Inside the Factory Using Rocks & Sunlight to Fix AI's Power Problem | Exowatt

[00:00] So what is Exowatt? Exowatt is developing the solar backbone for the AI infrastructure. So think of this as like a solar panel, but that has a built-in battery. One of the elements of our product is actually a Fresnel lens. This essentially focuses the light coming in from the sun onto the battery material. It gets very hot and then stores that energy. So what Exowatt is really doing, our mission is to build a solar backbone. [00:30] The thing about the AI race is so intense [00:33] that's now [00:34] time to power, right? I need power, I don't care how you make it. If you talk about a gigawatt data center, [00:40] That's almost a million U.S. households of energy. And the grid is just really not built for that. The most important thing is, at the end of the day, the raw materials that we use, sand and dirt and steel, right? And so you don't have to go import something from China. Really, the vision and idea is to scale this to the millions and billions of units, then build it better, faster, cheaper, and don't take no for an answer. [01:16] Welcome to Exowatt Miami. Thank you. Thanks for having us here. So what is Exowatt? So Exowatt is developing the solar backbone for the AI infrastructure. So we're [01:29] powering data centers that required enormous amounts of power,

[01:33] with solar energy that's available around the clock, which is [01:37] Not something that people usually think of when they think sun, because the sun goes down at the end of the day, but we've developed a technology that we'll show you today that allows us to capture that energy and make it available and dispatch around the clock to power baseload applications like data centers. [01:52] Amazing. So we're going to do a little tour around the office. I did [01:57] see the little model in there. - Yes. - Can we go check that out? It looks like you have merch too. - Yes, we do. You wanna start there. - I definitely do. - Yeah, okay, cool. [02:07] Yeah, please come in. So our flagship product is called the ExoWatt P3. And so [02:16] This is a desktop model, as you said, and essentially, as you can say, it's a modular system. Each module the size of a standard shipping container. [02:24] and it's made of three elements. That's why it's called the P3. So the three elements are the optical table, so these lenses, and then you have the heat battery and the PC or the power conversion unit all in this kind of container. And the idea is that we capture energy from the sun throughout the day. These lenses are tracking the sun. And then they're focusing that light onto a series of [02:48] rocks, right? That's the heat battery. [02:50] And then that energy is then transferred via an air loop to the engine, which produces electricity. So think of this as like a solar panel, but that has a built-in battery. And that essentially you can scale linearly to any project size. So if a data center customer needs electricity,

[03:07] 100 megawatts or a gigawatt or however many gigawatts, you just stack these next to each other. And kind of the hypothesis behind ExoWatt is that [03:15] None of the elements of what we're doing here are net new. People have been using lenses [03:20] centuries long ago. The engine technology that we use is very established and old as well. And essentially what we're using to store the energy is ancient as time. So it's essentially rocks. And so none of that is new. What's new really about our approach and our hypothesis is making this in a modular fashion, making this in a factory, and then scaling the production to millions and ultimately billions of these. [03:45] which ultimately allows us to go down the cost curve and generate electricity at our ultimate goal of one cent per kilowatt hour. [03:52] That's why we started the company. That's kind of been the goal with Northstar from the beginning. And now we're on a track and a path to get there. [04:01] How did you actually [04:02] found the company and who was involved. [04:05] Yeah, it's... And why Miami? [04:07] Yeah, yeah, interesting story. So we started here because I moved here from San Francisco after I sold my last company to work with Jack and Chester at the Atomic team. So Atomic is a venture studio. They start a... [04:21] bunch of companies internally. And so I moved in as a founder in residence back in [04:26] 2022 and started working with Jack ideating around what could become exowatt. Really, the hypothesis was [04:35] the modularity. So we knew solar technologies, especially solar PV and solar thermal are established technologies. Solar PV has actually been really successful in coming down the cost curve. If you look at the trajectory over the last 50 years, the cost of making a solar panel has gone down by 99.6%. But solar thermal,

