theCUBE + NYSE Wired: Powering Tomorrow | Alan Cooper, ON.energy

Clips
More from theCUBE + NYSE Wired: Powering Tomorrow

Alan Cooper

CEO

ON.energy

play_circle_outline Transmission capacity constraints versus abundant generation; peak-hour bottlenecks

play_circle_outline ON.energy AI UPS Decouples Data Centers from Grid for Perfect Power Quality and Stability

play_circle_outline Partnership with Crusoe for multi-gigawatt deployments (3–5 GW currently)

play_circle_outline Streamlining Safe Interconnection and Validation for Behind-the-Meter Storage and Private Generation as Grid Assets

play_circle_outline ERCOT Voltage Ride-Through, Data Centers as Regulated Grid Participants, and AI Load Ripple Transients

Info
Transcript

Alan Cooper, ON.energy

Alan Cooper

CEO ON.energy

This discussion examines power solutions and grid architecture for artificial intelligence AI infrastructure, focusing on decoupling data centers from the grid and enabling AI-scale deployments. It addresses power engineering, grid resilience, interconnection, behind-the-meter orchestration and strategies to turn compute sites into grid assets.

Alan Cooper of ON.energy joins theCUBE Research conversation hosted by John Furrier and Dave Vellante to examine how power engineering intersects with large-scale AI compute. Cooper describes ON.energy's patente... Read more

explore Keep Exploring

What are the main energy-related constraints—especially at the interface between the grid and data centers—that affect the deployment and operation of AI/data centers? add

What makes this "AI UPS" different from a conventional uninterruptible power supply, and how does it operate between the grid and the data center? add

What are the main issues with the electrical grid, and if you could wave a magic wand to fix it, what would you change? add

What needs to change in the interconnection process and the electricity grid to allow data centers (and behind-the-meter generation/storage) to connect safely, affordably, and help support the grid? add

How much capacity are you deploying with Crusoe, and how does your solution address regulatory and grid‑stability issues—particularly ERCOT interconnection rules, voltage ride‑through, and the "AI load ripple" from large data‑center load swings? add

bolt Powered by CUBE AI

Alan Cooper, ON.energy

search

(INTRO)

John Furrier

>> . I'm John Furrier, host of theCUBE, here in theCUBE's NYSE studio, of course. We have our Palo Alto studio connecting Silicon Valley to Wall Street. This is our Powering Tomorrow series. We're kicking off, we're meeting with the leaders. We're bringing the energy solutions and all the engineering around AI infrastructure to power the AI factories and the AI value proposition. Energy bounds the value and we're here to discuss it. Alan Cooper's here, the CEO of ON.energy. Alan, great to see you, thanks for coming on theCUBE.

John Furrier

>> Thanks for having me, John.

John Furrier

>> So energy is obviously the most important asset for AI. We've heard Jensen Huang at NVIDIA talk about it, bounded by energy. I've been introducing the idea of capital, finance is now in the equation because of the importance. But everybody's talking about NVIDIA, the chips, the AI factories, the density, but not a lot of people are talking about what happens between the data center and the grid. That is kind of the biggest story because there's a lot of data that's networked and the energy that powers these factories. There's a lot going on there and it could tip the scales and swing the balance of power, literally.

Alan Cooper

>> Yeah, that's exactly right. I think that's the universe in which we live, which is exactly the place where the grid and power generation meet data centers. And really that's the constraint, right? We're talking about data centers being energy constrained. The reality is that energy largely is abundant still in this country. The problem largely is capacity, which really is a different way of saying that we're transmission constrained. And even that's a different way of saying that we're transmission constrained during different times during the day. Specifically, largely it's like 40 to 70 hours during the entire year that we're constrained. And that creates a real problem for connecting data centers. And we're trying to create real solutions to solve that.

