preferred supplier. I choose those words carefully because customers are actively making a choice. Customers who had already ordered gas turbines and reciprocating engines have canceled those orders and chosen Bloom instead—this happened in the quarter. Existing customers are returning with expansion needs. The world's largest infrastructure investors are underwriting our deployments at scale. We are a trusted partner to customers, communities, and capital.

The market has taken notice. Customers are now approaching us later in their development process, requesting Bloom as the primary on-site power solution. The entry point varies, but the outcome is consistent. Once they see our capability, our execution, and the total value we create over the asset lifecycle, the conversation expands from a single project to a portfolio of projects.

Therefore, let me be very clear: Bloom Energy is not dependent on a single customer or a single project. We have multiple customers and multiple projects at every stage of development. Moreover, because we use a 'copy-exact' Lego-like server architecture that can be seamlessly redeployed across sites—unlike customized traditional equipment—every project in the portfolio is interchangeable. This diversification, interchangeability, and flexibility allow us to navigate the fast-changing AI landscape adeptly. Consequently, we have strong visibility and confidence in our growth trajectory for 2026 and beyond.

Finally, let me talk about how we run this company, because I know what's on your minds: Will AI investment continue at this astonishing pace? Our deep engagement with customers gives us confidence that the pace of investment will not only continue but also accelerate further. However, I cannot be 100% certain, and I won't pretend I am to fool you. Henry Ford couldn't control whether Americans wanted cars; he could control the cost, quality, and supply of the Model T. Like Ford, we focus on managing what we can control. While other companies benefiting from this build-out frenzy only talk about raising prices, we are consistently reducing costs year after year. We continue to innovate to better serve our customers and deliver on every promise we make.

We are simultaneously enjoying the tailwinds of a rapidly expanding TAM (Total Addressable Market) and increasing market share, and we are grateful to meet market demand. At the same time, we are building a durable competitive advantage by earning the trust of our customers and communities.

With that, let me hand the call over to Simon to walk you through the details. I'll be back for Q&A. Simon?

Simon Edwards, CFO: Thank you, KR, and good afternoon, everyone. This is my second earnings call as Bloom's CFO and my first with a full quarter of results to present. In April, I told you why I joined Bloom—belief in the mission, the structural shift happening in power, the high quality of the team, and the opportunity to help build a company for the ages. Three months in, that belief has only strengthened. The demand environment is very strong. We are not capacity-constrained or supply-chain bottlenecked. Operating discipline is real.

The factories are improving week by week. Cost reduction and efficiency gains are a daily rhythm here, not a project. Financially, we have a strong foundation, and my focus is on expanding our systems and processes to match our rapid growth pace. Before diving into results, let me go back to the Brookfield announcement KR mentioned—a five-fold expansion of our strategic partnership, significantly increasing their framework to finance Bloom Energy projects for AI infrastructure.

A commitment of this scale from one of the world's largest infrastructure investors is significant in its own right. It also serves as the perfect entry point to understand our revenue model, because one cannot truly grasp the value of this partnership without understanding the business model behind it. Given our shareholder base has grown substantially over the past year, let me take a minute to explain how a Bloom transaction works. Every Bloom transaction begins with a contract with an end customer—the party that will actually use the electricity. This contract takes two basic forms: either the customer purchases the equipment directly, a CapEx sale, or the customer signs a power or capacity contract but does not own the equipment.

As described in our 10-K annual report, the second form has several variants: a kWh-based PPA, a capacity agreement, or an equipment lease based on installed capacity. Economically, they operate the same way—the customer chooses to pay over time rather than own the asset. In either case, each contract has a clearly committed commercial operation date. Since most customers choose the pay-over-time model rather than direct purchase, we introduce a financing party—an institution like Brookfield—that purchases the energy service from Bloom, holds the asset, and provides service to the end customer under the contract we signed. Therefore, when you see the term 'customer' in our filings, it can refer to either party: the financing party buying from us and appearing in our revenue and concentration disclosures, or the end customer whose demand originated the transaction.

