Bloom Energy Corporation Q2 2026 Earnings Call

Good day everyone, welcome to the Bloom Energy second quarter 2026 earnings call. Just a reminder that today's call is being recorded. At this time, I would like to hand things over to Mr. Michael Tierney. Please go ahead. Thank you, good afternoon, everybody.

Thank you for joining us for Bloom Energy's second quarter 2026 earnings call. To supplement this conference call, we furnished our second quarter 2026 earnings press release and supplemental financial information with the SEC on Form 8-K, and have posted these materials, which we will reference throughout this call, to our investor relations website. During this conference call, both in our prepared remarks and in answers to your questions, we may make forward-looking statements that represent our expectations regarding future events and our future financial performance. These include statements about the company's business results, products, markets, customers, strategy, financial position, liquidity, and full-year outlook for 2026. These statements are predictions based upon our expectations, estimates, and assumptions.

However, as these statements deal with future events, they are subject to numerous known and unknown risks and uncertainties as discussed in detail in our documents filed with the SEC, including our most recently filed Forms 10-K and 10-Q. We assume no obligation to revise any forward-looking statements made on today's call. During this call, in our second quarter 2026 earnings press release and supplemental financial information, we refer to GAAP and non-GAAP financial measures. The non-GAAP financial measures are not prepared in accordance with U.S. generally accepted accounting principles and are in addition to, and not a substitute for or superior to, measures of financial performance prepared in accordance with GAAP. A reconciliation between the GAAP and non-GAAP financial measures is included in these materials, which are available on our investor relations website.

Joining me today on the call are KR Sridhar, Founder, Chairman, and Chief Executive Officer, and Simon Edwards, our Chief Financial Officer. KR will begin with an overview of our progress, and then Simon will review financial highlights for the quarter. After our prepared remarks, we will have time to take your questions. I will now turn the call over to KR.

Good afternoon, everyone, and thank you for joining us. It took Bloom 21 years to deliver its first billion-dollar year in 2022. It took us another three years to double our 2022 revenue. Now, we are guiding to double that revenue in just one year, having achieved our first $1 billion quarter. Bloom Energy's business is accelerating. We are expanding at breakneck speed and continue to grow more profitable. We have successfully demonstrated the earning power of our agile business model, doing exactly what we built it to do. Through all of it, we kept the promise that we make to our customers. Bloom will not be your bottleneck. We deliver power at AI speed and enable our customers to grow. This makes us a vital strategic partner and builds customer loyalty. Demand is compounding. New customers are arriving at a faster pace than ever.

Landing a mega customer used to take us years, a moat that protected the incumbents. Now, the proven success of our technology, coupled with an increasingly urgent need for efficient, clean, and reliable power, has collapsed the time from first engagement to first order. Because we can book, ship, and convert orders to revenue inside the same fiscal year, that demand shows up in results now, while also adding to and diversifying our backlog. Notably, just this year, several customers who had alternative solutions in place abandoned them and came to Bloom. Once customers are in, they see the total economic value of our solution and place more orders. We have booked major customers as of today who were not in our reported backlog at the end of last year, for whom we will ship some of our systems this year.

These customers are now placing longer-term orders, leading to our backlog growing at a faster pace than revenue. Let me repeat, leading to our backlog growing at a faster pace than revenue. Bloom Energy has emerged as a standard for on-site power, as we predicted we would in our third quarter call last year. At that time, we had just announced our first direct hyperscaler customer, Oracle, and delivered them power for a data center within 55 days. Today, all the major U.S. hyperscalers and over a dozen U.S. Neo Clouds, AI labs, and co-location data center operators have validated and approved our power solutions for their AI factories. Our commercial and industrial business continues to grow. We are the standard for on-site power for hospitals, factories, telecom providers, university campuses, and retail stores. It took us nearly a decade to become the accepted solution in these verticals.

Contrast that to AI data centers, where we have become a standard in less than one year. I founded Bloom on the conviction that on-site power would be essential to powering the world and ushering in the digital transformation. We have built our company to offer the best on-site power solution that removes friction for our customers. We are clean and reliable and fast and affordable. Customers do not have to choose or compromise. Over time, we are reducing friction and turning headwinds to tailwinds. Let's take a moment to discuss four friction points: capital, community, permitting, and speed. Start with capital. For a century, the cost of power plants and the grid was spread across millions of ratepayers and amortized over decades. A new load simply plugged into surplus capacity and only paid a monthly bill. Surplus grid capacity is now gone.

