The rush for clean, on-site power [partner content]

0:02This is the Frontier Forum, brought to you by Latitude Studios.

0:13Before fuel cells, Aman Joshi spent more than two decades at GE. A lot of that time was focused on aeroderivative gas turbines, the kind used for peaker power plants. And that gave him a unique window into the coming supply crunch. Even in my previous life, I was starting to see constraints. Years before turbine shortages became an industry-wide crisis, Iman could see signs. First, it was demand for peakers to support renewables. Then came the data center rush. And as orders for turbines backed up, Iman knew firsthand that the industry would have a hard time keeping pace. When you look at their tier two, tier three, tier four level of supply chain, they're all threaded to same suppliers and same sources.

0:56I knew that scaling supply chain up for, be it reciprocating engines or turbines, is going to be a long drawn out process. that is not going to happen as quickly as the power needs are growing. Today, Amon is the chief commercial officer at Bloom Energy, the leading maker of fuel cell systems for on-site power. And the supply crunch he saw coming has created an enormous opening for the company. Bloom can manufacture and deploy its modular systems much faster than conventional power plants, and without waiting years for a grid connection. In 2024, Bloom started surveying customers about how their energy strategies were shifting.

1:31And since then, interest in on-site power has nearly tripled. And at that point, there's about 13-ish percent of people who were acknowledging that on-site power generation will be needed by 2030. You fast forward, that number rapidly grew. In April this year, when we did the survey, it was more like 35 percent. And then it's also interesting when we asked them the same question, saying, hey, how many of you actually think that on-site power will be needed by 2035? 50 % of them said yes, that we're planning our own energy needs based on on-site power. But moving power generation on-site, and hence closer to customers, creates a new challenge.

2:12During the opening innings of the data center boom, developers rushed to lock down electricity in any way they could. Then communities began pushing back. And suddenly pollution, water use, and noise matter a lot more. So when you think about on-site power, addressing these community aspects becomes really important. And that is where, quite frankly, fuel cells have just a science-based advantage because we don't combust. You know, our emissions level are the lowest that it can be. Virtually no NOx or SOx. We don't consume water, except for startup. When we are in an operation mode, it's actually fuel cells self-generate water.

2:52And our lowest levels are the lowest. With the grid hitting its limits, clean on-site power generation is becoming a must-have for a wide range of CNI customers. And for many of those customers, it's not just a temporary fix. It's a permanent feature separate from the grid. And so as a result, it shouldn't be of a surprise to anyone that as on-site generation and distributed generation picks up, I fundamentally believe Bloom Fuel Sales is going to be the dominant technology that gets deployed in this market segment. I'm Stephen Lacey. This week, a conversation about how a scramble for electricity is changing how some of America's largest energy users get their power.

3:37This conversation with Bloom Energy Chief Commercial Officer Aman Joshi was recorded live as part of Latitude Media's Frontier Forum series. Data centers are driving a huge share of America's electricity demand, but they're not the only companies trying to connect new loads to an increasingly constrained grid. Chip fabs, battery factories, and other industrial facilities are competing for the same connections, equipment, and labor. It is becoming a zero-sum race for power. In this conversation, Eman explains why more customers are planning to generate electricity on-site permanently, why the familiar cost-per-megawatt-hour calculation can be misleading, and why speed, reliability, and community acceptance are central to power decisions.

4:25Let's talk about how this on-site power is going to get used. What you found is actually that a lot more customers are seeing this as permanent power rather than bridge power. So how are you seeing that change strategically from customers? Absolutely. Again, it goes back to that fundamental assumption that if the grid isn't available or is going to be available maybe in 5-10 years out, but the economics of what the pricing is going to be is going to be very high compared to what it is today. the customers are increasingly wanting to take control of their energy costs because it is a critical input to their own manufacturing costs.

5:05And as a result, a lot of them want a predictable energy cost. And because of that, we're seeing an increasing adoption of on-site power generation. And is it simply because of long interconnection timelines, or are there other factors that are contributing to this? I think there's also other factors. So as it relates to data centers, you see, you need a very highly reliable source of power that gives you very high reliability and availability, almost four to five nines availability, right? And that is also fuel cells, bloom fuel cells, inherently have a science-based advantage because we are a solid state device.