John Furrier

>> Every single person has had this experience. Okay, it's a heat wave, conserve power. We need more power. In fact, the requirement for power has been forecast to be greater than the entire home power load in the United States for this AI. Where do you guys fit in the equation? Explain what you guys do. It's important to understand the context.

Alan Cooper

>> Yeah, so ON.energy created a patented technology. Our flagship solution is called AI UPS. It's an uninterruptible power supply that sits in between the data center load and the grid, and it fully decouples them from each other. And so essentially what it does is it serves perfect power quality into the data center with full uninterruptible power supply, and it makes the data center invisible to the grid, which is a really useful asset now as data centers are getting increasingly large, increasingly volatile with AI training loads. And with regulations becoming increasingly stringent with how AI data centers have to perform and behave on grids with all of these changes to their data center architecture, we created a unique architecture for how they connect.

John Furrier

>> So UPS on every power supply, think of, oh, backup in case the electricity goes out on my PC. What's different about what you're doing? Because you're an interface of the grid. So what's the uniqueness of it? What is it doing specifically?

Alan Cooper

>> Yeah, so our system sits completely in series to the load. So it means all of the energy is flowing from the grid or from on-site primary power generation through our system and into the data center. So everything is flowing through the system into the data center, not in parallel to the load. So it doesn't require any switching times, doesn't require any specialized metering equipment. And that's very unique because essentially what you can do is interconnect a single standardized asset onto the grid that is highly predictable. It's basically a battery that's always charging. You can standardize across all of your data center fleet. The utility knows very clearly what it's getting at all times which is just a very predictable battery storage-like asset, and from a data center perspective, it's getting perfectly clean power all the time.

John Furrier

>> So it's not a generation problem, generating energy, it's the stability.

Alan Cooper

>> Correct, it's a stabilization asset. It's a power control system for the entire campus.

John Furrier

>> And what's it like for the data centers, they don't have it, they have to deal with what? What's the problem for them?

Alan Cooper

>> A million different problems, right? So a data center that doesn't have it still has to have a UPS to protect the compute. They're investing billions of dollars into compute architecture, so they're typically using low -voltage UPS to protect the compute. Then they're likely installing traditional battery storage to meet the voltage ride -through regulations and other grid codes that they need to pass all these interconnection requirements. they need to make all these things talk to each other, they need to have backup generators, and basically our technology smashes all these things together into one.

John Furrier

>> So a boatload of hassles.

Alan Cooper

>> Boatload of hassles.

John Furrier

>> And potential catastrophic problems.

Alan Cooper

>> Correct, and single failures could mean not only failure inside the data center, but potentially grid failures and grid outages, right?

John Furrier

>> All right, so you must have a lot of people buying the system, given the size of the build out. Explain, before we get into that feature, What is the scope from your standpoint of the AI build out? The demand is super high, we're seeing the big hyperscalers take out debt via instruments. How big is that demand in your mind? Scope that for people who aren't tracking it.

Alan Cooper

>> Demand is massive so far. We're actively building three gigawatts out of five gigawatt commitment from our wonderful key customer in Crusoe, Chase was on here a few weeks ago. We are capturing many more customers. We're seeing, depending on who you believe, between 20 and 50 gigawatts of real load being constructed per year between now and 2030. That's exactly why our technology is so important, because with that amount of load growth, you're coming up on real capacity constraints from the grid, and grid operators are trying to come up with a real solution to help the US win the AI race. And the reality is, there's enough generation on the grid to support it, there just isn't enough capacity and these types of solutions are exactly that are designed to make it work.

John Furrier

>> What are some of the issues with the grid, if you had to wave a magic wand at the grid, say hey, fix that, one, two, three, what would it be?