New data center load now means new infrastructure, heavy capital, long lead time, and ratepayers who will not fund or subsidize capacity for a corporate customer. The faster, cheaper, more predictable path for that customer is islanded on-site power. That solution demands either a capital budget most end customers do not have, or a financing partner behind a power purchase agreement. Negotiating bespoke terms across financiers, developers, operators, OEMs, and end users is complicated and slow. We kill that friction with strong financial partners who provide our customers project capital. Brookfield anchors that financing shelf. We formed the partnership last fall at $5 billion. Nine months later, in June, Brookfield expanded its commitment fivefold to $25 billion.

One of the largest and most experienced infrastructure investors in the world evaluated our technology, our delivery record, and our pipeline, backed us with $5 billion, watched us execute, and then multiplied that backing by 500%. Capital of that quality and quantity does not follow letters of intent, MOUs, or press releases. It follows performance, happy customers, and firm bankable orders. Brookfield is not alone. This quarter, Industrial Development Funding, who has previously funded Bloom deployments, partnered with Oaktree, MUFG Bank, and Morgan Stanley to fund Bloom deployments, cumulatively bringing their total commitment to $2.6 billion. More financing partners are in the wings. Gigawatt demand needs gigadollars of capital. We arranged that in advance. Next friction point, community. Communities have learned they cannot live with combustion. However, they've also learned they can live next to Bloom Energy Servers.

No combustion, negligible air pollution compared to turbines and engines, negligible water use. A power server that is aesthetically appealing and runs quieter than air conditioning equipment. No construction project is NIMBY-proof, Communities welcome Bloom, and that is proving to be a real competitive advantage. Our customers can get air permits with our technology faster than they can with a combustion alternative. Every month saved on permits is a month closer to power availability. Which brings me to speed, because time to power is really time to token revenue. Bloom is increasingly seen as the solution that eliminates the power availability friction point for AI. Chips without power are inventory, not intelligence. Grid operators quote years, timelines on which billion-dollar compute clusters go obsolete in a warehouse. Legacy suppliers celebrate backlogs stretching to 2029 and beyond. We think a four-year backlog is not a trophy.

It's a confession of constrained supply. By contrast, Bloom meets the time-critical needs of our customers and delivers power in months. Since the beginning of the year, we have been continuously adding to our American manufacturing capacity in Copy Exactly! increments and will continue to do so ahead of committed orders. Our speed is dependable because we have built a resilient supply chain with broadly available materials by design, multiple qualified suppliers across multiple countries for every critical input, inventory ahead of the ramp, relationships forged over two decades. No single supplier and no single country determines our destiny, and every part of our supply chain is prepared to scale with our growth. When the business case is measured in months of AI compute. The fastest, dependable path to power wins. We are on that path, and nobody else is close. None of this is by accident.

We built scale into our model years ago for a simple reason. If we were right about the market need for on-site power, the company had to be able to scale rapidly. Now, you're seeing that play out in practice. Capital, community, permitting, speed. Remove all four frictions and the market renders its verdict. Supplier of choice. I choose those words deliberately because customers are choosing. Customers who had ordered combustion turbines and reciprocating engines cancel those orders and choose Bloom. Nebius did it this quarter. Existing customers are coming back with expansion opportunities, and the largest infrastructure investors in the world are underwriting our deployments at scale. Trusted partner to customer, to the community, and to capital. The market has noticed. Customers now come to us late in their development process and ask Bloom to step in as the primary on-site power solution.

The entry point differs, the outcome does not. Once they see our capability, our execution, and the total value we deliver over the life of ownership, the conversation expands from one project to a portfolio. Let me be very clear. Bloom Energy is not dependent on one customer or one project. It is multiple customers and multiple projects across every stage of development. Because our Copy Exactly! Lego block servers redeploy from site to site painlessly, unlike bespoke traditional equipment, every project in the portfolio is fungible. Diversity, fungibility, and nimbleness allows us to navigate the fast-changing AI landscape. Consequently, we have visibility and conviction in our growth trajectory for 2026 and beyond. Let me close with how we run this company, because I know what's on your minds. Will AI investment keep growing at the breakneck pace?