The recently announced IDF partnership is another practical example of this model and was a significant contributor this quarter. Mechanistically, this is our standard structure—signed PPAs, with IDF acting as an independent third party purchasing the energy service on cash terms against defined sites and delivery schedules.

There is one characteristic of this model worth understanding, especially for newer investors: deliveries for large campuses are uneven. In any given quarter, one or two customers may dominate our revenue, and the customer mix rotates as different projects enter their respective delivery windows. Revenue in a single quarter may appear highly concentrated, but this concentration reflects delivery timing, not the composition of our backlog. Our backlog spans multiple hyperscalers, emerging cloud providers, colocation data center operators, and commercial and industrial operators. Our contracts also have payment security mechanisms commensurate with their transaction size.

Now, on to the results. A reminder that I will focus on non-GAAP adjusted metrics. A full reconciliation of GAAP to non-GAAP measures is included in the press release and supplemental materials on our investor relations website.

Revenue was $1.065 billion, up 166% year-over-year and 42% sequentially. This is another record quarter for Bloom and our first quarterly revenue above $1 billion, reflecting the acceleration of data center deliveries and disciplined execution converting signed demand into revenue. Product revenue was $935 million, up 215% year-over-year and 43% sequentially, representing nearly 90% of total revenue this quarter. Gross margin was 34.3%, an improvement of 604 basis points year-over-year. This improvement reflects a favorable business mix and gross margin expansion in both product and service segments. Product gross margin was 37.2%, up 193 basis points sequentially and 291 basis points from Q2 2025.

On pricing, we deliver value, not commoditized kilowatt-hours. Our customers pay for the ability to get power fast and for the capabilities this solution enables—whether it's following the load curve of an AI campus or being ready for carbon capture. Our pricing reflects this. On costs, we continue to drive down product costs across materials, labor, and overhead, while simultaneously scaling capacity and adding new capabilities.

Service gross margin was 22%, up 977 basis points year-over-year, marking our fifth consecutive quarter of double-digit service margins. Service revenue is recognized quickly after deducting warranty costs, while service costs are booked when incurred. Therefore, the timing of fleet maintenance—especially stack replacements—can cause gross margin to fluctuate between quarters. Excluding such timing differences, service margins have stabilized above 20%, benefiting from improved fleet performance, extended stack life, and scale economies. We believe this level is sustainable over the long term.

Overall gross margin will vary depending on the pricing mix of projects delivered each quarter. It involves deliberate trade-offs between cost optimization and accelerating delivery—when a customer's 'time-to-power' value exceeds the incremental cost, we prioritize the customer, and the same applies to the service timing I described. All considered, we remain confident in our full-year gross margin outlook of approximately 34%, a target we raised last quarter.

Operating profit was $240 million, up 737% year-over-year, with an operating margin of 22.5%, an expansion of approximately 1,536 basis points. Adjusted EBITDA was $253 million, representing about 24% of revenue. Non-GAAP diluted EPS was $0.78, and GAAP diluted EPS was $0.62. The improvement in profitability reflects both higher volumes and significant operating leverage from scale.

It's worth noting that the operating leverage in these numbers is structural, not a one-quarter effect. Revenue grew 166%, while operating expenses increased by only 48%. The mechanism behind this should persist. The leverage comes from how we have structured our cost base. Our R&D foundation and G&A infrastructure are largely fixed costs, and the revenue base is growing rapidly. Therefore, each additional gigawatt of delivery adds near-zero incremental management overhead. We manage our support functions the same way we manage our factories—with extensive use of automation and data analytics across SG&A, service operations, and supply chain. As a result, these functions scale with technology, not headcount. We will continue to invest across the business, including in G&A and R&D, but we expect operating expenses to grow at a much lower rate than revenue for the foreseeable future, driving ongoing operating margin expansion.

Cash flow from operations was $226 million, an increase of $439.5 million compared to the prior year, driven by improved profitability and good working capital performance. Free cash flow was $175 million, and the cash balance at the end of the quarter was $2.7 billion. Generating this level of operating cash flow while growing revenue at this pace is a true testament to the strength of the underlying business and the team's rigorous working capital management.