5:52We don't have any rotating equipment. We can enable either data centers or consumer or industrial customers to have the confidence that they can generate their own highly reliable power that will not have any outages. And are they thinking that they may get an eventual grid connection five, six, seven years down the road? Or what are their expectations for eventual grid connection for some of these customers? It's going to be a longish answer. So short answer is most of the customers are going to Bloom Solution as a permanent power. They're not thinking of it as a bridge. And for understanding that mindset, you also have to understand that there are places where we are cheaper than the grid.

6:41Why is that? Because we don't have any transmission or distribution or anyone else in the value chain that you have to deal with. So it's essentially converting natural gas molecule into electron in a highly efficient manner on site. So when you think about the grid today and you look at the delivered cost of the electricity, you actually have to add various elements to it to really get to what would be the future delivered cost of electricity. What do I mean by that? You know, a lot of time people, Stephen, talk about, hey, you can get six cents, eight cents power. True, but that is the cost of power at the point of generation.

7:25If you're grid connected, then you need step-up transformers, right? That's going to be anywhere between$5 to$10 a megawatt hour. Then you need high-voltage transmission lines, which also requires a build-out and can be another$20 to$30 a megawatt hour. Then you need substations for step-down transformers. Then you need distribution lines. Then you need distribution level substation transformers. So essentially, once you're done with all this on the other side of it, even as the grid infrastructure built out happens, there's just no way that the cost of electricity from the grid is not going to be significantly higher versus what it is today.

8:11And so a lot of data centers and consumer and industrial customers have started to recognize the fact that even if grid is available in five years or seven or 10 years, depending on where you are, the cost of that electricity is going to be much, much higher. Is it a particular type of customer that sees this as permanent power? I'm sure a lot of customers are nervous because it complicates balances of systems. Operations and maintenance is much more high stakes. It's just a much more complicated on-site endeavor. Is it a particular type of customer that's willing to embrace the more complicated on-site permanent power solution?

8:51Anyone who needs reliable, clean power that is not willing to deal with the grid interruptions, frequency fluctuations, fits in this category. The swim lanes over time will become clearer. grid-connected power will largely be made available to retail customers like you and I for our residential use. We should add as much solar as we can, we should add as much wind as we can, put battery systems, but the grid power will become largely a power that is used by individual customers. And all the CNI customers or data centers will eventually move towards some form of on-site generation. It's no different than even today, right?

9:43When you think about it, you will see a lot of big petrochemical refineries actually have their own form of self-generation. Why is it that when you were setting up your own refining operations, you had your own power? They may be connected to a grid for some kind of a backup, but they are always using their own generation. And when you think about where we are headed, data centers are nothing but a modern-day big power consumer, just like refineries used to be. So the way I think about it is, grid will become largely for individuals, and then data centers and any large industrial load will have its own self-generation.

10:29Just one more question on this permanent power piece. I think that there's really strong competing views in the market now about just how many customers are going to go fully off-grid. And I think people who are maybe more skeptical think that it's going to actually be far more costly, far more complicated for customers, and that many of them will just choose not to go fully off-grid. So what do you say to skeptics who think that the market is going to be severely limited? And I use the word complicated many times. Do you think it is more complicated when we think about fuel cells specifically for this on-site generation?

11:07I think it can be complicated if you use legacy combustion generation because the rotating equipment inherently doesn't have that level of reliability and availability. When you think about solid-state devices like Bloom Energy, actually it's not complicated. It's actually a lot more simpler than grid power. Maybe it was 25, 30 years ago, we used to have landline phones, right? And no one thought that, hey, cell phones or smartphones will become more reliable, will actually increase the penetration and give people a lot more opportunity to be able to connect broadly. because when the telephone industry went decentralized versus having landline phones, not only did the adoption increase, but the reliability of the overall network and the system rapidly improved.