[04:55] hasn't gone down that trajectory because most of the time people have built very large projects and so we said like there's an opportunity to take that modular approach to solar thermal which actually is contrarian right because a lot of people say [05:07] If you think about thermodynamic systems, you always want to build the largest system possible. That's why you want to build the largest rocket, you want to build the largest engine, you want to build the largest thermal mass. [05:18] Because the larger the system is, the better the thermodynamic efficiency. So to go and say, like, we're going to build a small modular [05:25] thermodynamic system is kind of contrarian and counter intuitive, but we actually saw the benefits of [05:32] doing that because we can very much control the costs and make that a process that we can replicate over and over and have a learning curve, similar to solar PV. And so that was a hypothesis. And we didn't really have any sense of what is the product or what it's going to look like. So what you see here and the future iterations [05:52] They're nothing we dreamt of. It was basically a process of iterating. So we went through various prototypes, we went through 50 different configurations, [06:00] to even come up with this one and we've continued to iterate on that. And every time the approach we've taken is like, how do we get to one cent per kilowatt hour, which is [06:10] By the way, for context, like, [06:12] electricity prices probably where you are at or where we're here at for residential customers are anywhere between 13 to 20 cents a kilowatt hour. For commercial customers, it could be as low as 8 to 15 cents a kilowatt hour. So to get to 1 cent a kilowatt hour is insanely, insanely low. And if you think about data centers, they're signing contracts as

[06:34] on the mid-range 20 cents a kilowatt hour, up to 30 cents a kilowatt hour. So they're really desperate to get power and really want to solve this. And we saw a path to doing it sustainably. We saw a path to doing it, by the way, in the US, right? This is before tariffs were issued and before all the policy was an issue. We said, in order to get to that cost, we need to manufacture domestically. We need to source domestically. We need to make this using raw materials that don't [07:04] So every configuration that we designed, we looked at the bill of materials or the BOM, and we asked ourselves, [07:10] What can we eliminate from this? How can we make this simpler? And the same approach you see at SpaceX, right? Or at Tesla. So that's where I was before my last company. So that constant iteration on let's remove parts, let's make it simpler, let's scale the manufacturing is what has led to this. And yeah, we'll kind of show you the real products outside. Okay. Yeah. Amazing. [07:36] Sorcery is brought to you by Brex, the financial stack trusted by more than 30,000 companies, including one in three venture-backed startups in the U.S. Nearly 40% of startups bail because they run out of cash. Brex is literally built to help founders avoid that. Unlike traditional banks that let your money sit idle, chipping away at it with fees, Brex is designed to help you spend smarter and move faster.

[08:06] powerful account. You can send and receive money globally at lightning speeds, get 20 times the standard FDIC coverage through their partner banks, and even high yield from day one. With same day and even same hour liquidity, access your funds anytime. Companies like Scale AI, DoorDash, Service Titan, HIMSS, Anthropic, Flexport, Robinhood, and Plaid trust and use Brex. [08:36] slash sorcery. [09:06] Visit Turing.com/sourcery. Spelt S-O-U-R-C-E-R-Y. That's Turing.com/sourcery. - All right, cool. So we're now standing in what we call Lighthouse Miami. So we were, we got out of the swag room. [09:23] And so Lighthouse Miami is [09:26] What we established here in Miami, it's about a 40,000 square foot facility. [09:30] where we have most of our engineering team, some light manufacturing, and we have the showroom back there that I'll show you outside. And we started out here

[09:39] after moving two times in Miami, we've been constantly outgrowing our space. [09:44] And I think we're at the limit of this space too, so we might have to move out of here. [09:48] soon as well. [09:50] We call it the lighthouse because one of the elements of our product is actually a Fresnel lens. And so as I was telling you, the Fresnel lenses are used, one of the applications for them is in an actual lighthouse. [10:02] So you can kind of use the Fresno Islands two ways. You can take lights that's coming from a wide source, like the sun, and then focus it onto one point. Or alternatively, you can take a point source and kind of beam it out. And that's what lighthouses do. So we call this lighthouse because [10:19] We actually added some light to it before this, it was dark. And yeah, it's really cool space now that the team's [10:26] using to do our development and testing and, uh, [10:30] and also other functions are here as well. So what we're going to do is [10:35] We're going to start actually in the second warehouse. So we have two warehouses here. And I'll show you some of the components that go into [10:43] what makes the pee free and then we'll walk across [10:46] to the other side and go outside and see the [10:49] the production units that are out there. For context, this is [10:53] one of the earlier prototypes that we built is the actual version of that model that you saw there. And we're right now using it as a history piece. Is this the actual size? Yeah, yeah, yeah. So this is a 40-foot shipping container. [11:07] essentially [11:08] This is of course now empty and we're not using it, we're just using it