Now, turning to guidance. Based on our strong year-to-date performance and our high visibility and confidence in the second half, we are raising our full-year revenue outlook to $3.9 billion to $4.2 billion. At the midpoint, this implies approximately 100% growth compared to slightly over $2 billion in revenue for 2025. Our outlook is built bottom-up, on two layers: the base layer is backlog conversion—signed commitments delivered according to customer site readiness dates; the second layer is new orders within the year. We have intentionally reserved manufacturing capacity for 'time-to-power' customers who need power urgently, and we expect to book and convert capacity at rates consistent with recent experience.

On gross margin, we are maintaining our full-year non-GAAP gross margin target of approximately 34%. I also want to take this opportunity to explain an operating principle: when we are faced with a choice between preserving a margin percentage point in a quarter and accelerating delivery for a customer we will partner with for the long term, we will prioritize the customer and the long-term strategic value of that relationship. 'Time-to-power' is what our customers value most right now, and we will continue to deliver on that promise. Over a full year, this operating principle is fully consistent with the margin levels we are guiding.

On operating profit, we are raising our full-year non-GAAP operating profit outlook to $800 million to $900 million. At the updated revenue midpoint, this implies an operating margin of approximately 21%. This is a substantial increase from our initial guidance of $425 million to $450 million in operating profit (midpoint implying a 14% margin). This is the operating leverage I described, directly reflected in the model. Full-year non-GAAP diluted EPS outlook is now updated to $2.55 to $2.85.

Finally, a few words on how to interpret our guidance. Demand in the AI business does not follow traditional sales cycles. Today, we see demand, we take orders, and when the customer is ready, we ship. Therefore, the level and shape of our outlook come from the same inputs—signed commitments and their timelines, our capacity, and our 'time-to-power' project pipeline. What was once seasonality is now just delivery timing based on customer readiness.

In summary, this was a milestone quarter. We exceeded $1 billion in quarterly revenue for the first time, achieved record profitability, generated strong cash flow, and raised our full-year outlook. We are executing with discipline against a backdrop of continued robust demand. With that, operator, we are ready for questions.

Question and Answer Session

Operator: Thank you. Ladies and gentlemen, we will now begin the question-and-answer session. Our first question comes from Mark Strouse of JPMorgan.

Mark Strouse, Analyst: Yes, good afternoon. Thank you very much for taking our questions. KR, I wanted to go back to a comment you made earlier about all major US hyperscalers, and over a dozen other operators now have validated—sorry—validated and approved your technology. Can you maybe talk a little bit about how many of those are currently actively using your technology? How many are in the backlog or near-term pipeline? And then I have a brief follow-up. Thank you.

KR Sridhar, Founder, Chairman, and CEO: Mark, as you well know, we leave it to our customers to talk about their deployments and their business. What I can tell you is that there is a combination of the three categories you mentioned: customers already using our technology, customers who have placed orders and we are shipping to them, with power delivery imminent and construction ongoing, and customers who have signed definitive agreements with us. They fall into that universe. We don't break these out separately, but it does cover, as you noted, all the major US hyperscalers and over a dozen emerging cloud providers (neoclouds) and the ecosystems built around them, including colocation partners. These are real. We don't segment them. But let's just pause and think about this. Nine months ago, we announced our first direct hyperscale cloud customer and said we wanted to enter this market and become the standard, just as we had done in the C&I space. It took us ten years to do that in C&I. And I can tell you, I never thought we would become the standard in nine months. That speaks volumes about our value proposition across the industry—this isn't a faster horse, it's a car. That's why this is happening, and it's irreversible. Thank you.

Mark Strouse, Analyst: Okay, very helpful. And if I could ask one quick follow-up. I know, going back to the last call, you didn't want to give a specific capacity number. But perhaps relative to that last call—given your repeated comments today about things accelerating—is it fair to assume that your planned capacity timeline or capacity number is also accelerating accordingly?

