The CEO Fixing America's AI Infrastructure Lag Behind China

In this episode of Motley Fool Hidden Gems Investing, Motley Fool contributor Rachel Warren talks with Hannan Happi, co-founder and CEO of Exowatt -- backed by Sam Altman and Andreessen Horowitz -- about why the AI build-out is hitting a wall that no amount of chips or software can fix. They discuss:

• Why a one-year grid delay costs a hyperscaler $12 billion in missed revenue.
• Why China has 10 times more capacity than the U.S. to build AI infrastructure.
• What investors need to actually be tracking as hundreds of billions of dollars flow into the AI build-out -- including whether the data centers being built today will still be operating in 10 years.

To catch full episodes of all The Motley Fool's free podcasts, check out our podcast center. When you're ready to invest, check out this top 10 list of stocks to buy.

A full transcript is below.

This podcast was recorded on Aug. 23, 2026.

Hannan Happi: China adds 540 gigawatts of power to the group per year, to more than 10 times our capacity. That means that China, if we consider China to be our rival in this AI dominance race, has 10-11 times more capability and capacity than we do to build infrastructure and bring that online.

Rachel Warren: That was Hannan Happi, co-founder and CEO of Exowatt, on why the AI race is less about algorithms and more about electricity and why the U.S. is dramatically behind where it needs to be. I'm Motley Fool analyst Rachel Warren. The conversation about AI almost always focuses on chips and models, but Hannan argues that the real bottleneck is something far more physical. He joined me to discuss why a single year of grid delays could cost a hyperscaler $12 billion in missed revenue, why communities across the country are pushing back hard on data center development, and what investors watching the AI build-out should actually be tracking as the capital flows in. We hope you enjoy.

Hello, everyone, and welcome back to Motley Fool Conversations. I'm Motley Fool analyst Rachel Warren. Today, I'm joined by Hannan Happi, the co-founder and CEO of Exowatt. Hannan brings an incredible deep tech background to the table. He studied mechanical engineering at the Technical University of Munich, later attended the Stanford Graduate School of Business. Prior to launching Exowatt, he held key engineering and leadership roles at industrial and tech powerhouses like General Electric, Accenture, Tesla. He also co-founded Volansi, an autonomous hardware and logistics company. But now he's tackling one of the biggest challenges facing the tech sector right now, is company Exowatt, which has captured the backing of elite investors like Sam Altman, Andreessen Horowitz. Is purpose-built to power AI, delivering dedicated renewable energy to the engines of modern intelligence. Exowatt calls its energy generation category, on-site firm solar. It's designed to bypass the years-long utility grid delays that are stalling the modern AI buildout. Hannan, welcome to the show.

Hannan Happi: Thanks for having me, Rachel.

Rachel Warren: Absolutely, the last few years, when I think a lot of investors talk about AI, the conversation is very much focused on semiconductors, networking capabilities. But you've argued that the next massive phase of AI investment isn't actually happening in software. It's happening in that physical infrastructure. Walk us through that. Why has the biggest bottleneck and the AI boom suddenly shifted from the silicon to the raw physical infrastructure?

Hannan Happi: Absolutely. If you think about just a couple of years ago, the largest data center that we had in the United States was about 100 megawatts in capacity. That was considered very large and the rack density was in tens of kilowatts, and a building block of a data center was maybe 10-20 megawatts. Fast forward, just in the last couple of years, the building blocks for data centers have scaled up in the order of 300-500 to even 700 megawatts, the building block. You could argue the building block of a data center is now five times larger than the largest data center we had in the country a couple of years ago. The data centers themselves are now on average about a gigawatt, and there are some that are in the order of 10 gigawatts. To put that into context, a gigawatt data center essentially is consuming the equivalent of 1 million U.S. households' worth of energy. You're basically saying I'm building a city from scratch for 1 million people, and I'm trying to do that as fast as possible because I want to stay in the AI race. I want to have the best model. I want to be competitive. This is where we're past the idea of how do I write the best LLM algorithms or models? How can we overcome the chip shortage? Because the chips become better, too, and more powerful, even if they have become more energy efficient, they're still consuming a lot more energy as a total.