Our engagement with customers suggests that the pace of investment will not just continue, but accelerate. However, I do not know for sure, and I will not insult you by pretending to. Henry Ford could not control whether America wanted to drive. He controlled the cost, quality, and availability of the Model T. Like Ford, we are focused on managing the controllables. We drive costs down every single year, while others benefiting from these build-outs talk only about raising prices. We innovate continuously to better serve our customers and meet every commitment we make. We are experiencing the tailwinds of rapid TAM growth and increasing market share simultaneously, and we are grateful we can meet the market's demands. At the same time, we are building a durable advantage by earning the trust of both our customers and the communities.

With that, let me hand it to Simon to take you through the numbers. I'll rejoin you for Q&A. Simon? Thank you, KR. Good afternoon, everyone.

This is my second earnings call as Bloom's CFO and my first with a full quarter behind me. In April, I told you why I joined Bloom. Belief in the mission, the architectural shift in power, the quality of the team, and the chance to help build a generational company. 3 months in, that conviction has only strengthened. The demand environment is robust, and we are neither gated by capacity nor our supply chain. The operating discipline is real. The factory improves every week, and cost out is a rhythm here, not a program. In finance, we have a strong foundation, and my focus is on scaling our systems and processes to keep pace with our rapid growth. Before the results, let me come back to the Brookfield announcement KR discussed.

A 5-fold expansion of our strategic partnership as Brookfield increases its framework to finance Bloom power projects for AI infrastructure. A commitment of that size from one of the world's largest infrastructure investors is significant on its own. It is also the best way into our revenue model because you cannot fully appreciate what that partnership does without understanding the model behind it. Given our shareholder base has grown considerably this past year, let me take a minute to walk through how a Bloom deal works. Every Bloom deal starts with a contract with the end customer, the party that will actually use the power. That contract takes one of two basic forms. Either the customer buys the equipment outright, which is a CapEx sale, or they contract for power or capacity without owning the equipment.

As our 10-K describes, the second form comes in a few flavors. A power purchase agreement priced per kilowatt-hour, a capacity agreement, or an equipment lease priced on installed capacity. Economically, they all work the same way. The customer pays over time rather than owning the asset. In any case, every contract carries committed commercial operation dates. Since most customers choose to pay over time rather than own, we bring in a financier, an institution like Brookfield, that purchases the energy service from Bloom, owns them, and delivers to the end customer under the contract we originated. When you see the word customer in our filings, it can mean either party, the financier that buys from us and appears in our revenue and concentration disclosures, or the end customer whose demand created the deal.

The IDF partnership that was recently announced is another live example of that model, and it was a meaningful contributor to the quarter. Mechanically, it is our standard structure. NEB has signed the offtake, and IDF, an independent third party, is purchasing the energy service on cash terms against identified sites and delivery schedules. One feature of this model worth understanding, especially for newer investors, is that large campus deliveries are lumpy. One or two customers could lead our revenue in any given quarter, and the customers rotate as different projects reach their delivery windows. Revenue can look concentrated in any single quarter. That concentration reflects delivery timing, not the composition of our backlog. The backlog spans multiple hyperscalers, neo clouds, colocation providers, and commercial and industrial operators, and our contracts carry payment security appropriate to the transaction size. To the results. As a reminder, I will focus on non-GAAP adjusted metrics.

A full GAAP to non-GAAP reconciliation is in the press release and the supplemental deck on our IR website. Revenue was $1.065 billion, up 166% year-over-year and 42% sequentially. This was another record quarter for Bloom, and the first time we exceeded $1 billion in quarterly revenue, reflecting accelerating data center deliveries and disciplined execution converting signed demand into revenue. Product revenue was $935 million, up 215% year-over-year and 43% sequentially, and represented nearly 90% of total revenue in the quarter. Gross margin was 34.3%, up 604 basis points year-over-year. The improvement reflected both favorable mix and margin expansion across both product and services. Product gross margin was 37.2%, up 193 basis points sequentially and up 291 basis points from Q2 2025. A word on price. We deliver value, not a commodity kilowatt.

Our customers are paying for time to power and what the solution can do, whether that is following the load profile of an AI campus or being ready for carbon capture, and our pricing reflects that. On cost, we continue to drive product cost down across material, labor, and overhead, and we are doing that while ramping up our capacity and adding new capabilities. Services margin was 22%, up 977 basis points year-over-year, and our fifth consecutive quarter of double-digit service margin. Service revenue is recognized ratably, net of guarantees, while the service costs are booked as incurred. The timing of fleet maintenance, stack replacements in particular, moves the margin quarter to quarter. Underneath that timing, margins have now reached the 20%+ level, driven by fleet performance, longer stack life, and scale, and we believe we will sustain them there over the long term.