KR Sridhar, Founder, Chairman, and CEO: Yes. Here is how we do capacity planning: based on our commercial pipeline and commercial orders, we have very good visibility into when customers need product and when they are ready to energize. As you well know, whether you read anyone's report, there will be about 30 to 40 gigawatts of new AI data center capacity coming online in 2027. These projects are at various stages of development, all greenfield. We have a very sophisticated algorithm to determine how many of these projects will land and when. Fortunately, unlike everyone else, our equipment is fungible. When it's on the truck, we can redirect it to a different site if needed. That's how flexible our system is. Based on this, we can predict our capacity needs to ensure we are never the bottleneck for our customers. I can tell you, from where we stand today, we are confident in our ability to deliver on this promise—for everything in our order book and the future demand we foresee. That's it. That's the extent of our capacity commentary. For now, capacity will not be a constraint. Thank you.

Operator: The next question comes from Chris Dendrinos of RBC Capital Markets.

Christopher Dendrinos, Analyst: Yes, thank you and congrats on a strong quarter. I wanted to ask about the supply chain. I'm curious, in your discussions with hyperscalers, what are they asking you, and how are you responding to convince them that you won't be a bottleneck in terms of on-time delivery? Thank you.

KR Sridhar, Founder, Chairman, and CEO: That is an excellent question. You bring up a very important point. These are extremely sophisticated consumers and customers. Their due diligence goes far beyond our product, our performance, or our economic value proposition. They want to understand our current committed orders and our progress on new orders we can sequentially bring on. And if you talk to most of them, they aren't just signing up for a single transaction. They are signing up with us for the future. They want to be a strategic partner with us going forward. They want to understand—they share their capacity expansion plans with us confidentially and ask if we can meet those needs. We must go through very detailed, one-on-one discussions with them under NDA and convince them we have the capability to scale. That's when we get the validation. So going back to the earlier question about 'what these validations mean,' this is the process we go through. It is a fairly rigorous process with every single customer.

Christopher Dendrinos, Analyst: Got it. And then I wanted to ask, do you go through the same process with IDF and Brookfield? And on Brookfield, you expanded that partnership by $20 billion. How should we think about the timing of executing against this program? Do you have an expected timeframe to deploy that $20 billion? Thank you.

KR Sridhar, Founder, Chairman, and CEO: Sure. That's a good two-part question, you snuck it into one—but I'm happy to answer both, they are very relevant.

Part one, ultimately, financial customers take ownership of our equipment. When they take ownership, they care not only about getting it to work but also about its ability to perform consistently and deliver over its entire useful life. Therefore, they add extra layers of scrutiny on our ability to perform, availability, and our capability to maintain the equipment over the timeframe their financial model requires, so they can actually get their return. So these are two additional layers they add in their due diligence, on top of the other processes we discussed. Even in today's world, $20 billion is a lot of money. So obviously, they go through the entire process with us in depth. Also, remember they didn't start by deploying that amount directly. They started with $5 billion, watched our performance, our capability, our execution, and talked to multiple customers—the oldest of which has been with us for over 15 years—to understand our performance and their satisfaction. Customer satisfaction is absolutely necessary. It's based on all this that they chose to continue investing. As for the timing, think of it as a financial shelf that is now set. The speed at which the shelf gets used will depend on the pace of deployment of these funds. I would say again, it's very similar to our journey of becoming the standard in AI in less than nine months. Nine months ago, when they put in $5 billion, I would not have predicted we would come back so soon for the next $20 billion. That speaks to the pace of acceleration in AI and our own business. Thank you.

Operator: The next question comes from David Arcaro of Morgan Stanley.

David Arcaro, Analyst: Oh, great. Thank you. Thanks for taking my questions. There have been some recent projects that have had some challenges in project development that have gotten media attention. I was wondering if you could just describe your financial exposure to project delays, typical protections that you have in your contracts, and maybe the alternative plans that you might work out with customers?

KR Sridhar, Founder, Chairman, and CEO: Yes, thank you for that question. We typically don't comment on specific projects, as you know. But I'll step back and talk about how we structure our contracts. Our contracts are structured with a master services agreement, and given the 'copy-exact' nature of our equipment, we have great flexibility in redeploying equipment to different customer projects. On top of that, we have strong protections in our contracts. We need those protections from a financial perspective as well. So when you think about how these contracts work, if there are any project delays, the end customer can redeploy that equipment to other projects, but ultimately, the financing party has the obligation to accept equipment delivery from Bloom.