Now, we have to build power infrastructure. As a country, we haven't had to build massive amounts of power infrastructure. For the last couple of decades, we haven't had much of a load growth in the U.S., and now we do, and data centers are eventually going to contribute or be taking about 9%-10% of total energy produced in the U.S., the grid capacity. We are in a mad rush to build power infrastructure. Building power infrastructure is not trivial. It's not like software that you can write code, and it just scales infinitely. You have to move metal. You have to move concrete. You have to move earth. You have to do all sorts of permitting, and construction, and move workers. We just don't have those capabilities at scale in the country. What we're trying to do at Exowatt is come up with a formula that allows us to do something in a factory setting with tight control over cost, tight control over the speed of execution, and try to scale that as fast as possible to address this power gap. But there's a massive power gap, and I think everyone's scrambling to figure out how to cover it. That's the reality that we're facing today.

Rachel Warren: I think as you noted, it's become increasingly obvious that the primary constraint is about delivering reliable power where AI infrastructure is actually being built. What are the solutions to that look like? Obviously, I know that's what Exowatt's all about, but I'd love to hear your thoughts, as well, broadly on how you're able to get past this bottleneck in the next five, 10, 15 years if you're building out all these data centers as a tech company.

Hannan Happi: Good question. Again, going back to the history of data center build-out. Three, four years ago, when you built a data center, again, whatever it was 25 megawatts or even 100 megawatts, you really didn't think of power as a constraint. You basically put the data center where the location was favorable, from a real estate perspective, maybe from a fiber perspective, typically close to urban and suburban areas, and power was not really necessarily a key factor. You just plugged it into the grid, and the grid was able to give you power. But fast forward today, where we these massive data centers, the grid has no capacity. They have to come up and scramble with solutions to build their own power. This started out with the mad rush to go build gas.

Over the last couple of years, especially in the last two years, every data center that has been announced is saying, "Well, I'm going to build some version of gas generation behind the meter gas, and eventually, there'll be some grid." Gas generation has its own challenges. We have, yes, abundant amounts of gas in the country, but we have a lack of supply chain for turning that gas into electricity. Turbines have been backlogged 5-7 years in some cases. We have also now fuel volatility, given the geopolitical things that are happening around the world, gas prices are also fluctuating. Then just by the simple fact of having a gas line doesn't mean you actually have access to that gas to turn it into electricity.

A lot of people started saying," I have a gas pipeline, build a data center. It's not that simple." I think the most recent thing that's happened, if you put aside all the supply chain constraints, all the factors that go into building a generation system, is now the communities are saying, "I don't want this many gas generators in my backyard." The NIMBYism for gas generation is now really spiking because gas generation by default is not necessarily a clean solution. It might be a fast solution, but it's not a clean solution. The pollution that it's creating, the emissions that it's creating, it's really hurting the communities. In even very gas-rich states like Texas, we're seeing a backlash against data center build-out. Our theory and hypothesis has been a couple of things we're doing wrong and we should do differently as a industry and also as a country. No. 1, we shouldn't go build these massive data centers in the same locations we used to build the smaller data center, so in urban and suburban areas, where they can have a massive impact on the communities. We should try to take these massive infrastructure projects, these data centers out to areas where the impact on the community is lower, where you can get abundant, cheap energy from solar resources, even if it is gas resources. You want to take it as far out as possible. Solar, in our opinion, is the cheapest, most abundantly available, least supply chain-constrained source of energy that we have today, not only in the U.S., but around the world.

The supply chain is very robust, and if we're talking about traditional solar panels, and it's a matter of now making that enormous amount of solar energy that we get that sufficiently covers all of our data center power gaps and making it such that it's dispatchable, so that it can serve a base load application like a data center. The argument for solar against solar and wind and other renewable generation sources, historically, has been these are intermittent resources, so when the sun is not shining, when the wind is not blowing, how can I power my data center that's supposed to run 24/7?

That's exactly what we're solving for at Exowatt by taking that intermittent resource and making it dispatchable by storing that energy in what we call a long-duration storage technology and allowing data centers to have access to the solar energy that we capture for today around the clock, 24/7, to power their baseload applications. I think that's the shift that we're going to see in the data center market. There are already hyperscalers that are starting to do that at scale, and I think more and more of that is going to come online hopefully very soon because it is the fastest path to electricity, essentially, and the lowest impact on communities. Also, it creates a ton of jobs because you have to manufacture these panels. You have to manufacture all the supply chain for the batteries and the components that go into that. That's where I think hopefully the sustainable data center build-out is going to move towards away from urban and suburban areas, away from very polluting generation solutions like gas and diesel to more sustainable solutions, not because data centers want to be sustainable, but because that is the cheapest, fastest path for them to get access to power and without hurting any of the communities where they're building that power infrastructure.