[11:13] As you can see every year, just to [11:15] write down the names of the folks who've joined us so hopefully yeah we'll keep on going [11:20] Alright, that was fun. [11:27] Do you play Padel? [11:28] Yeah. [11:30] So this is Ultra. This is the one. That's Ultra? Oh, so we're in Magic City. Yeah, exactly. Oh, that's amazing. [11:35] Yeah, so this is the largest Fidel Court in Miami, of course. Oh, that's so fun. [11:40] Yeah, so... [11:41] On this side, we originally started doing some light manufacturing and assembly. Now we've repurposed it to extended R&D. [11:50] One of the coolest things we set up here is this [11:53] kind of room here I'll show you. This is a [11:56] solar simulators. This is the largest [11:59] solar simulator of its kind in the world. [12:02] state of the art and what that means is [12:04] we essentially have the sun indoors. So that allows us to do 24 set of testing. [12:09] So I can show you what it looks like. [12:11] That's an image of it from [12:13] Do you, can you get tan in there? Can you also use it as a tan? Yeah, you can get very tan, yeah. [12:20] Yes. [12:24] Um, [12:25] Inside here [12:29] and they're doing some construction on it. [12:32] You see you have these lamps essentially [12:35] and [12:36] They're going to be beaming light [12:39] It gets so hot that the paint will actually chip off

[12:43] Yeah, people don't expect to see a company like this in Miami. No, I'm used to this in El Segundo. It's actually I'm getting very excited right now. I'm like, is that a CNC machine? [12:52] Yeah, people like... [12:54] Wow, you could do this in Miami? That's so cool. Yeah. [12:57] Whenever we say we're in Miami, people think we're at the beach and, you know, [13:01] No, this is hardcore. [13:04] So I wanna show you the lenses here just to begin with. [13:09] Um... [13:10] So this is, again, as I mentioned, it's called a Fresnel lens. So this is the first element of our system. What a Fresnel lens is, is think of it as a... [13:18] giant magnifying glass. [13:19] So you have this curvature of the lens and if you flatten it, [13:23] you kind of get these grooves. So if you touch this, you feel that there's some grooves to it. So that's kind of the corresponding to the [13:30] curvature of the lens. [13:32] This essentially [13:34] focuses the light coming in from the sun, [13:36] onto the battery material. [13:38] that gets very hot and then stores that energy so [13:41] essentially our battery, [13:43] is a heat battery. And what that means is we're heating up [13:47] Uh... [13:48] rocks, right? And then storing that energy in formal heat. [13:52] which is very cheap. [13:54] as compared to doing lithium ion or any other type of electrochemical battery. [13:59] and you don't have to get lithium or magnesium or cobalt or any of these other fancy chemicals. So that's why we chose a heat battery. [14:06] And this allows us to essentially, as I said, [14:09] captured incoming light in form of heat [14:11] stored in Flint with heat over longer durations, up to five days per cell,

[14:16] and then dispatch it around the clock for data center applications. [14:20] And so the way we do that is, [14:25] You can see some units here in the back. [14:29] So we have these blowers. So the blowers are think of it as [14:33] like your oven, right? When you put something in convective mode, there's a fan spinning that's basically blowing air through the system. [14:39] that they're blowing air through the rocks and then they're bringing it to our engine. So that's the engine. [14:44] One of the engines brings it to the engine heat exchanger [14:47] and then that's [14:48] essentially a Stirling engine. I'll show you what a Stirling engine is. It's a very simple engine. [14:52] a 200-year-old technology, very reliable, [14:55] um [14:56] hundreds of millions of hours of runtime on it. [14:59] and that's what produces electricity. And so you kind of combine these together, string them along, and then you can get [15:05] to any power profile or any energy profile you need. [15:08] The simplest way of explaining the technology [15:11] is like the five-year-old version [15:14] is imagine if you're out [15:16] in the woods or at the beach and you find a rock, right? And you grab the rock [15:19] and you have a magnifying glass in your pocket, [15:22] and then you hold a magnifying glass to the sun and then you heat up the rock. The rock's gonna get hot [15:26] You just wrap that in a blanket and put it in your pocket, that's kind of your battery. [15:30] and then you stick a sterling engine on, which I'll show you, [15:32] and that's going to basically start running based off that heat [15:37] differential. [15:38] and make electricity. So that's really what we do. It's not really rocket science. [15:42] But it's rock science. It's rock science, exactly. Yeah, yeah.