12:05And you're seeing early innings of that starting to happen on on-site generation. Let's go back to the community piece. You talked about lower emissions being a critical piece of what these fuel cells offer. And it's interesting because I think when we had chatted before, you said that maybe a couple years ago, people didn't really care about the sustainability edge. And now developers are saying that it's really critical for bringing a community along, for getting permits. How much is the emissions and local power piece factoring into power decisions now? Oh, big piece of that. Because again, what happens is either you are a data center or CNI, you weren't used to building these big on-site generation power plants.

12:58And only when you started doing it, you started realizing some of the additional challenges that come along with it. And I'm going to give you an example. So if you take a one gigawatt of combined cycle turbines, right, typically the NOx emission is 0.03 pounds per megawatt hour. You know, it's one gigawatt times 1 ,000 times 24 hours times, you know. So you multiply all that, that's about, give or take, 720 pounds per day of NOx. You know what that equates to? that's like having 150 ,000 new cars on the road. That's one gigawatt of power plant at one site. So think about what NOx would be when you start deploying not just one, five and tens of gigawatt of on-site gas turbine-based generation.

13:58And I think that is what and why you see communities pushing back. Similarly, and I'm using big blocks of power, and that's why I'm using the combined cycle gas turbine as an example, because for big blocks, they are the most efficient turbine. You look at water. A combined cycle turbine would use more or less about 190 gallons per megawatt hour, give or take, for a gigawatt of project. you do the same math, that's about 4.5-ish million gallons per day of water usage. You know what that equates to? That's like people having 300 ,000 showers a day. That is the impact that why you're starting to see communities push back.

14:46And this is where Bloom has an incredible technology advantage because we don't combust. So we're not consuming water, we're not contributing to that level of emission. We're the cleanest and the most efficient way as to how you can convert natural gas molecule into an electron. So we don't think it is an or. I think it is an add. You need to add on-site power because grid power isn't available, but you should not be compromising on these community attributes. And Bloom gives everyone that perfect pathway to be able to do both. I'm actually kind of curious about what you're hearing from customers.

15:31So these environmental benefits are certainly more attractive to them. I think everybody has been surprised by just how ferocious the pushback has been. So when you talk to customers, how do the environmental issues play into the broader pushback that they're seeing? How do they weight all of these things that are causing communities to oppose many of these data centers or other commercial facilities? Stack those up for me and tell me where you think the environmental attributes play. I think they've definitely become front and center of how decisions are being made. I will publicly cite one of our large customers, Oracle, the New Mexico project, was with combined cycle and open cycle turbines earlier in the year.

16:24They switched that project completely. That's about a two and a half gigawatt of a single site deployment by Bloom Energy that is going to be constructed. And one of the reasons was exactly for these community benefits, they reacted to the feedback from the community and chose to go with Bloom to show that they are going to be a very responsible corporate citizen. And so it is definitely becoming front and center of every decision that is being made. It's no longer, hey, I can be the lowest cost of electricity, which by the way, Bloom still is. But it's also about how do we do it responsibly so that communities do not feel that the water and the emission levels in their communities are going to go through the roof and it's going to become an uncontrollable problem in a few years.

17:20And so do you feel like you are seeing that those environmental decisions are actually having an impact on whether a project is allowed to go ahead? Yes, we definitely see that. We definitely see that. Again, it kind of connects back to your original question of permanent power versus bridge power. My own feeling has been that in early 2025, data center customers were still going with combustion technologies thinking that it's just a bridge to the grid becoming available. And so maybe these natural gas based generators are going to be on site for maybe two years, three years, five years and then the grid will show up.

18:01Now they've realized that, well, that's not going to be the case. These are permanent energy generation solutions. And as a result, it's also natural that they've started to evaluate all of them very, very differently. You and I chatted before, and you said that you think we're in the second inning of the data center demand story. What do you think the transition from the second to the third inning looks like? Just like for onsite generation, community attributes have become extremely important. I think on-site generation is going to become a lot more efficient in terms of the architecture as to how these power plants are constructed.