Alan Cooper

>> I think fundamentally they're doing this to some degree across the board. So shout out to all the people who spent tireless time working on this. But number one is interconnection is in the process of being resolved and revolutionized, but it needs to get better and faster. The process of interconnection is complex because they're protecting the average citizen and energy is a basic human right at this point, and so they need to make sure that affordability is maintained and that things that are being connected are safe for everybody. And I think that they need to make sure that that is faster and that projects that are being submitted into interconnection are real, and that they are being implemented with technologies that can make data centers grid safe and that can build compute without breaking the grid. That's exactly the kind of stuff that we're doing. I think number two, there needs to be increased flexibility on the grid. And I think everybody's been chasing that from the increase of decentralization of the grid, from large concentration of generation to decentralization and now battery storage. And I think now this is a huge opportunity for storage behind the meter and even generation behind the meter. So all this primary generation that these data centers are building can be a great asset for the grid if somebody can orchestrate that and sell it back into the grid. So that's another opportunity.

John Furrier

>> So a lot of either dormant power or not connected power.

Alan Cooper

>> And all these things like fully islanded or hybrid solutions or all the story about speed to power, why don't we turn all of those assets into assets for the grid? Yeah. And I think those are great for moments of constraint, right?

John Furrier

>> I remember there was a concept called grid computing back in the 90s. and I think we're going to get to a place where energy and compute is going to be connected. I want to get your thoughts on your product, and you mentioned Crusoe. Chase, we saw him last week in San Francisco. He was on many times. But that is an example of some of the build -out we're seeing, these big AI factories. This is what my narration at the beginning was, what happens between the AI factory and the grid, that's where you come in. You're a critical ingredient for him because he's building essentially massive AI factories. And it's completely built from the ground up. He needs to have that. What's it like working with Crusoe? Share some of the experiences and how it all came together.

Alan Cooper

>> It's beautiful and exciting, honestly. Every day is a new adventure. They're working on many different projects and solving a bunch of different puzzles. I definitely feel our companies have very similar mottos and similar rhythms and paces. And so I think that has helped the relationship quite a bit. They're building some very ambitious stuff at a pace which very few companies can keep up with. and we are doing everything necessary to keep up with that pace that Chase is running. And a shout -out to all of our friends at Crusoe. They're really at the top of their game.

John Furrier

>> And there's tons of Neo clouds coming out. I'm seeing sovereign compute clouds come out. We're seeing a ton of activity. The market's reacting behind us. They're making trades. The market's hot for AI infrastructure.

John Furrier

>> It is, yeah.

John Furrier

>> Explain some of the momentum you have, and how is that being understood by the capital markets?

Alan Cooper

>> I think the capital markets seem to have shifted, particularly in the last couple of weeks, at least from my interpretation of it, from big plays in the hyperscalers and AI monetization. And they seem to have gotten a little bit cooler on their infrastructure bets in their own portfolios, but shifted into exactly the space that we're in, which is sort of power generation and industrials plays and power gen. and I think these are cycles. So I wouldn't read that much into it in these horizons. I think ultimately this is a long -term strategy that we're building. We've been building these types of solutions for 10 years. It's interesting.

John Furrier

>> The capital markets and the financiers, the banks too, they're looking at this market. It's almost like a whole clean sheet of paper. It's a brand new generational shift. The risk variables, at least in my observation, there's no comparable. And so they're trying to squint through the risk. You solve a big piece of the puzzle because you can quantify downtime, you can quantify problems that you solve. They have to model that out because the CapEx investment, people are always like, when's the bubble bursting? Well, the payoff's coming, the demand's still there. So the bubble conversation to me is irrelevant because there's no bubble. There's maybe some speculation, right, on prices, but there's demand.

Alan Cooper

>> And maybe what I would comment, John, is that it seems to me that the market, every time it shifts from definitely not a bubble to maybe a bubble, it shifts from bet on the hyperscalers to maybe bet more on AI picks and shovels. And maybe that's the toggle that keeps shifting, at least in my impression. And I think that's exactly the shift that we're seeing today.

John Furrier

>> How should people think about your company?

Alan Cooper

>> I think we're an AI picks and shovels play, of course. We build specialized industrial power technology. We build it very well. We've been building it very well for a long, long time. We build patented equipment in specialized factories. We build it right here in the US very proudly. And we're deploying it at the largest scale possible at some of the most ambitious projects.