Simon Edwards, CFO: And one very important point specifically on project issues, because that is certainly top of mind. We can tell you that when we provided guidance and raised our 2026 revenue guidance, that guidance is not dependent on any single project. We have a sophisticated algorithm. We expect certain projects to be delayed, certain projects to land on time, and certain projects to pop up and be absorbed within the year, as we described in our prepared remarks. So all of that is factored into our guidance. Construction projects, as long as construction projects exist, I believe there will be delays, okay? I'm not a historian, but that's my expectation. We should bake delays into our assumptions, but it doesn't affect our revenue guidance because we have a sophisticated algorithm to handle things throughout the year. Therefore, our 2026 guidance is not dependent on any single project.

David Arcaro, Analyst: Got it. Thanks for the extra color on that, very helpful. KR, I also appreciated your added color and confidence on the supply chain. I was wondering if you could talk about your access to Scandium, which has been getting a lot of attention lately. Could you describe your use of it, what you see as the available supply in the market, inventory buffers, etc.?

KR Sridhar, Founder, Chairman, and CEO: We have published a detailed blog post and filed an 8-K on this topic. There are three things you need to know as an investor: One, there is enough Scandium on Earth to be economically recovered to power the entire planet—that's the actual resource on Earth. Two, based on our current progress, we have clear visibility for a deployment scale of 25 gigawatts. Three, we are not dependent on China. Those are the statements we have made, and that's all we are going to say. Everything else is company proprietary. Next question.

Operator: The next question comes from Nick Amicucci of Evercore ISI.

Nicholas Amicucci, Analyst: Hey everyone, how are you doing? Simon, sorry to put you on the spot a little bit. I'm just curious, obviously strong Q2, big guide raise, but the free cash flow guidance was pulled. Just wanted to get some context on that, and also thinking about the $2.7 billion of cash on the balance sheet and thoughts on capital allocation?

Simon Edwards, CFO: Yes. Hey Nick, thanks for the question. First, let's set the context. The company historically included some metrics in the supplemental deck that weren't formal guidance. So we are now just aligning the deck with what we actually guide on. But getting back to your question on cash, which I think is the key question, we see a very strong conversion from operating income to free cash flow. Think about where we started the year, operating income guide midpoint of $450 million equating to $200 million in operating cash flow, we then raised it to $675 million, and now the midpoint is $850 million. That's about a $175 million increase from our previous operating income guide, and we think that $175 million will convert at 100% to operating cash flow. So you can think of over $375 million as our new baseline. Of course, as you know, we don't provide formal guidance on earnings releases, but I just wanted to make sure you are comfortable that we are seeing strong cash conversion.

Nicholas Amicucci, Analyst: Yes, no, that makes a ton of sense. And then, when we think about this longer-term AI demand and why everyone's CapEx is so high right now, and thinking about where the actual returns will be generated. When we think about that and combine it with your backlog and the conversations you're having, KR, have you started having conversations around inference reasoning yet, or is it still mostly just time to power on training?

KR Sridhar, Founder, Chairman, and CEO: Yes, both, absolutely both. First, time to power is extremely important, that's number one. Let me explain 'time to power' in a slightly different way, because it's very important for many analysts who study utilities and Bloom Power together, and for some of our investors. A full-stack AI provider responsible for all data center finances, a 1-gigawatt data center, in a single year, depending on the nature of the AI customer, can generate $12 billion to $24 billion in revenue. If you can give them power in one month, that's the key, that represents $1 billion in revenue they wouldn't otherwise have, with 40-50% gross margins and 20-25% net margins. Out of the 350 to 400 gigawatts that need to be deployed next year, guess how many projects will be delayed because the power provider can't deliver on time? We are the best place to solve the 'time to power' problem. If we can do that, you don't even need to do the math. That's why 'time to power' is so important for large data centers.