18:45And what I mean by that is, I'm telling you something that has already been announced by NVIDIA and a lot of hyperscalers, 800-volt DC architecture is coming. You see, if you were getting the power from the grid, you were required to use AC power that you then had to convert, right? And I think when you go to on-site power, a DC architecture makes a lot more sense. It's less capex intensive for the customer. And so I think that is coming. Similarly, as the cooling needs grow, having absorption chillers so that you can use the heat from the on-site generation for cooling needs is also coming. So the way I see it is the on-site power infrastructure will continue to become architecturally a lot more efficient.

19:49And that to me is the second innings. Let's shift to the broader CNI space now. How are you seeing some of these constraints we've been talking about in the data center market start to impact other commercial and industrial customers? In the exact same way as it is for data centers, right? Because grid power is a zero-sum game. If it's not available for data centers, it's not available for any CNI customer. and with a lot of on-shoring that started to happen. The second, third, fourth derivative of AI, semiconductor manufacturers, chips, fab, server manufacturers, all of them are seeing that, hey, you can't get grid power in time or at a price that economically makes sense for their business model.

20:45So we are actually seeing an unbelievable growth. I sometimes joke that even our C &I customers are growing at a 50 % year-over-year growth rate, and we don't get to talk about it. And so we're seeing an incredible surge from C &I customers all over the U.S. and different states. Because again, if the grid is not available for data center, what makes any industrial think that they can set up a factory and certainly power will show up for them? And you don't get to talk about it because you're so focused on talking about data centers? Yeah. I mean, of course, with the data center infrastructure built out, right, it catches more headlines because deal size are much more bigger, right?

21:27A small data center could be 200, 300 megawatts versus industrials. Their power needs could be in tens of megawatt. Really, does it go above 100 megawatts? Of course, if you're setting up a fab, then you may need that. But generally, the industrial customers are, let's call it, 500 to 100 megawatt range. And so it doesn't catch as much the news cycle. But the spread and the distribution of customer is a lot more broader. And Bloom's been doing that for the last 15 years. I mean, that's the backbone of Bloom. That's how Bloom as a company grew because we were serving CNI customers. And we're continuing to do that.

22:10That's part of our core mission. Yeah, what are the hottest areas of industrial growth that you see right now outside of data centers? Semiconductor fabs, and I would say everyone who's like a second, third, and fourth derivative of AI data center wave is all of them. As the U.S. administration encouraged onshoring of manufacturing operations, all of them have ended up or are in the process of setting up their manufacturing within US, and all of them are suffering with the same power shortage issues. So, Aman, you have like a wide range of customers. ConAgra is a customer, Quanta Computer, Ferrari.

22:50It's a very different set of needs. How do the power needs of like a food producer compare to an automaker? Actually, it's pretty similar. The megawatt may vary depending upon their individual plant size, but all of them are looking for reliable power that is not very expensive. Fundamentally, that's the base layer, right? Everyone wants reliable power. No one wants to be having to worry about power outages so that their production batches go waste if there's a power trip up, right? No one wants that. and be it a car manufacturer, be it a hospital, be it a food manufacturer, fundamentally that's what they want and then rest is, hey, Bloom as a solution is cheaper than the grid because you don't have to deal with transmission, distribution, retail, transfer, the whole value chain.

23:48So it's cost of natural gas and the equipment is still the same. So I actually don't see much difference, Stephen. Are you seeing any differences in terms of sophistication of energy teams at these companies? This is becoming a very complex market. And so I'm just curious, do you see companies under-investing in energy teams? How would you characterize the range? It's a great question. I will start with CNI first, and then I'll build on to data centers. For CNI customers, quite frankly, all they care is reliable power. And if it matches or is cheaper than the grid, instead of paying grid bills, they'll pay Bloom Energy bills.

24:34That's really how they evaluate these things. When you think about large data centers, in the energy industry, people are used to this concept of LCOE or levelized cost of electricity. That concept is for grid-connected power. That equation is fundamentally wrong. If you're thinking about on-site power, right? Because when you're thinking of on-site power, that LCOA equation needs to be adjusted for overbuild requirements, which is required to get to three or four nines of on-site power reliability. That equation needs to be adjusted for all these community attributes in terms of, hey, what's the lowest emission source?