Blended gross margin will move quarter to quarter, driven by the pricing mix of projects delivered in any given period, deliberate trade-offs between cost optimization and expediting, where the customer's time to power is worth more than the incremental cost, and the service maintenance timing I just described. Across all of it, we continue to feel good about our full-year gross margin outlook of approximately 34%, which we revised up last quarter. Operating income was $240 million, up 737% year-over-year, and operating margin was 22.5%, an expansion of approximately 1,536 basis points. Adjusted EBITDA was $253 million, approximately 24% of revenue. non-GAAP diluted EPS was $0.78, and GAAP diluted EPS was $0.62. The increase in profitability reflects both higher volume and substantial operating leverage as the business scales.

To that point, the operating leverage in these numbers is notable, and it is structural, not a one-quarter effect. Revenue grew 166%, while operating expenses grew just 48%, and the mechanics behind that should persist. The leverage comes from how we have built our cost structure. Our R&D base and our G&A infrastructure are largely fixed against a rapidly growing revenue base. Each incremental gigawatt of deliveries carries little incremental overhead. We run our support functions the way we run our factories, using automation and analytics across SG&A, service operations, and supply chain. Those functions grow with technology rather than people. We will continue to invest across the business, including G&A and R&D, but we expect operating expense growth to remain well below revenue growth, which should drive continued operating margin expansion.

Cash flow from operations was $226 million, an increase of $439.5 million from the same period last year, driven by profitability and favorable working capital performance. Free cash flow was $175 million, and we ended the quarter with $2.7 billion of cash. Generating significant operating cash flow while growing revenue at this current rate truly reflects the strength of the underlying business and the working capital discipline of the team. Now turning to guidance. On the strength of our year-to-date performance, as well as the visibility and confidence we have for our second half, we are raising our full-year revenue outlook to $3.9 billion-$4.2 billion. At the midpoint, that represents 100% growth over 2025 revenue of just over $2 billion. Our outlook is built bottom-up in two layers. The foundation is backlog conversion, signed commitments delivering against customer site readiness dates. The second layer is in-year bookings.

We intentionally reserve manufacturing capacity for time to power customers who need power in months, and for whom we expect to book and convert into that capacity at rates consistent with recent experience. On gross margin, we are holding our full-year margin rates at approximately 34% on a non-GAAP basis. I'll also use this as an opportunity to make one point on how we run the business. When we have to choose between protecting a point of margin in a given quarter and expediting to deliver an order quickly to a customer who is going to be with us for years, we are going to 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 are going to keep delivering against that.

Over a full year, that discipline is fully consistent with the margin rate we are guiding to. On operating income, we are raising our full-year non-GAAP operating income outlook to $800 million-$900 million. At the updated revenue midpoint, this implies an operating margin of approximately 21%. That is a material step-up from the $425 million-$450 million operating income guide at the beginning of the year, which represented a 14% margin at the midpoint. This is the operating leverage I described earlier, flowing directly through the model. Full-year non-GAAP diluted EPS outlook is now expected to be $2.55-$2.85. Finally, a word on how to read our guidance. Demand in the AI business does not follow the traditional selling cycle. Today, we see the demand, we book the order, and when the customer is ready, we ship the equipment.

Both the level and shape of our outlook come from the same inputs, which are signed commitments and their schedules, our capacity, and our time to power pipeline. What was once seasonality is now simply delivery timing based on customer readiness. To conclude, this was a milestone quarter. We exceeded $1 billion in quarterly revenue for the first time, delivered record profitability, generated strong cash flows, and raised our full-year outlook. We are executing with discipline against a demand environment that keeps getting stronger. With that, operator, we are now ready for questions.

Thank you. Ladies and gentlemen, at this time, we will take your questions. If you have a question today, please press star one on your telephone keypad. To start out, we do ask that you keep your questions to one initial and one follow-up. Our first question comes from Mark Strouse, J.P. Morgan. Yes, good afternoon. Thank you very much for taking our questions.