Rachel Warren: The factors that we've seen, like deployment speed and permitting, energy availability, which, of course, you've just touched on, execution risk. This is a much bigger part of the picture as we're watching the AI build out. Demand seems to be just one very small piece of the pie. But we hear a lot about grid interconnection delays stretching for years. Can you paint a picture of how devastating a 3-5-year grid wait time can be for a hyperscale or a tech company that's trying to build out these data centers to train their next-gen models?

Hannan Happi: Just think about over the last, what is it now? July, or last 7-8 months this year, how many model releases we've had between just OpenAI and Anthropic and then also Meta and Grok and Google and also all the Chinese open-source models. I can't keep count of it anymore, but I'm sure it's in the order of dozens now models that have been released over just the last seven months. The pace of these models being developed and released is extremely high. The competition is really high. If you look at the revenue profiles of OpenAI and Anthropic. It's very obvious that this is a very profitable business, every day counts, every minute counts. Whether you're essentially an AI application company or an AI model company, or you're the company that's providing infrastructure, if you look at the recent transactions that SpaceX did with Anthropic and with Google, we're talking about $1 billion a month. A year delay, if you just take that as a number, is about $12 billion of missed revenue opportunity.

Obviously everyone is a mad rush to get any access to any power they can get as fast as possible. It used to be that when we talk to customers, they said, "Unless you can give me 300, 500 megawatts of gigawatt, I'm not going to be interested." Now, it's come to a point where any power you can give me, I'm going to take it, even if it's in small quantities, because the equation is very simple. If I have 10 megawatts or 100 megawatts, I can monetize that to the tunes of hundreds of millions of dollars, if not billions of dollars. That's why everyone's in a mad rush, putting aside the fact that we're also competing with China for AI dominance. Just internally, the economics of AI right now are so lucrative that everyone's willing to spend whatever it takes to get access to power infrastructure, to computer infrastructure as fast as possible.

Rachel Warren: Given all these factors, do you think that the next generation of AI winners will be defined less by their breakthrough technology and more by their ability to build out?

Hannan Happi: Yes, and I think that's definitely the case. I think that's where as a country, we have a massive problem. If you think about, in the U.S., we're adding about 50 gigawatts of power to the grid per year, and we're talking about the hyperscale spending hundreds of billions of dollars to trillions of dollars on their computer infrastructure, and they need to cover a gap that's in the order of 70-100 gigawatts at a minimum. Then you take that as comparison to the fact that China adds 540 gigawatts of power to the grid per year. More than 10 times our capacity. That means that China, if we consider China to be our rival in this AI dominance race, has 10-11 times more capability and capacity than we do to build infrastructure and bring that online.

Now, we can argue they don't have the most advanced chips that we do, but it seems like even without those chips, they're already outperforming us, given the most recent open-source models that Chinese companies have released at that are outperforming the closed source models that we have in the U.S. I think we're at a massive disadvantage there as a country, and I think we have to scramble to figure out what energy infrastructure we can bring online, how fast what partnerships we can establish around the world because there's massive challenges here in the U.S., not only the permitting challenges the grid backlog, and not only the fact that we have to build so much infrastructure, we have the community backlash. We have the essentially lack of resources, labor to build and construct and electricians and technicians. All of that combined with the fact that our so-called adversary in this race is 10-11 times more powerful than we are. I don't really have a solution for right now, but I think that's a challenge that we have to all acknowledge.

Rachel Warren: Well, we've spent a lot of time talking about some of the infrastructure build-out challenges and the various integrated energy solutions, like what Exowatt is providing that could help to stem that tide. But I want to go back to something you mentioned earlier, which is this idea of NIMBYism. Maybe you can define what that means for our audience, and then I'll follow up with my next question.