[15:47] Now, so we'll show you some of the rocks as well. [15:50] I love a good dad joke. [15:53] - - [15:57] So what are the rocks? What type of rocks are they? [16:00] You could use any kind of rock. So we have developed a specialized rock that can handle the temperature profiles and go through the cycles a lot. [16:07] but [16:08] We're maintaining the rocks at about 800 degrees Celsius to 1000 degrees Celsius. [16:13] So that's [16:15] equivalent to about 2000 degrees Fahrenheit and [16:18] You can use any rocks, riverbed rocks, you could use fancy ceramics. There's nothing [16:25] that's necessarily [16:29] secret to what kind of rocks you use. What's really important is how you transfer the heat to it, make sure that you've done testing. [16:36] to make sure that the rocks don't break over time and they can handle the cycles and that's [16:40] I love the testing that we've been doing and [16:42] have done and developed our material around it. [16:45] The most important thing is [16:47] Think about [16:48] the raw materials that we use [16:50] At the end of the day, [16:52] It's essentially [16:53] sand and dirt [16:55] and steel, right? [16:57] The sand is essentially quartz that's converted to the lenses. [17:01] to the installation that we use, [17:03] to even the [17:04] Um, [17:05] the rocks essentially and same with the dirt. [17:08] And then the steel is just for the enclosures and the engine parts. [17:11] and [17:12] really nothing fancy about it and that's really one of the [17:14] key [17:16] uh, [17:17] value props here is that you can make this domestically using very simple raw materials, and you don't have to go import something from China.

[17:25] And it's really important because you think about we're building the backbone of [17:30] our AI infrastructure [17:32] on borrowed technology, right? And so we're trying to do this here domestically and I think [17:37] We're just getting started so we have lots of room to grow but [17:41] there's a very clear path to how we can support [17:44] data center growth and in general, [17:46] other power applications, heat applications, industrial applications, [17:50] at the multi-gigawatt scale very soon. [17:52] This is the other side where we do a lot of R&D. Just to tell you why I said this is a production studio. So this room you see here, this was a projection studio. [18:01] So this is where the projector was, that's the kind of opening where the projector was [18:05] And inside there was a, like a movie, like a, like a... [18:09] Cinema, yeah, screening studio, yeah. [18:12] So, [18:13] Anyways, we've got lots of various [18:15] uh r d components here that that i can talk you through but the the main highlight is if we go outside [18:22] We can show you actually how all this comes together. [18:26] We're currently maintaining some of the machines, so all of it is not running, but I'll show you kind of the elements of it as we go outside. [18:34] Again, you ask like why in Miami? People don't expect to see this in Miami, but the fact is [18:39] First of all, you have this sun, which you wouldn't otherwise have in San Francisco all the time. Second of all, you've got a lot of great talent. [18:46] because South Florida is actually [18:49] the center of [18:51] a lot of the energy activity in the US, like the largest energy company, a renewable energy company in the country.

[18:57] is here and then you got a lot of data center developers also in South Florida as well so [19:01] We've got lots of great talent, lots of great investors, and yeah, lots of great stuff to show you here. [19:07] I'm gonna put on my sunglasses because otherwise I'll get a headache. [19:12] But yeah, so what you see here is how this all comes together. So you've got the lens table at the top, right? [19:18] each one of these is sitting on a 20-foot skid. So inside you saw a 40-foot skid. We've now downsized it to a 20-foot skid for ease of handling and manufacturing. And then that light is [19:28] getting focused onto the battery cells or the rocks that's essentially in these enclosures. And then we blow air using the blowers that you saw into this. [19:37] first shipping container here, that's essentially the [19:40] where the PCU, the power conversion unit engine sits, and then that air gets looped back in. So it's not like we're taking ambient air and then heating it up and then making electricity out of it. It's actually that we are cycling the air every day about 200 degrees Celsius. So [19:57] We're going from [19:58] 600 to 800 or 800 to 1000 degrees Celsius and that's what's helping run the engine. So it's [20:04] high temperature, but not extremely high temperature where you've seen some of [20:08] you know, some companies make batteries that are 2,000 or 3,000 degrees Celsius. So this is [20:13] what we have today. [20:14] And then the optical table again, of course, tracks the sun. Right now, this is not hooked to the grid, but rather our EV chargers.