KR, I want to go back to your comment about all of the major U.S. hyperscalers and over a dozen other operators are now, excuse me, validated and approved to use your technology. Are you able to talk about a bit more how many of those are actively using your technology today, and how many of those are kind of in your backlog or in your near-term pipeline? Then I've got a quick follow-up. Thank you. Mark, as you very well know, we let our customers speak about the deployments and what they do.

What I can tell you is the combination of the three things that you spoke about, customers already using it, customers who booked orders, and we have shipped units to them for whom the power will be delivered and is in construction, and customers who have given us definitive agreements. They fall in that category. We are not going to split that up, but it is all the major, as you pointed out, it's all the major U.S. hyperscalers over a dozen of the neo clouds and the ecosystem around it, co-location partners. That is all true. We don't break it up. Again, let's just take a moment to think about this.

Nine months ago, we announced our first direct hyperscaler customer and said we want to enter into this market and do what we did in C&I and become a standard. It took us 10 years to do that previously for commercial and industrial. I can tell you, I never thought within nine months we would become the standard. It just speaks to our value proposition in this entire industry. Okay. This is not a faster horse. This is a car. That is why this is happening, and this is not reversible. Thank you. Okay. Very helpful.

If I can sneak one more in. I know going back to the last call, you don't want to get into specific comments on capacity anymore, but maybe just relative to the last call, just given your comments today several times about things accelerating, is it safe to assume that maybe your timeline or your magnitude of capacity that you're planning is accelerating as well?

Yeah. Here's how we do the capacity planning, right? It is very clear to us based on our commercial pipeline and our commercial orders, when our customers need their products, and when they're ready to turn the units on. As you understand, very clearly, depending on whose reports you read, somewhere between 30 and 40 gigawatts of new AI data center capacity is going to be turned on in 2027, right? They are in various stages of development. They're all greenfield. We have a very sophisticated algorithm that we use to figure out how many of those projects are going to land when. Luckily for us, unlike everybody else, our units are fungible. When they're on a truck, we can redirect them to a different site if we need to. That's how fungible our systems are.

Based on that, we can predict what our capacity needs are going to be such that we never become the bottleneck to the customer. I can tell you, as we sit here today, we are confident of keeping that promise for everything we have in the order book and everything that we see coming forward. That's all we are going to comment on our capacity. Capacity is not going to be our constraint as we see right now. Thank you. Your next question comes from Chris Dendrinos, RBC Capital Markets.

Yeah. Thank you, and congratulations on the strong quarter. I wanted to ask about the supply chain here. I guess, when you're having conversations with hyperscalers, what are they asking you, and what are you telling them or responding to them to give them confidence that you won't be the bottleneck in terms of delivering on time? Thanks. That's a very good question.

Look, you are raising something very important. These are extremely sophisticated consumers and customers. The amount of diligence that they go through is not just about our product, not just about our performance, not just about our economic value proposition. They want to understand where we are with respect to committed orders, where we are with respect to new orders that can come online. Look, most of them, if you talk to them, are not 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 for the future.

They want to understand as they share confidentially with us what their capacity expansion plans are and ask us if we can meet that. We have to walk through with them, under NDA, in great detail and convince them that we will be able to scale. That's when we get validated. To that previous question of what does all that validation mean? That's the process we go through. It's a pretty rigorous process. It's a pretty rigorous process with each of those customers.

Got it. I guess, maybe fair to say that you go through that same process with IDF and Brookfield. Maybe just as the follow-up here, on the Brookfield topic, you expanded that partnership by $20 billion. How should we think about the timing to execute on that? Is there sort of a window to which you all think about being able to execute on $20 billion of transactions? Thanks. Sure. A great two-part question you snuck into one question.

I'm happy to answer both of them. They're very relevant, right? The first part is, look, at the end of the day, the financial customers take title of our equipment. When they take title of our equipment, it is not just about can they place it in service. They care about how will it operate through the entire period to be able to deliver. They go through additional layers of our ability to perform, our ability to be available, and can we upkeep that equipment for the time that is needed based on the financial model that they came up with, so they can actually get their returns. It is two additional layers of detail that they would go through in their diligence, and they go through the same other process that we discussed.

Even in today's world, $20 billion is a lot of money. Very obviously, they go through this process deeply with us. Again, remember this, they didn't come into it in first step. They put the $5 billion in. They watched how we perform. They watched what we are capable of. They watched the execution. They spoke to several of our customers, the oldest of them now being more than 15 years our customer, to understand how we perform and how satisfied they are. Happy customers is absolutely essential. It is on the strength of all that they come and invest. As to the timing, look, think of this as a financial shelf. This shelf is now available. The pace at which it will get used up is going to depend on the uptake of those funds.

