There’s an unwritten rule in climate tech: the flashier and more expensive the reveal of a “transformational” product, the more skeptical you should be.
Exowatt, the startup backed by Sam Altman and Andreesen Horowitz promising a “groundbreaking modular energy system” for data centers, threw a blow-out party at an airport hangar in Los Angeles last Tuesday.
Leonardo DiCaprio was in the front row. Attendees walked down a carpet, Oscars-like, as they arrived. The product was concealed behind a long, black curtain.
As a live DJ queued up “Can’t Stop” by the Red Hot Chili Peppers, the curtain fell away to unveil a 40-foot long, bright orange battery module topped with solar lenses: the Exowatt P3. CEO Hannan Parvizian led the crowd in a call and response reminiscent of the early days of WeWork: “I say Exo, you say Watt!”
As one attendee pointed out, the hype-filled launch recalled Bloom Energy’s unveiling of its “power plant in a box” in 2010; then-Governor Arnold Schwarzenegger joined tech and energy executives to tout Bloom’s hydrogen fuel cells as a clean energy revolution for data centers and commercial buildings. (Today, the company’s fuel cells still run primarily on natural gas.)
The P3, Parvizian explained, combines solar lenses, a heat engine, and a thermal battery into a module that can be deployed at any scale for as little as four cents per kilowatt-hour. If you want one, he added, get in line — there’s already a 1.2-gigawatt backlog of orders.
“Compared to other baseload solutions available to us today, Exowatt P3 is cheaper, more linearly scalable — and most importantly, ready to be deployed today,” Parvizian told the launch party crowd.
Anticipation for Exowatt’s reveal had been building for months. The startup came out of stealth mode in the spring with a Wall Street Journal exclusive, touting OpenAI CEO Sam Altman’s investment in the “game changing” approach to clean, firm power for data centers.
That initial news was short on details about how exactly the tech worked — but its lofty promises and apparent plan to revive concentrated solar (for heat, rather than for power) raised questions at the time.
“Anyone else run the napkin math on the Exowatt idea? I feel like I’m missing something?” Duncan Campbell, VP of project analysis at Scale Microgrids, wrote on X.
Earlier this summer, Parvizian told Latitude Media that the decision to keep specifics of the company’s design under wraps was intentional, an effort to build suspense for the upcoming launch, where it revealed additional pricing and technical details. The reality: the P3 is essentially a repackaging of technologies from the early days of renewables, many of which were abandoned due to cost and performance issues that prevented them from scaling.
The reaction from the media since the launch has been broadly positive, with occasional nods to the complicated history of solar-thermal, and especially concentrated solar power. But questions remain, especially about whether Exowatt’s product can overcome challenges inherent to the technologies it has cobbled together.
A thermal battery, a heat engine, solar lenses
Once the curtain dropped, Parvizian talked through the system on stage.
The P3’s “long-duration heat battery cell” is made of “proprietary silicon-based composite materials.” Those cells can be charged either directly from the sun using Exowatt’s “proprietary designed lens,” or else by using “built-in resistive heaters that can take excess electricity from the grid.”
Five of these cells make up a battery pack. Two packs are connected to a Stirling engine, which, by circulating air, takes energy from the batteries and converts it into electricity, Parvizian told Latitude Media. That setup is encased in Exowatt’s distinctive shipping container-sized bright orange outer shell.
Modules can then be configured in a variety of ways, he added: “Need more power? We add more engines. Need more dispatchable energy? We add more battery packs.”
Each of P3’s components rely on technologies that have already been used in other applications — often unsuccessfully.
First, the Stirling engine. Exowatt’s heat engine of choice has been around for centuries but had limited success in modern energy applications, in part because it tends to be larger and heavier than other options, has a complex design, and comes with high costs.
In the early 2010s, a handful of solar companies leveraging Stirling engines went bankrupt, including Stirling Energy Systems in 2011 and Infinia in 2013. (For his part, Parvizian said Exowatt tested multiple heat engines, and ultimately landed on the Stirling engine because of its “reliability, performance, and supply chain.”)
Meanwhile, at its core, P3 is a modular solar-thermal system. The technology, and specifically its use for power generation, gained a lot of investor and developer attention over a decade ago, but the technology was largely abandoned due to its high price points and geographical requirements.
Jenny Chase, solar analyst for BloombergNEF, said that that’s because solar-thermal generation “combines the disadvantages of fossil fuels — moving parts, the need for a heat engine or turbine — with the disadvantages of photovoltaics (having to collect energy from a large area from a variable resource).”