25:16How much money can I save if I don't need water? that equation needs to be adjusted for AI load following capabilities, right? Because with Bloom, we don't need batteries. That equation needs to be adjusted for, even after all this, for other technologies, you have to load firmware on their rack to dampen the AI workload, which is basically wasting 15 to 20 % of the energy just trying to run dummy workloads, right? And so for on-site power, still early in second innings, but there is a clear recognition that what actually matters is not LCOE. It's the total dollar per watt delivered cost of electricity into the rack.

26:03And that is where Bloom is already cheaper versus any other solution. Because a lot of times people are talking about cost of generation, but that's not delivered cost of electricity. and people will talk about, hey, you know, you could get eight cents power using XYZ generation source. True, but that power is not at four nines of availability or reliability or can handle AI workloads. So once you supplement all the additional things that are needed for the quality of power that data centers or these large load industrials need, LCOA is the wrong math. And so still early days, but clearly in last one year, we've seen customers starting to understand this equation as they had other sources of generation get deployed in some of their power plants.

26:54You and I were chatting earlier and reflecting on just what a historic moment this is. I mean, I've never seen anything like this in my career, and you're sort of reflecting on this as more than just a generational buildout. This is a historic buildout. Have you ever seen anything like this? And ultimately, when you say this is a historic build-out, how do you frame this in the context of other huge capital-intensive endeavors we've taken throughout history? Clearly in my lifetime, this is the biggest infrastructure built out that I've seen and at a pace at which it is happening. It's not just an external infrastructure built out.

27:31Sitting here in Bloom Energy, I feel like I'm also going through a massive adoption of Bloom as an asset class in itself. So it's external infra built out and then internally within Bloom, we're getting adopted as a full-on asset class like a wind turbine or utility sale, solar or battery storage happened over the last few years. And I don't see any signs of any of this slowing down. You know, it's interesting. I hear a lot of and I read a lot of articles where people are questioning, well, you know, what if this data center build out slows down? I just don't see it because at the end, fundamentally, it's all of us as individuals starting to use that more compute that sits within the data center, be it we using cloud or chat GPT or consuming more data for our own personal use.

28:30And so in the end, this build-out is not based on speculation. This build-out is based on that underlying layer of individual customers like you and I who are starting to use compute in the forms of tokens that we are consuming. And all hyperscalers are doing is facilitating that in the form of big infrastructure investments that they're doing in data centers. So if anything, Stephen, in the last one year, I've seen more acceleration of the build-out. So to wrap up here, I'm curious, we've been talking about the transition from the second to the third innings. Would you care to speculate what the ninth inning of this looks like?

29:12In the ninth innings, all this has come together. I fundamentally believe that the modern day digital age energy infrastructure is going to be predominantly distributed in nature versus an industrial age infrastructure that was more centralized generation with transmission distribution attached to it. That is the ninth innings wherein, you know, I go back to my telephone example, you know, landlines came, cell phones came, but there was a time, Stephen, that I still had a landline phone, even though I was very comfortable with the cell phone technology. I was very comfortable with the iPhone technology.

29:56And at some point, I gave up on the landline phone because I realized it wasn't needed anymore. So for data centers and for C &I customers, that's the ninth inning, wherein you realize that self-generation that is clean and reliable gives you the independence, the certainty, and the control that you need to run your operations effectively. Aman Joshi, the Chief Commercial Officer of Bloom Energy, thank you so much. Really appreciate it. Thank you, Stephen. Likewise. Aman Joshi is the Executive Vice President and Chief Commercial Officer for Bloom Energy. To learn more about fuel cells and how on-site power can help your business stay ahead, read Bloom's 2026 Mid-Year Data Center Power Report.

30:43You can find that in the show notes. For 25 years, hospitals, college campuses, manufacturers, retailers, data centers, and communities around the world have relied on Bloom for clean, reliable power. Visit bloomenergy.com to learn more. This is an edited version of a Frontier Forum recorded in front of a live virtual audience. We took lots of questions from the audience with much more detail on economics and technology. So if you want to go deeper, you can click the link in the show notes or watch the full video at latitudemedia.com slash events.