I think I can again say it's very similar to us becoming the standard for AI in less than nine months. I wouldn't have predicted nine months ago when they invested $5 billion, we'll come back that quickly for the next $20 billion. This just speaks to the pace of acceleration, both in AI and in our business. Thank you. David Arcaro from Morgan Stanley has the next question.

Thank you so much. Appreciate you taking my questions. There have been some project development challenges that have gotten headlines at a few large projects. I was wondering if you could characterize your financial exposure to project delays, any contractual protections that you typically have in place, and alternatives that you may work on with customers.

Thanks for the question. Look, I think we don't comment specifically on individual projects, as you know. If I take a step back and just think about how we contract, right? Our contracts are structured with master service agreements, and there's flexibility for us to deploy this equipment, given its Copy Exactly! nature, to various customer projects. With that said, we have strong protections with our contracts. Then ultimately, our financiers also need those same protections, right? As you think about the way these contracts kind of flow, to the extent that there are any project delays, the end customer is able to redeploy that equipment to other projects. Ultimately, the financier is on the hook to take delivery of the equipment from Bloom.

The other thing, extremely important point to talk about in terms of projects, because it would be on the minds of people. We can tell you when we gave you this guidance, and we upward revised it, right? The 2026 revenue guidance is not dependent on any single project. Again, we have a sophisticated algorithm, and we expect certain projects to push out, certain projects to come in, certain projects to just come out of the blue and absorb it in the same year as we described in our script. We take all that into account when we give the guidance. Construction projects, as long as construction projects have existed, I'm sure there have been delays. Okay? I'm not a historian, but that's what I would expect.

We should bake that in, but that's not going to affect our revenue guidance because we have a sophisticated algorithm to figure that out for the year. Our 2026 guidance is not going to have any dependence on any single project.

Yep. Understood. Yeah. Thank you both for that extra color. That's helpful. KR, appreciate your additional comments on the supply chain and your confidence there. I was wondering if you could address your access to scandium, which has gotten a lot of attention. Is there any way that you might characterize your use of scandium, how much you see as available supply and stockpile, et cetera?

I think we have put out a detailed blog on this topic, and we have filed the 8-K. The three takeaways for you all as investors to understand are the following. There is enough scandium on the planet that can be recovered economically viably to power the planet. That is what's available on the planet. We have visibility currently based on what we are working for 25 gigawatts of deployments. We are not dependent on China. Those are the statements we made. That's all we are going to say about it. Everything else is proprietary to the company. Next question. The next question is from Nicholas Amicucci, Evercore ISI.

Hey, guys. How are you? Simon, sorry, I'll put you on the hot seat for a little bit. Obviously impressive raise in guidance On a strong 2Q, but the free cash flow guide was pulled. I just want to get some context around that and just considering that you have $2.7 billion worth of cash on the balance sheet, just thinking through capital allocation.

Yeah. Hey, Nick. Thanks for the question. Just to level set here, the company has historically included a supplemental presentation that has included some metrics that are not formal guidance. We've just really aligned now the presentation to where we truly guide. To come back to your question on cash, which I think is a relevant one. We see significant conversion from operating income down to free cash flow. As you think about where we started the year with op income guide of the midpoint of $450 million, with $200 million of CFOA, we raised that to $675 million, and now we're looking at $850 million at the midpoint. There's about a $175 million raise versus our prior op income guidance, and we see 100% of that dropping down to CFOA. You should think about $375 million plus as being our new baseline.

As you know, we don't provide formal guidance as part of our earnings release. I just want to make sure that you get the comfort that we see strong conversion here.

Got it. No, that makes perfect sense. As we think about just, I guess, longer term AI demand and just deal like why the CapEx from every hyperscaler right now is so high, and we think about where the returns are actually going to be generated. As we think about that and then think about your backlog and kind of the conversations you guys are having, KR, are you starting to have any conversations just along the lines of inference reasoning, or is it still really just kind of speed to power on AI training?