Craig Lawrence, partner and co-founder of Energy Transition Ventures (which is not an investor in Exowatt), pointed out that when solar thermal technology first emerged, the costs of the competition — batteries and solar PV, largely — were 10 times what they are today.
“If they couldn’t win back then, how do they win now?” he asked rhetorically. “You’d better have an answer to that, if you are doing this.”
One clean energy executive with a background in CSP and thermal storage told Latitude Media that part of his skepticism about Exowatt comes down to the physical constraints of solar power and storage.
“It’s not like some software thing that we can disrupt,” he said. “We are subject to the laws of physics…I’m a little baffled.” He went on to reference a line in Carl Sagan’s 1979 book Broca’s Brain: “extraordinary claims require extraordinary evidence.”
Running the numbers
For this modular design to work, Exowatt will have multiple constraints. The first is a matter of geography.
Exowatt chairman Jack Abraham said the company’s review of data centers in the U.S. found that roughly 60% of them, or roughly 750 facilities, are located in areas with enough direct sunlight to power Exowatt’s systems.
However, Columbia University professor and longtime industry analyst Travis Bradford said that historial precedent suggests that the reality may be much lower.
“CSP only really works in certain, very limited locations, desert locations in mid-latitudes where you get a significantly high number of insolation hours and not a lot of light scattering because of ambient moisture in the air and other things,” said Bradford. “There’s some pretty narrow bands of geography where you can get cheap kilowatt-hours out of any kind of CSP thermal process.”
In other words, Exowatt is starting out with a “really really small market,” said Bradford, and one that is constrained to places where data centers will inevitably also struggle with heat and access to water.
If they couldn’t win back then, how do they win now? You’d better have an answer to that, if you are doing this.
Furthermore, no matter where they’re located, powering a data center with Exowatt’s modules will require a lot of space. For its first year or so, Exowatt will be targeting data centers that are smaller than 50 MW.
A single module can dispatch 25 kilowatt-hours of electricity at 480 volts. Based on its initial target customer, the company’s own estimate calculator suggests that serving a 50-MW peak for 24 hours in a sunny state like Arizona would require up to 1,123 acres. That’s bigger than Central Park, nearly 1,500 football fields — or about one-third of the acreage of the famous Ivanpah CSP facility in San Bernardino, California, just to serve a small data center. A typical 50-MW solar PV system might take up one-quarter the space.
Those space requirements would complicate any of the “off grid” or “baseline power” applications that Exowatt’s executives said the company is planning for.
According to Parvizian, Exowatt currently has over 1.2 gigawatts of capacity in a “reserved backlog” — primarily in the form of MOUs and LOIs. If all pans out for the company, that backlog would require between 22,118 and 26,957 acres of space in a state like Arizona; that’s nearly twice the size of Manhattan dedicated solely to powering data centers.
Of those pre-orders, Parvizian said, approximately half constitute new data centers that are looking for off-grid solutions.
Land use aside, that type of setup won’t come cheap. Exowatt’s site estimates that powering a 50-MW data center in Arizona with 24-hour dispatch durations would cost up to $420 million, with at least $1 million in annual operations and maintenance. And even the cheaper option, with 8-hour dispatch durations, would cost between $118 million and $144 million, with at least $360,000 for O&M. And merely reserving a place in the queue to purchase comes with a $10,000 fee.
Abraham has been quoted as saying that they could charge three times as much for the P3, and data centers would still find them attractive. And that might be true, given that data center operators are less price sensitive than other industrial customers when it comes to procuring energy, especially amid the artificial intelligence race.
But that price insensitivity may not be enough to make the technology attractive for data center operators with very specific requirements for energy delivery — high volumes, and 99.999% reliability.
“It’s a very tough match to make without lots of supplemental backup generators or grid connections, which are going to be pretty expensive,” said Bradford. “The amount of storage you would need to match the generation that could ensure you had that kind of reliability is just a crazy amount of storage, there’s almost no way you could do this on a standalone basis.”
“I think they’re trying to fit a very uncertain solution to a very certain requirement,” he added.
Ultimately, Chase said, Exowatt’s talk of data centers “has nothing to do with the tech.”
“You know what else can supply electricity to data centers?” she said. “Photovoltaics.”
How it all came together
The initial idea for Exowatt came not from its CEO, but rather from Abraham, who is chairman of Exowatt and also the co-founder and CEO of the venture capital firm Atomic.
He said he encountered a heliostat solar plant — a type of CSP — in Israel, on a trip a few years ago. Research sparked by that trip yielded what Abraham described as “a thesis of how to make it viable at scale,” prompting him to call up Parvizian, who he had met previously at an Atomic gathering.