[20:31] You can see the Rivian Charter here. [20:33] Actually, if we go to the other side, I'll show you some other cool test stuff that we're doing. [20:42] If you come inside here, [20:45] We have these containerized engine test stands. So these are test stands that can run round the clock for our various engines. [20:53] They're currently, as you can see, fixing one of the test stands. But we build these because they can be separated from the building and we can run them all the time without having to worry about any security or safety issues. [21:05] Here you have another test stand which [21:08] Um... [21:10] is used for essentially various temperature profiles that we run the battery at. [21:15] And I'll show you what's inside here when we go inside. [21:21] So this is an engine running. You can see the engine at the very far back. If you come around, that thing at the far end is the engine. And then all these pipes are just blowing hot air into it. So this is inside that shipping container that I showed is a similar setup where we're just getting the hot air from the battery cells, the rocks into the engine. Right now this is just doing it from the ambient. [21:51] He's in heaters. Yeah, we can close this for safety. It's like every time you open a different container there's like a little city inside. Yeah, yeah, exactly. Surprise!

[22:03] Quick note before we keep going. I've been investing on public for a while now, and it's a really solid platform that combines stocks, bonds, options and crypto with incredible AI tools that let you do things like build a completely custom investable index from a prompt. Right now, public will give you an uncapped 1% match when you transfer your portfolio. Check it out at public.com slash sorcery public investing for those who take it seriously paid for by public investing. Full disclosures in the description. [22:30] VCX by Fundrise, the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View the portfolio at GetVCX.com. That's GetVCX.com. Founders ship faster on deal. Set up payroll for any country in minutes, hire anyone anywhere, get visas handled fast, and get back to building. [23:00] D-E-R-Y. [23:01] Enterprise AI runs on Merge, the AI infra platform for integrations, agent tooling, and model orchestration, so your teams ship product, not plumbing. [23:11] Mistral, Dropbox, and Drada already trust Merge and production. [23:15] start building at merge.dev. We have rocks here, lots of rocks, but I'm not collecting them yet. [23:26] Yeah, if you come here to the left [23:31] Gosh, this space is huge.

[23:33] Yeah, it's about, as I said, 40,000 square feet. [23:37] So, [23:38] As I mentioned, this is kind of what we're talking about. [23:41] various [23:42] Rocks. [23:43] But... [23:44] This is where you heat up and store the energy. And essentially the rest of the stuff you see also [23:50] This is our materials lab. We have various kind of devices here for testing. [23:54] We have ovens, we have all sorts of [23:56] microscopes and everything else but [23:58] We have a [24:00] Um, [24:01] Where's our cabinet? [24:03] various [24:04] rock samples, as you can see. Wow. [24:06] here as well. But yeah, this is essentially [24:11] um [24:12] how the XO at P3 works and really the vision and ideas [24:16] to scale this to the millions and billions of units, we started productionizing this first generation that you saw outside. [24:23] last year and now we're working on the next generation that's even simpler [24:26] and cheaper and gonna scale that significantly this year to ship to [24:32] some of our Hyperscalar customer projects. [24:35] We've been very fortunate to have a backlog of [24:38] over 10 gigawatts at this point working with the leading hyperscalers. And I know that sounds like a lot, but [24:44] um [24:45] the pace at which data centers are now building, it's not a lot in charge. [24:51] When we started the company three years ago, I remember [24:54] And I was new to the data center world. [24:56] a big data center, I think the largest data center at that point was [24:59] 100 megawatts. [25:00] Mm-hmm. [25:02] like a rack was like

[25:04] 50 kilowatts or 100 kilowatts at most. [25:07] Even less than that. [25:08] and um [25:10] You know, when we were talking to customers, it was like, oh, we're building in 25 megawatt blocks or something like that. [25:14] And then... [25:16] As time went by, the kind of large data center became like a gigawatt, and then recently now it's become 10 gigawatts, right? [25:25] and it keeps on growing and then the racks keep on getting more energy dense from like 50 kilowatts an hour [25:30] people are building racks that are a megawatt, which is insane. And sometimes people say, well, [25:37] Um, [25:38] I don't think people contextualize how much energy that is. So if you talk about [25:41] a gigawatt data center [25:43] That's almost a million US households of energy. [25:47] So if a data center says I'm building a 5 gigawatt data center, they're basically building a city that's 5 million households. [25:53] and they're building in a very tiny footprint. [25:56] Although not so tiny. The funny thing about data centers is [25:59] um, [26:01] when they start talking about buildings in acres, you can start realizing how big these buildings are. So like a five gigawatt data center, [26:08] itself, the building, not the power infrastructure, nothing else. [26:11] is going to be 5,000 acres. Oh my god. So that's just the building for the data center. And then you have all the other infrastructure around it [26:20] And I don't think people really appreciate how big these buildings are because historically data centers were built normally around urban suburban areas. That's why you have a lot in Virginia and in North Carolina and Chicago and other places like that.