Yes. It's both. It's absolutely both. Look, time to power is super important, number one. Okay. Let me try to explain time to power in a slightly different way, because for many of the analysts who look at utilities and power, as well as some of our investors, it's important for you to understand. A full stack AI provider that is responsible for everything in the data center financially, the 1 gigawatt data center in one single year, depending on the nature of the AI customer, will deliver between $12 billion-$24 billion in revenue per year. You pull in power for them within a month, which is the tall pole, that is $1 billion-$2 billion of revenue that they would not have had, on a 40%-50% gross margin and a 20%-25% net margin.

Among the 35-40 gigawatts that needs to get deployed next year, take a guess on how much of those projects will get delayed because the power provider is not able to provide power on time. Okay. We are the place to go for time to power. The math is, you don't even have to do the math if we can provide that. This is why time to power is important for the large data centers. Now, as inference comes along, if the transmission distribution infrastructure in the country is having difficulty doing transmission, building highways, imagine how difficult it'll be for them to upgrade distribution, which is surface streets. That's where inference power is going to be needed. Bloom is ideally suited for that. You cannot put a gas turbine in the middle of Manhattan.

We see both opportunities as extremely robust for us, not this quarter, not next quarter, for years to come. Thank you. Your next question is from Ben Kallo from Baird.

Hey, thanks for taking my questions. I have two. I don't know if you guys think about your competition and supply-demand curves out there. Maybe, if you do, could you give us your thoughts about where we sit just in overall new capacity coming online, whether it's reciprocating engines or it's combined cycle turbines or other, versus your decision process? I have a follow-up which is kind of at a higher level.

Look, I think given the huge supply-demand gap right now, I think every single technology that can provide power quickly for the next few years is going to have a play. Let's just start with that. If engine makers, turbine makers increase their capacity, there's going to be a need for it. If Bloom increases its capacity, there's going to be a need for it. Let's fast-forward and think through competitively at a point, if a single customer has to choose between a turbine, an engine, and a fuel cell. Okay? The first thing that matters is not LCOE, which is just an absurd construct for on-site power. It is the total cost of power to token revenue.

Bloom's ability to provide 800 volt DC power, Bloom's ability to provide reliability without overbuild, Bloom's ability to be able to locate inside a city, outside a city, because we don't pollute the air. Bloom's ability to get a permit. Okay? None of the other competition, there is not a single commercial vendor today who can provide that total value proposition other than Bloom. What we do then is we don't obsess on the competition. We obsess on the customer. Okay? Thank you, KR. Just maybe from a commodity model or Chinese open source models that have, I think, people worried or concerned, can you just talk to us about if you view that as an opportunity, either of them, or a threat going forward?

Look, whether it is the Chinese, whether it's the American labs, it just doesn't matter. I would say, as a technology optimist, to me, it's a given that token cost and token efficiency and token's ability to do things are all going to improve. In that, cost will get better, cheaper. Efficiency of the token will get much better. What a token is able to do with respect to productivity will keep increasing. All of them will happen. That means the token price point will come down, but the total token usage will go up like crazy because that's Jevons paradox. When that happens, you need more power, not less power. If anything, this is going to accelerate. If anything, whatever we are predicting on AI is an underestimate, not an overestimate. Thank you. Everyone, at this time, in the interest of time, we will switch to just taking one question.

Our next question is from Manav Gupta, UBS.

KR, you've started building this company 21 years ago. You had a vision of the cell, obviously you've come a long way. I'm trying to understand what's the vision of the product for the next three to four years, coming back to your primary comments, what you can control in context of Henry Ford. Where do you see this product moving in the next three to four years?

Look, I think very clearly, imagine our onsite power. DC is going to be the primary source, whether it's data centers, whether it's anywhere else, whether it's fleet charging of electric vehicles, whether it is large apartment complexes and microgrids being built for residential complexes. DC is where the world is going to go, right? Predominantly. DC power being generated on-site, being able to use the heat to do both the heating and the cooling, then on top of that, decarbonization, in my view, is going to become super important. Bloom is able to do carbon capture better than anybody else. Okay? We are going to be focused on how do we give a comprehensive solution where the fuel is getting utilized at 90+% efficiency, combined efficiency. It is not polluting the air. It's not using water. The same format that a customer gets used to to power large data centers is the same format, the same technology that powers your neighborhood store, that powers your neighborhood inference data center.

That is the vision. We want it to be like an appliance that you can plug in and get power. This is where we're going to go.

Next up, Maheep Mandloi from Mizuho Securities has the next question.