John Furrier

>> Can you share some examples of some deployments, momentum?

Alan Cooper

>> So we're actively deploying about five gigawatts with Crusoe. We have closed on other projects which cannot be named. Projects, unfortunately, are extremely confidential, as most of the industry has become. The product market fit has been extremely attractive, particularly in markets like ERCOT, where regulation, again, around the interconnection, requiring solutions such as ours to keep data centers safe on the grid for things such as voltage ride -through and what we call the AI load ripple, where basically data centers, especially AI...

John Furrier

>> What's it called?

Alan Cooper

>> The AI load ripple. Essentially what happens is these data centers and high -performance computing are large clusters. They're all processing about a gigawatt of compute at a time, and they're all processing essentially the same problem at the same time. And then as they cache and process the information, they drop the load about 70 % before they pick up again. And in that process, essentially, it's like strobing on the grid. And that presents a big problem. It spits out transients onto the electric grid. and essentially the grid operators are saying, let's hold our horses and let's solve this problem. Let's make sure that at least you're not negatively impacting the grid if we're going to let you connect. And so that's a regulation that has been formalized across ERCOT in Texas and likely is expanding nationally here soon. Our solution solves a lot more than that, but it's indicative of the same trend, which is sort of the barometer shifting from, hey, I don't want you to catastrophically impact our grid and our communities, to, hey, I don't want to feel you at all, to, hey, data center, how about you become a grid asset and increase flexibility and become positive to the grid? And I actually see that trend happening across all of our customers and hyperscalers, shifting to how do we become positive participants?

John Furrier

>> We're seeing, I'm doing a story right now on carbon credits. That's a lot of this kind of community stuff where there's a lot of economics involved in giving back. basically what you're saying is that the grid wasn't built for this.

Alan Cooper

>> It was not.

John Furrier

>> So you saw that immediately.

Alan Cooper

>> Yeah.

John Furrier

>> But on the business side, there's requirements to run these factories or anything at scale. a business would have electricity to power the lights.

John Furrier

>> That's right.

John Furrier

>> Run the operations, but not run monster data centers. So you're essentially bridging this generational shift.

Alan Cooper

>> That's right.Essentially they're saying, the grid is saying, you want to connect this thing, you got to put a firewall between me and you. And that seems completely fair. And from the data center's perspective, they're saying, if the investment is modest, then that seems fair too.

John Furrier

>> I love how computer networking technology comes into all examples. Then you got the DMZ, you got inside the meter, so this is what happens, this is the new architecture. You're redefining the architecture.

Alan Cooper

>> That is, and we really are redefining the architecture, and we define it as the topology of the data center has to be re-architected, it's a great opportunity because they sort of have to be re-architected anyway because of the nature of high performance computing moving from being like 30, 60, 130 kW racks and now moving towards extreme high density computing. So I think we're taking advantage of that opportunity to really rethink how a data center should be built.

John Furrier

>> Powering Tomorrow, pun intended, as a double entendre on the words, but talk about the journey that you've had. Obviously you're in a good position, the product market fit, the shift to the new architecture is happening with the grid and its relationship to what's connecting to it in terms of size and scope. Has it always been easy? I'm sure I've heard stories of regulatory nightmare, hoops you got to jump through, but now it's pretty clear that there's economic value in these data centers, notwithstanding people who don't understand what it means and protesting the data centers. but you're connecting an economic engine to the Internet and the grid.

Alan Cooper

>> the answer is none of it's been easy. All of it's been rewarding. It's been 10 years of trying to help customers work through deeply unsexy problems, both behind the meter and in front of the meter, of how do you solve this resiliency challenge? and using a deep know -how of batteries and power control systems to connect these two problems and solve them. And I think finally it's being applied to a space that is creating real change and really is a generational opportunity so we're fully taking advantage of it

John Furrier

>> When was the markersin time where people started to get it? And then when did it click in? Or where are we, are people still figuring it out? Where's the journey?