Hannan Happi: Basically, NIMBYism means that a community or someone that is in the vicinity of the data center is asking not to build that data center in their backyard. For good reasons, they're saying, "Look, I'm paying the price of all the pollution that the data center is creating." The real estate is gone for that data center. It's not creating search. A lot of jobs. There's, maybe some marginal tax revenue, but for the most part, data centers don't have a lot of ongoing jobs or sustaining jobs. You're taking away my water resources. You're increasing essentially my energy bills because I have to pay for your infrastructure upgrades. For all those reasons, you could see why communities don't want data centers in their backyards. I think there's been an effort now from the administration and also from the hyperscalars to try to reduce this burden by saying, "Well, the hyperscaler or whoever's building that data center has to start paying for that infrastructure upgrades themselves." They have to bring this online in a way that it doesn't impact the ratepayers and the communities. They have to try to reduce these emissions that are hurting the communities, one of the things that is really tragic about this is we talk about how fast the data center is coming online in certain places.

The other thing we don't talk about is like, well, everybody around that data center now has cancer, and is dying as a result of that. Yes, while one company may be saying, "I put out a slightly better model than this other company." The communities around that data center have been massively impacted. I understand and I have sympathy for the why you wouldn't want a data center in your backyard. I think that's why we're advocating for a model which we call "to build on frontier land." Not in urban areas or suburban areas or even rural areas, but to go out in areas where we have massive amounts of land, massive amounts of energy, and solar resources and energy resources as a whole, and we're far away from impacting communities. We're not going to take away the energy resources. I think that's the shift that needs to happen for us to be able to really build data centers.

Otherwise, as you probably know, I think there's over $160 billion of data centers that are currently not coming online this year that we're supposed to be coming online because of all the community backlash. Many states have already put moratoriums out. Many are considering them, both Democrat and Republican states; this is a bipartisan hate issue. Nobody wants this data center build-out. I think we need that data center build-out. The infrastructure is required, of course, because we want to advance. We want to stay dominant on the AI front, but we need to do it in a sustainable way. I think what we've been doing so far hasn't been sustainable. It's been just driven by how fast can I get something online and really not take into consideration what the even short-term impacts of this are going to be let alone the long-term impacts.

Rachel Warren: I want to talk about this frontier land concept that you have been talking about. It really flips the legacy model. Historically, data centers were built where networks and talent already existed. Maybe talk through this idea of why we should be looking more at locations where the reliable energy can be built alongside compute power and how that flips the script on what we think of in terms of real estate and utilities and infrastructure.

Hannan Happi: Again, going back to the history of data center build-out. In the era where power was into constraint, you just built it at a location where, there was at some point, a constraint around fiber connectivity, so you want to have access to fiber. I think the other thing was that a lot of data center build-out was happening with a real estate lens. Also, the developers of these data centers are always thinking about, well, what else could I do with this plot of land if it's not a data center, or what do I do after the data center? Boom is gone? Is there some other use to this land? I think that's why data centers were really concentrated around urban and suburban areas, or especially in places like Virginia, you have like this data center alley, high concentration of data centers. By the way, Virginia is not cheap. Land is not cheap, power is not cheap. There's nothing cheap in Virginia. You should ask yourself, why do we build so many data centers there? Just doesn't make sense.

I think our opinion is now that we've really come to a point where data center developers just don't have any more solutions to go build urban and suburban data centers in prime real estate markets, essentially. Now, there's a shift to let's go build farther out from cities and towns and even rural areas, frankly, and go build in what we call barren land and frontier land where you have abundant land, 41% of the country is empty, literally empty. There's nobody living in 41% of the country. I'm not saying all of that is flat, and all of that is prime for data center development, but more than enough of it is available to power all the data center capacity that we need in the country. We can do that sustainably, because in those areas, you have abundant sunshine, you have abundant wind, you have abundant gas resources, even, especially if you think about Texas and New Mexico and other places like that and Oklahoma. We have now also started building data centers in a way that they're going to be reliant on having massive amounts of construction workers.

This whole hypothesis behind Exowatt is if we can build a modular system in a factory setting where you can have your professional labor in the factory. You don't need to go have them in the middle of the desert. You can then ship out these modules and then install them very easily on site. The same thing is happening on the data center development as well. Companies are developing much more rapid ways and modular ways of building data center infrastructure. I think that's where the industry needs to go and is somewhat going, but we're still going to be seeing a little bit of this short-term impact on the reaction that some states are going to have a data center build-out about completely banning them, completely coming to a stop or taking away their incentives. I think we could revive that in certain areas where it makes sense by showing that there's a model for sustainable data center development.

Rachel Warren: It's fascinating to hear about as well. I guess another question I'd like to ask, are we seeing the hyperscalers, the tech companies that are building out their data centers? Are they looking at this type of energy generation as the solution to a lot of the problems and bottlenecks that they're seeing right now in the infrastructure buildout? What does that adoption curve look like from your vantage point?