Now, with the onset of inference demand, if the transmission and distribution infrastructure in this country is already challenged with transmission (think building highways is hard), you can imagine how hard it will be to upgrade distribution networks (think local streets). And that's exactly where inference power is needed. Bloom is perfectly suited for that. You can't put a gas turbine in the middle of Manhattan. So we see both opportunities as extremely strong for us, not this quarter, not next quarter, but for years to come. Thank you.

Operator: The next question comes from Ben Kallo of Baird.

Ben Kallo, Analyst: Hey, thanks for taking my questions. I have two. Maybe I'm not sure if you look at competitors and the supply/demand curve of the market. But if you do, could you just talk about where we are overall for new capacity, whether it's reciprocating engines, combined cycle gas turbines, or other technologies, relative to your own decision-making process? And then I have a follow-up, which is more macro.

KR Sridhar, Founder, Chairman, and CEO: Look, I think given the enormous supply-demand gap, there is room for every technology that can deliver power quickly in the next few years. Let's start with that. So if engine manufacturers and turbine manufacturers expand capacity, there will be demand. If Bloom expands capacity, there will be demand too.

Ben Kallo, Analyst: Okay.

KR Sridhar, Founder, Chairman, and CEO: But let's think forward, from a competitive standpoint, assume a customer needs to choose between a turbine, an engine, and a fuel cell, okay? First, the most important thing is not LCOE, which is an absurd metric for on-site power; the real metric is total cost per token, the total power cost relative to token revenue. Bloom's ability to provide 800-volt DC power, Bloom's ability to guarantee reliability without overbuilding, Bloom's ability to be sited anywhere in and out of cities because it doesn't pollute the air, and Bloom's ability to get permits—these are things no other competitor can do. There is no commercial supplier in the market today that can offer this complete value proposition except Bloom. So, Ben, our approach is—we don't obsess about competitors; we obsess about customers. Okay.

Ben Kallo, Analyst: Thank you, KR. And on the commoditization model or open-source models from China, I think a lot of people are concerned or uncertain about that. Could you talk about whether you see those as opportunities or threats for you going forward?

KR Sridhar, Founder, Chairman, and CEO: Look, whether it's from a lab in China or a lab in the US, it doesn't matter. As a technology optimist, I think it's inevitable—the cost of tokens will go down, the efficiency of tokens will improve, and the tasks tokens can accomplish will keep growing. Costs will go lower, token efficiency will skyrocket, and the ability of tokens to improve productivity will continue to grow. All of this will happen. It means the price per token will drop, but the total usage of tokens will explode, because that's Jevons Paradox. When that happens, you need more power, not less. So if anything, it will only accelerate things. All predictions for AI today are underestimations, not overestimations.

Ben Kallo, Analyst: Thank you.

Operator: Next, to save time, we will switch to one question per person. The next question comes from Manav Gupta of UBS.

Manav Gupta, Analyst: Hey, KR. You started this company a year ago with a vision for the product. Obviously, you've come a long way. I'm trying to understand, over the next three to four years, what is the vision for this product? Going back to your earlier comment about Henry Ford, the things you can control—where do you see the product heading in the next three to four years?

KR Sridhar, Founder, Chairman, and CEO: I think that's very clear. Imagine, our on-site power, DC (direct current) is going to be the primary power source—whether it's for data centers, or any other application, whether it's fleet charging for electric vehicles, or large apartment buildings and microgrids for residential areas, DC is the future of the world. So, generating DC power while also utilizing heat to provide both heating and cooling functions. On top of that, decarbonization will, in my view, become extremely important. And Bloom is better than anyone at carbon capture. So we will focus on how to provide an integrated solution—combined efficiency of over 90% fuel utilization, without polluting the air, without consuming water. The same solution, the same technology that customers use to power a large data center is the same one that will power a store near your house and power an inference data center near you. That's our vision. We want it to be like a home appliance, plug-and-play. That's where we are going.

Operator: The next question comes from Maheep Mandloi of Mizuho Securities.

Maheep Mandloi, Analyst: Hey, thanks for taking the questions. I have a question on capacity expansion over the next few years. One common thing we hear from other manufacturers in the industry is inflation in CapEx estimates. I'm curious about your view, knowing that your manufacturing equipment is different. So as you scale from 2 gigawatts to higher, how should we think about that? Thank you.