Alan Cooper

>> So I think there was a very clear marker about 18 months ago. We started bringing our full data center push about 18 months ago and I think the team over at Crusoe was clearly the first to get it. Particularly the digital infrastructure group over at Crusoe with Chris Dolan and Jason Hutzler over there. And since then there's been just such widespread buy -in from everybody inside of the data center architects inside of a lot of these fast -growing, particularly the energy -first leaning firms. I think now more widespread adoption amongst maybe some of the more traditional data center developers who have more three -year roadmaps for technology development, the QTSs, Aligned, firms like that. And obviously into hyperscalers, who ultimately have to be convinced as the end users.

John Furrier

>> Is there a trend that you're watching, carefully, that impacts you, or product opportunity on the roadmap? What jumps in my head is energy switching, energy routing. I learned here at the NYSE that their parent company, ICE, they do energy futures. I see people vertically integrating their energy into their data center. They might want to maybe share that, broker it. There's money involved in energy. So I see a lot of action happening.

Alan Cooper

>> You see energy seems so decentralized still. The energy generation side of it still seems extremely decentralized. I do see a trend of more private equity scooping up power generation companies. So that seems to be a really interesting trend. The acquisition of, the take private of AES by BlackRock GIP was an interesting example. If I look into the future, I think a lot more utilities are going to be taken over potentially by more technology companies with exactly this in mind, and I think that's probably a wise strategy as data centers become a more interesting load profile as part of their portfolios. Lots of new technology. Lots of technology risks still. I think maybe that's one aspect of our solutions that I really liked, where we leveraged really reliable existing supply chains at full scale to deploy this thing at gigawatts, which do not underestimate what the Crusoes of the world and what these people are trying to construct at the pace at which they're trying to construct it because there are 9 ,000 people on location where we're building today and they need their stuff on time, the day they need it and you cannot be a minute late. And so I think you need to be up to that standard and your whole supply chain needs to be ready for that. And I think some of these new technologies are very interesting, very attractive, but will they be ready for that level of scale?

John Furrier

>> Timing's everything there.

Alan Cooper

>> Timing is everything and capital's everythingand manufacturing is hard.

John Furrier

>> Yeah, and the speed that they have to deploy these factories. Crusoe's got an ambitious goal, but this operating value when it's up and running, not just build, when they start operating it, they're going to need to have that steady energy to make sure that the systems are running and have resilience.

John Furrier

>> That's right.

John Furrier

>> Final question, what's your focus now? What are you optimizing for in your current situation? Is it more technology build, more distribution, more customer, go to market? What's the focus?

Alan Cooper

>> Yeah, I think like everyone, a little bit of everything. So the whole world is coming at us really fast. Building out multiple concurrent projects in multiple states at the same time for multiple off -takers is its own beautiful challenge. It's a great opportunity. Continuing to improve the product relentlessly, which I think is a big part of our roadmap. How do we continue to have the best product on the market? We currently have the best one and I guarantee you that we will continue to have the best one. And how do we continue to add more value to the customer than they would ever expect from us? And I think that's a lot of stuff that's beyond the product. So how do we continue to add flexibility? How do we make them have a more compelling value proposition to their end users, which are hyperscalers, and their communities? So how do we make them more flexible to the grid?

John Furrier

>> As we said at the top, AI value is bounded by the energy available. Alan, thanks for coming on and sharing what you're doing and your insights here on theCUBE NYSE Wired program.

John Furrier

>> Thank you very much, John.

John Furrier

>> All right, Powering Tomorrow, our new series. Talking to the leaders in energy, making it all possible and powering the future of AI. Energy is the bounding function for the value being created by AI. It's happening super fast. The industry leaders are coming together to do that. I'm John Furrier, host of theCUBE. Thanks for watching. Thank you.