[26:34] because [26:35] The biggest data center was maybe 100 megawatts. Typically, it says we're like [26:39] 20, 30 megawatts. And the typical approach was [26:43] "Well, I'm going to build a data center. I don't really have to think about power because I'm just going to plug it into the grid and there's going to be power." There was really never this constraint of power. [26:52] as an equation. And now as data centers have become bigger and bigger and bigger, and the hyperscalers want to invest more and more, [26:58] As you've probably seen, like the [27:00] Google's of the world and so and so forth, they're planning on investing a trillion dollars in CapEx and data centers over the next couple of years. [27:07] That means that you have to [27:09] consume more power and the grid is just really not built for that. What that means is the data centers now have to get how to bring their own power online. [27:17] Because it's also important that [27:19] and [27:20] When you plug in a data center, it consumes a lot of power [27:24] from the grid, which also means you and I as rate payers will have to pay for that additional infrastructure that has to be built to support that. And I think that's, [27:33] become a big issue today in the US where these data centers are being built and they're harming the communities, increasing the energy prices, they're taking the [27:41] the scarce resources that the communities have like water. And now we're seeing a shift to building data centers that are no longer connected to grid. [27:49] two or three years ago it would have been [27:51] a sin or impossible to tell a data center customer, we're going to make a behind the meter power generation asset. Behind the meter being like, [27:58] or off-grid being you're not connected to grid, you're not flowing through the grid. [28:02] But over the last year, you've probably seen that it's become

[28:05] you know, norm and unfortunately, [28:09] most of these customers are using natural gas generators to power the data center because it's available and you know, NatGas is [28:15] whatever you just [28:17] burn some gas and you can make electricity. It's, [28:19] certainly not cheap, [28:21] It's, of course, not sustainable. And as you can, if you think about the context of like the war we're in right now, it's not really a good idea to bet your data center power on a gas backbone. So what ExoWatt is really doing, our mission is to build a solar backbone, which is, [28:38] free energy, right? The sun is free, it's available. You can power the entire US 100,000 times over with tiny speck of land of solar panels. Even Elon Musk famously said that a couple years ago and a couple of times. And the only kind of [28:53] pushback sometimes we get is like, "Well, solar takes a lot of land, [28:57] But the reality is we have so much land, 41% of the US, [29:01] is empty like [29:03] zero people live there, according to the Census Bureau. So land is really not a constraint. I think it's going to be also good for the communities to not go build data centers in the metro areas where people live and it's going to affect them and affect their health, think their energy prices, to go to these barren land areas and build these data center projects and [29:24] Just recently we launched an initiative that we call AxelRise [29:27] which is to go develop these land sites [29:30] and build multi-Gigawatt data centers on top of them. [29:33] So that's something that we're super excited about.

[29:36] Our Chief Data Center Officer Nick, who recently joined us, is leading that effort and [29:41] You'll be hearing a lot of those projects coming online in the next [29:45] Next year, I would say we'll have some news there for sure. Wow. So in terms of the funding side of ExoWatt, how much have you raised to date and who are your investors? [29:55] Yeah, we've been super fortunate to have the best investors back to company. So as I mentioned, we launched it from [30:01] a venture studio called Atomic. They initially gave us the pre-seed funding, if you will. [30:06] And then in 2024, we came out of stealth and we launched our raise our seed round. [30:11] which was led by Catherine Borlett, Andreessen, and Sam Altman. [30:15] Um, [30:16] And then later that year, about six months later, we raised our Series A. [30:20] and I was led by Felicis and then 8090 and MVP. So in total we've raised about $140 million. [30:29] over the last two years and [30:31] It's been great. I mean, we've got [30:33] tons of great investors with [30:35] tons of knowledge about the industry and also very deep pockets to help us scale the business. But if we think about [30:41] how do we take this to the next level and how do we deploy multi-gigawatts [30:45] scale projects. [30:47] will need even more funding. And so we're going to be making some announcements on that front very soon as well. [30:53] Oh! [30:53] Interesting. Yeah. [30:55] That's exciting. Yes. [30:57] - So with investors like Sam Altman, how are you working [31:02] with strategic partners. And I don't know if it extends just to OpenAI, but these other hyperscalers that are customers, how are you strategically being involved and what have you learned from them?