KR Sridhar, Founder, Chairman, and CEO: Oh, thank you very much for that question, because this is a key differentiator for us, right? Our factory investment returns in months. Okay. We are not from the old era. Okay. We are not an industrial-era power company. We leverage the same technologies that have driven consumer electronics and semiconductor devices to be produced in greater quantities and at lower prices for everyone globally while creating greater value. That's the model we are using, and it's the model we will stick to. So let others deal with their own problems. From our perspective, our investment to expand capacity has a payback period of just months, and as long as market demand exists, we will continue to add capacity. Thank you.

Operator: The next question comes from Sunaina Ocalan of Bernstein.

Sunaina Pai Ocalan, Analyst: Hi, team, thanks for taking my question. I wanted to ask about the competitive landscape, also a follow-up to the comments made on this call and to the previous question. I think what you are saying makes a lot of sense—the Bloom on-site solution doesn't require transformers, doesn't consume water, it's a superior solution, that's clear and sensible. So, for the next 12, 24, 36 months, how are you thinking about market share relative to other fuel cell solutions targeting the same data center market? For example, I get asked about molten carbonate fuel cells. If you could provide some color on market share, that would be great.

KR Sridhar, Founder, Chairman, and CEO: Look, how many megawatts or gigawatts they can install, that's for them to tell you, not for us to comment on. As far as the data center sector is concerned today, I think our market share is in the very high 90%. If someone wants to enter this market and thinks they can compete with us, competition is a good thing. Competition makes us hungrier, it makes us run faster, it keeps us on our toes, and we will thrive on competition. So, I welcome competition from anyone, everyone. Thank you very much.

Operator: The next question comes from Colin Rusch of Oppenheimer.

Colin Rusch, Analyst: Thanks so much. I wanted to dig into the data center business and the time to power advantage you highlighted. Can you talk a bit about the evolution of your thinking around pricing strategy and target margin profiles for the platform? And also give us a sense of how much of the business you are taking is replacing other technologies that were originally planned for these sites?

KR Sridhar, Founder, Chairman, and CEO: Look, we don't think in terms of LCOE pricing, because we are not a utility company. We are a strategic partner for our customers, and we create value for them. Based on that value, they should be happy to buy from us, and they should be happy to let us capture an appropriate return. So for us, it's not just a story of cents per kilowatt-hour; it's about the value we bring to them, it's about the total solution we provide. For us, it's critical to increase margins by creating value. You see, when we talk excitedly about growth, we often overlook a hugely important thing happening inside the company. I remember seven years ago, most of you—the same analysts—were only worried about our service losses; that was the only thing you cared about. And we kept telling you that the technology product would mature, and you would see us reach the 20% gross margin we promised. This quarter, we just announced a 22% gross margin. I want to take this opportunity to give the highest accolades to the team that has worked day and night to make this number a reality. Think about it, when we went public eight years ago, our gross margin was negative 21%; this quarter, it was positive 22%. A 43-percentage-point improvement in service margins. Now let me emphasize one more point. Service margin is a financial metric, but the first word is 'service.' Who are we serving? We are serving our customers. Ultimately, customer satisfaction is the most important thing. So we are not just achieving the financial metric. If you look at our 2025 data, 80% of the orders we received were from repeat customers. That speaks more than any other metric about how satisfied our customers are. So I am very, very proud of this achievement. I truly believe—and I'll end with this—I truly believe that the service part of the business, the service revenue and its associated margins, is a significant driver of our enterprise value, and more importantly, it demonstrates we are serving our customers the right way. I am incredibly proud of that team and give them the highest praise. If you combine everything I've just said—look at our backlog, market demand, how perfectly we fit the needs of the future digital world (unlike past technologies); look at our team's execution; and combine it all to feel the trust we are building in the communities and businesses we serve—I have a lot to be thankful for and I am so grateful for this fantastic Bloom team for doing a phenomenal job. Thank you, everyone.

Operator: Ladies and gentlemen, this concludes today's conference call. Thank you for your participation, and you may now disconnect.