Parvizian had some energy experience — a year as an engineer at Siemens, and another as an operations analyst at Tesla, for instance — and in 2022 was fresh off the sale of his VTOL drone delivery company Volansi. He took the lead on product designs in early 2023, cycling through dozens before settling upon the P3.
Atomic is a funder-incubator that has backed the fast-growing direct-to-consumer healthcare brand hims&hers and the shared living startup Bungalow. However, none of Atomic’s other investments are in the energy space.
In April, Exowatt announced its $20 million seed round from Atomic, Andreessen Horowitz, and Sam Altman. The money flowed on the promise that the P3 will meet the energy needs of data centers, desperate to power the artificial intelligence boom initiated by Altman’s own company, OpenAI.
“Renewable infrastructure for data centers is really urgently needed,” the same clean energy executive with a background in CSP and thermal storage said. “It’s an enormous opportunity to build out all this new computation, and their demand is driving innovations of all kinds in storage.”
However, he added, we already have technologies that are theoretically equipped to meet that need, from enhanced geothermal to long-duration energy storage. “There are a lot of actual, real technologies that we know work that people are rushing to deploy to meet this data center demand.”
Exowatt is promising quick deployment, facilitated by domestic, modular manufacturing. The company relies on third-party manufacturers in the U.S., to insulate the company from geopolitical tension and tariffs. It will also seek to take advantage of tax credits from the Inflation Reduction Act.
And as compared with the massive CSP projects of a decade ago, modular units are certainly quicker-to-deploy, and customizable to the loads of customers of various sizes.
But John Woolard, who is currently the CEO of Meridian Clean Energy and was previously the president and CEO of the solar-thermal company Brightsource, said there is a risk of taking modularity too far. Woolard didn’t want to comment specifically on Exowatt’s design, but he advised paying attention to “the amount of materials per square meter of sunlight collected.”
“Think of a mirror or a lens with minimal infrastructure around it, just a little post. That’s going to cost less than something like a huge box with lenses or mirrors on top,” Woolard continued. “Modularity is good, but it’s got to be modularity combined with minimization of the input materials.”
And Exowatt’s predecessors have also experimented with modularity. Fifteen years ago, the startup Ausra emerged from stealth in 2007 with $40 million in hand, and also boasted a modular approach to its concentrated solar thermal technology. That company, led by the current chief innovation officer of Rondo Energy John O’Donnell, was acquired then ultimately shut down in 2014.
Ambition against the odds
Abraham said that many other clean technologies have been “designed incorrectly” by focusing on “small gains in efficiencies” rather than levelized cost of energy (LCOE), a metric that quantifies the cost of generating electricity over a system’s entire lifetime — including the prices of materials, initial capex, operations, and capacity factor.
Exowatt claims it can achieve a levelized cost of energy of roughly four cents per kilowatt-hour, not including subsidies; the company sees a path to one cent per kilowatt-hour, Abraham said.
That wildly ambitious goal is far beyond what other renewables have achieved. As a point of comparison, a 2023 Lazard analysis found that the LCOE of commercial solar PV started at 4.9 cents per kWh, though federal subsidies brought that cost down to 3.2 cents. (For utility-scale applications, the LCOE starts at 2.4 cents per kWh, though incorporating the production tax credit brings that cost down to near zero.)
And, as Lawrence warns, competing with these mature technologies is highly difficult for any new technology: “The lesson of the last 20 years is, “Don’t try to compete head-to-head with silicon PV and lithium-ion batteries,” he said, adding that those technologies have global supply chains and have been standardized and commoditized. “It’s very difficult as a startup with a proprietary technology to compete with the whole world.”
“But,” he added, “I’m glad entrepreneurs continue to try.”
It is still unclear whether the Exowatt P3 is truly “game-changing.” On the one hand, CSP and thermal storage “have always made sense on paper,” Lawrence said. “The challenges often come down to cost and scale.”
But on the other, he added, with Sam Altman and Andreesen Horowitz backing the company, Exowatt may have the resources to get data centers interested in the product — even if it is a composite of technologies that had been set aside.
A few months ago, when the specifics had not been unveiled, Parvizian told Latitude Media that the company’s true innovation is in the design of the system and how it’s manufactured. And as he closed out Exowatt’s launch party in southern California, Parvizian described the P3 as “an engineering masterpiece,” adding that “it’s a result of years of testing and development, ingenuity, and design.”
What he didn’t say: it’s also built on years of attempts to make other versions of the technology work — attempts that ultimately couldn’t compete.
Editor’s note: This piece was updated on September 20 to include thoughts from Jenny Chase, solar analyst at BloombergNEF.