Hannan Happi: I think, yes, in general, I think the majority of the hyperscalers out there are right now trying to build some dispatchable renewable capacity, whether it's taking traditional solar panels and combining them with some short-duration batteries or whether it's to combine them with long-duration storage technologies or whether it's to buy and build a full-stack energy solution like the Exowatt P3. We have a massive backlog right now for our product, for our solutions, something that we're working really hard to try to serve as we scale up our manufacturing. But there are definitely a lot of interests for these types of solutions because everyone at the end of the day knows that this is the path, the fastest way to get to electricity. This is the fastest path to electrons that are cost-effective, that are sustainable, that are not going to be impacted by supply chain risks, that are not going to be impacting the communities. I think it's happening more and more.

Again, the last two years, I would say it was very focused on gas. Everyone thought gas is the end all be all. But now we're seeing this transition more to dispatchable renewables as being the solution. There are some other technologies out there, too, that folks are investing in or thinking about. I should say, at the end of the day, there is no silver bullet for energy or power. You have to build an energy mix, that's ultimately what our grid is. It's an energy mix. Every solution out there should be tested out and tried as much as possible. Some of them have shorter-term timelines coming online and lower costs. Some of them have longer-term potential. Hyperscalers are at the end of the day trying everything, to be honest. If you can go sell them power from anything that you can generate power from, they'll probably be a buyer.

Rachel Warren: Well, as we draw to the end the last few minutes of our conversation today going back to all the different points that we have discussed in this last bit of time if you're an investor that is watching the AI infrastructure build-out, that is something that pertains greatly to a lot of the companies that we as investors own and invest in. What are the things we should be looking for to track the health and sustainability of the AI infrastructure build-out, to ensure that all of the capex that we are seeing flowing into the markets, that it's actually having a valuable, profitable effect for the companies that we invest in.

Hannan Happi: I think as an investor we look at both the short term and the long term. The short term is how viable is the plan for this capex build-out going to be in the short term? Are there going to be any obstacles facing that from communities, from an energy perspective, from a labor perspective. These are short-term things. Then the long term is how viable is this going to be in the long run in the sense that, are they building this infrastructure around energy sources that are going to be sustainable, not in the sense of green, but sustainable in the sense of they're not going to have supply chain volatility. They're not going to have fuel volatility. They're not going to have, with a change of administrations, potential huge penalties to pay, which is going to be probably the case for a lot of gas data centers, the change and tune of the administration and policies change around, like, gas data centers, and I think a lot of hyperscalers are aware of that. They might be up for paying massive penalties for the pollution and the emissions. Also, the communities will probably come out and ask for a lot of damages or maybe put these data centers to complete halt. Who knows? I would look at both the short-term and long-term.

I think it's very enticing to think about the short term because, as we discussed, the faster you can get these data centers online, billions of dollars get printed out very quickly. But I think as an investor, I would want to say and look at how sustainable is that revenue stream? Can we bank on it for another five, 10, 15, 20 years, and look at the long-term components of the infrastructure build-out as well. Really, I think, in that sense, think about where are these data centers being built is that essentially going to be in an area that's sustainable in the long run or not?

Rachel Warren: I so appreciate your time today, Hannan. It's been really a fascinating discussion and really interesting to hear about the pace of the build-out, what it's looking like and the work that Exowatt is doing. Thank you so much for joining me.

Hannan Happi: Thanks for having me, Rachel.

Rachel Warren: As always, people on the program may have interests in the stocks they talk about, and The Motley Fool may have formal recommendations for or against, so don't buy or sell stocks based solely on what you hear. All personal finance content follows Motley Fool editorial standards and is not approved by advertisers. Advertisements are sponsored content and provided for informational purposes only. To see our full advertising disclosure, please check out our show notes. For the Motley Fool Hidden Gems Investing team, I'm Rachel Warren. Thanks for listening. We'll see you next time.

Rachel Warren has positions in Alphabet. The Motley Fool has positions in and recommends Accenture Plc, Alphabet, GE Aerospace, Meta Platforms, and Tesla. The Motley Fool recommends the following options: long January 2028 $260 calls on Accenture Plc and short January 2028 $280 calls on Accenture Plc. The Motley Fool has a disclosure policy.