Hey, thanks for taking the question. One just question on capacity expansion for the next three years. The one theme we're hearing from other manufacturers across other industries is inflation on the CapEx estimates. Just curious on your thoughts on these. I know your manufacturing equipment is different, so how should we think about that now as you expand from 2 gigawatts to beyond that? Thank you. Thank you so much for asking that question because that's a significant distinguisher for us, right?

Our factories, the return on investment is a few months. Okay. We don't come from the old world. Okay. This is not your industrial age power company. We are relying on the technologies that made consumer electronics and semiconductor devices become better available in larger quantities to everybody on the planet, and get lower in cost and provide greater value. That's the model we are adopting. That's the model we're going to follow. Let them deal with whatever they're dealing with. From our perspective, for us to expand capacity, the return on investment is going to be a few months, and we'll keep adding capacity as long as the demand is there. Thank you. The next question will come from Sunaina Ocalan from Bernstein.

Hey, team. Thank you for taking my question. Maybe if I can ask about the competitive landscape, and just on a follow-up on the comments that were made on the call as well as somebody's question ahead of mine. It makes sense. I think what you're saying makes sense, which is on-site power, Bloom solution, not having NOx and SOx, not using water is a superior solution. That makes sense, and it's clear. How are you thinking about the market share over the next maybe 12 to 24 to 36 months on some of the other fuel cell models that are targeting essentially the same data center space? I'm getting asked about molten carbonate, for instance. If you can just provide some color on the fuel cell market share, that would be great.

Look, I think it's for them to tell you how many megawatts and how many gigawatts they can install, okay? It's not for us to comment to you. Today, within the data center space, I would say we would be in the very high 90s in terms of our market share, okay? If somebody is going to get in and think that they're going to compete with us, competition is a very good thing, okay? Competition makes us hungrier, competition makes us run faster, competition makes us paranoid, and we will thrive in competition. I welcome competition from anybody and everybody. Thank you. The next question will come from Colin Rusch, Oppenheimer.

Thanks so much. Gus, as you get into this a little bit deeper with the data centers and have those time to power advantages, can you talk a little bit about evolution of your thinking on pricing and target margins for the platform, as well as just giving us a sense of how many projects you're selling into where you're displacing a different technology that was originally planned for those sites?

Look, again, we don't think about LCOE price of electricity because we're not a utility, okay? We are a strategic partner to our customers, and we deliver value to them. Based on that value, they should be happy buying from us, and they should be happy allowing us to capture value. It is not just a cents per kilowatt hour story for us. It's about the added benefits we bring to them. It's about an entire solution that we look at. For us, being able to increase these margins by capturing value is extremely important. Look, when you think about this, we are so excited about talking about growth, we forget to highlight something extremely important that's happening in the company. Okay? I remember seven years ago, most of you, the same folks, the analysts, the only thing you worried about is our service losses. Okay? That's the only thing you worried about.

We kept telling you the technology product is going to get robust, and you will see us get to the 20% gross margin that we talked to you about. We just reported this quarter a 22% gross margin. I want to take this opportunity to give a huge shout-out to the team out here that's worked tirelessly to go make that number happen. Think about it. When we went public eight years ago, minus 21% gross margin, negative 21%. This quarter, plus 22%, 43% swing in service margin. Let me focus on one more thing. Service margin is a financial metric, but the first word in that is service. Who are we serving? We're serving our customers. At the end of the day, it's happy customers that matter. We are not just achieving that financial metric.

If you look at 2025, 80% of the orders that we booked were repeat orders from customers who have given us multiple repeat orders. That speaks more loudly than anything else about how happy our customers are. I'm very, very proud of that accomplishment. I truly believe, and I'm going to close with this, I truly believe that that service part of the business, and the service revenue and the margins, is a big driver to our enterprise value, and more importantly, how we serve our customers in a proper way. Very proud of that team. I want to give a huge shout-out.

If you combine what I just told you by looking at our backlog, the demand out there, how we fit perfectly to the needs of a future digitized world, unlike previous technologies, look at how we are executing as a team, and then combine that all together to see the trust that we are building within the communities we operate and the companies we serve. I have a lot to be grateful for and lot to be appreciative of a great Bloom team that has done a phenomenal job. Thank you all. Ladies and gentlemen, this does conclude today's conference.

We would like to thank you all for your participation.