[31:13] Yeah, I mean, we've been very fortunate to be able to talk directly to all the hyperscalers, learn about their needs, learn about [31:20] kind of [31:21] how we can support them and have relationships with all of them. We haven't announced any of these projects publicly, the commercial projects that we have with them, but [31:30] Hopefully soon we can talk about them as well. [31:32] um but but it's been it's been a learning experience for me i have an energy background i understand how power works [31:38] But [31:40] This how data centers consume power has been a learning experience for us. [31:43] And I think it's also [31:45] been a shifting, evolving story. Just as I said, just [31:49] where they take power from is one topic and then [31:51] how that power is consumed is another topic, especially when you think about [31:57] data centers not only consuming a lot of power, [31:59] But if it's an inference data center or a training data center, they also consume that in a [32:05] not necessarily base load way. So people think, oh, data centers are base load means like [32:09] you have a flat load. [32:11] But it's not like that. It's like going like this all the time. And there's a lot of power electronics that you have to develop to actually handle those transients. And we've been learning about that. [32:21] and developing our own kind of solutions towards that as well. [32:24] And yeah, we're really excited to [32:27] support these hyperscalers in their projects that are coming online next couple of years. [32:31] I really [32:32] hope that we can scale our business as fast as possible to do that, to build this sustainable infrastructure, because otherwise, [32:39] we're going to be left with [32:42] tens or even

[32:43] Yeah, tens of gigawatts of [32:45] Mac gas burners that are going to last for another 20 years. And they might be a quick fix now, [32:52] but [32:53] Um... [32:54] I think the hyperscalers know they're going to pay a price for this later, both in terms of penalties for emissions and also just the cost of energy that's going to go up. So the sooner we can get this to the market and scale it in a very significant way, the better everyone will be off. [33:10] No, no. [33:11] start their carbon offset campaigns a little bit more to make up for that. Yeah, that's an unfortunate fact that is kind of like, [33:19] gone out of the vocabulary of all the hyperscalers is there's no more like sustainability is our priority. [33:26] You know, it used to be one of the objectives, but it's now [33:30] Time to power, right? I need power, I don't care. [33:32] how you make it, squeeze penguins if you have to, and I need the power. [33:38] And then second to that is like the cost of power. [33:42] But that's, to be honest, very secondary at this point. Data centers are so desperate, the AI race is so intense, [33:51] that cost is kind of like secondary for the most part. [33:54] And then [33:55] Third, fourth and fifth are other things probably. And then six or seven is now like, I want to make sure it's sustainable. [34:01] and that's not a fortunate thing it doesn't have to be that way but we're here to fix that [34:07] Given, and you did mention this a bit beforehand, but [34:10] given you're so close [34:12] to the actual build out of these data centers, you understand where they are in this cycle and adoption and that kind of thing.

[34:20] What are the biggest misconceptions or I guess hidden truths about where we are right now [34:26] in the world of data center adoption. [34:31] Yeah, I think... [34:32] Um... [34:34] There's a lot of headlines about data center [34:38] projects being planned or initiated [34:41] I don't think all of them are actually happening. So there's a lot of like phantom data centers that are being... [34:47] announced, let's call it that way. I think... [34:51] data centers are definitely in a rush to get online fast. [34:54] they do have massive supply chain constraints. Two years ago it was maybe NVIDIA GPUs, now it's Power. [35:01] And now it's actually not even [35:04] hardware, it's actually labor. So the biggest kind of [35:07] challenge for data centers today is there aren't enough construction workers [35:11] to build a data center so you have to like import people from other states to a site to build a data center [35:17] And being an electrician these days is actually very [35:20] a high-paying job because there aren't enough electricians to even twist wires and and you know just [35:26] hook things up. [35:27] So I think, um, [35:29] Some people always think that there's a bubble and this all is going to go away, but the reality is if you look at the [35:37] the capex budgets and the spend that's happening from the hyperscalers [35:41] it's [35:42] very large and it's [35:45] It's pretty much planned for the next couple of years. So I would say until 2030, there's no doubt that this is going to continue to scale at this rate. The question is like, what happens after 2030? And is it going to continue beyond that?

[35:56] My data centers are super expensive, by the way, [35:58] You can kind of say, [36:00] rule of thumb, one gigawatt data center or per gigawatt [36:03] you have to spend about 50 billion dollars [36:06] of CapEx. [36:08] which is an insane number to think about, right? [36:12] You know, I was talking when you talk to folks in the industry like, [36:16] Yeah, it's the same numbers more or less that we had like three or four years ago. It just has a couple more zeros behind it. It just becomes larger and larger and larger. [36:23] to the point where [36:25] even the hyperscalers want to go get debt to fund this because they don't want to, you know, maybe they don't even have the capability to fund this from their, [36:33] from the balance sheets. [36:34] But yeah, the data center build out is happening. I think what we're trying to really influence is to make sure that it's done sustainably, [36:43] We're trying to make sure that it's done outside of urban and suburban areas where it could affect the communities. [36:49] I talk about energy crisis. The other aspect is really health. If you think about [36:53] they're burning diesel, they're burning gas, [36:56] in the gigawatts, [36:57] this is going to have a real health impact for people living around those data centers. [37:04] where I think we've got the kind of solution here. Now it's really about scaling and adopting it [37:11] at scale and bring this online [37:14] very quickly. [37:16] So, [37:17] Through all of this and building ExoWatch what it is today, [37:21] What were the biggest lessons that you learned early in your career? [37:24] Yeah, I mean, as I mentioned, I used to work at Tesla for the launch of the Model S when the Fremont factory was getting built. And it was it was an amazing eye opening experience.

[37:34] Tesla wasn't a popular company, we had to fight for a lot of currents to uphill. [37:41] to get there but [37:42] kind of [37:43] really the lessons learned were to first of all have [37:46] an amazing, dedicated, mission-focused team. Like, I think that's the primary ingredient that actually got Tesla through. The second one is to build a great product. [37:56] that customers really want and enjoy. And that's something I also learned at YC, right, from Sam. [38:01] build something people want. [38:03] And then really hardware is hard. [38:06] So you want to try to figure out how to... [38:10] create the shortest feedback loops possible, and learn from that as fast as possible, and scale that as fast as possible. And that's kind of why we chose this modular approach, [38:19] instead of going and building a giant infrastructure project and saying, "Well, we need 10 years and a billion dollars to get to our first [38:24] Then on the side, we said let's build the smallest thing [38:27] learn from that, iterate quickly, and build the next generation, next generation, do that quickly and fast. And then... [38:33] The last thing about everyone coming out of Tesla, you're [38:36] obsessed with vertical integration you want to build everything a house we haven't done that at xawad to the to that extent yet we've started working with contract manufacturers because we know we need to scale fast quickly so they have that infrastructure [38:50] But, you know, Elon... [38:52] really taught us how to vertically integrate everything and build it yourself and build it better, faster, cheaper, and don't take no for an answer. Yeah. [39:01] Amazing. [39:02] Well, it's awesome to see everything that you've built and not only like

[39:06] how cool it is to be involved in this next generation of the AI supercycle, but do it in a sustainable way and create a [39:13] ton of jobs here in Miami. Yes. [39:15] Yeah, exactly. That's the other thing. [39:17] I forgot to mention about data centers. Data centers build-out itself creates some construction jobs, but after it's built, there's not a lot of jobs that are required to maintain it. But we're building factories to build infrastructure for those data centers, and those are going to create tens of thousands of jobs per factory. So I'm really excited about that as we continue to vertically integrate into the future, and [39:39] to create a lot of jobs for all the communities that we're going to be building with. [39:43] Amazing. Thanks so much for coming. Thanks so much for coming, Lili. [39:47] Hey, it's Molly. If you enjoy our interviews, check out our newsletter, Sorcery.bc, where we deliver a once a week top deals and tech headlines email and also go deeper on our podcast interviews. Subscribe to Sorcery today. And don't forget to subscribe to the podcast on YouTube, Spotify, Apple or wherever you listen. Link in description to sign up.