AI traders fear that the power grid gained’t have the ability to feed AI’s “insatiable demand” for energy—so that they’re taking issues into their very own fingers.
OpenAI CEO Sam Altman and enterprise capital agency Andreesen Horowitz are two of the traders who’ve put $20 million into Exowatt, a inexperienced power startup focusing on knowledge facilities, The Wall Street Journal first reported.
They’re a part of a rising listing of traders, from Big Tech to private equity, who’re collectively pouring tens of billions into power and knowledge infrastructure to help future AI improvement. However alongside bets on huge data centers and computing hardware, Altman and different CEOs are placing cash towards unproven, moonshot power startups that promise to ship limitless energy for subsequent to nothing.
Exowatt is a tiny, little-known startup that guarantees to provide AI firms with modular photo voltaic methods offering nearly-free power. Renewable power is a pretty funding for a lot of AI firms as a result of it presents a less expensive long-term various to present power sources, and since some renewable tasks can connect with knowledge facilities immediately without having to undergo the regulatory crimson tape required to interface with the electrical grid.
“The goal of one cent per kilowatt-hour is very achievable,” Exowatt CEO and co-founder Hannan Parvizian advised Fortune. (The Vitality Data Administration reported that industrial energy cost just above 8 cents per kilowatt-hour in January.) “As we build more and more modules, the cost goes down. Eventually, we project that it’s going to be down to the cost of raw materials.”
Miami-based Exowatt, based final yr by Parvizian and Jack Abraham, employs a small crew that features PhDs and engineers with prior expertise at firms resembling Tesla and SunPower, Parvizian stated. Exowatt’s P3 power platform is a modular unit in regards to the measurement of a transport container that works by utilizing solar energy to warmth a bit of fabric known as a thermal battery. The thermal battery can retailer warmth after which convert it to again power via an engine.
Parvizian didn’t specify what materials the thermal battery was made from, however stated it was “similar to some of the other raw materials that have been used for heat batteries in the past by other companies.” The science isn’t new—different firms have for quite a few years used supplies resembling liquid tin or carbon blocks for thermal power storage within the U.S., however the expertise hasn’t been broadly adopted but.
Parvizian stated AI stood out as a very sturdy use case for Exowatt due to the sector’s demand for power and since many knowledge facilities are already situated in sunny areas.
“The majority of the data centers in the U.S…are all built in very high solar radiation areas, which is perfect for solar technology,” Parvizian stated. “We think we kind of found a perfect match here [in terms of] a problem and solution.”
AI improvement requires a lot power due to its computing calls for. Giant language fashions (LLMs), the expertise behind chatbots and generative AI instruments resembling Google’s Gemini, Anthropic’s Claude, or OpenAI’s ChatGPT and Sora, absorb huge quantities of information for coaching and improvement. The information facilities the place that knowledge is saved make the most of essentially the most highly effective and superior pc chips accessible—however these chips are additionally essentially the most power-hungry.
Nvidia provides greater than 90% of the AI chip market, and its hottest chips draw about 1,000 watts of energy, Dell COO Jeffery W. Clarke stated within the pc firm’s most up-to-date earnings name—about as a lot as a typical microwave. Think about that AI firms have tens of millions of those chips working across the clock—Mark Zuckerberg lately stated that Meta alone may have over 350,000 by the tip of this yr—and the power calls for change into large.
Raul Martynek, CEO of information heart operator DataBank, advised Fortune he’s already seeing the distinction.
“Before the [AI chips] showed up, we were…delivering 10 to 20 kilowatts per rack [of computer chips]. Now, we need 40 to 100 kW per rack,” Martynek stated.
Talking on the Fortune Brainstorm AI conference in London last week, Ami Badani, CMO of chip designer Arm, stated that AI might eat a full quarter of all energy within the U.S. by the tip of the last decade.
“It takes 100,000 AI chips working at full compute capacity and full power consumption in order to train [OpenAI’s video generation tool,] Sora,” Badani said. “ChatGPT requires 15 times more energy than a traditional web search.”
Altman and others within the AI house consider that moonshots like Exowatt’s photo voltaic expertise or nuclear energy are the one approach to offer sufficient energy to fulfill AI’s power calls for.
“We still don’t appreciate the energy needs of [AI]…There’s no way to get there without a breakthrough. We need [nuclear] fusion or we need, like, radically cheaper solar [power] plus storage, or something at massive scale,” Altman said at a Davos panel in January.
Altman has also personally invested undisclosed amounts in Helion and Oklo, two nuclear power startups that might gas AI knowledge facilities. And he’s not the one nuclear believer within the AI house: Amazon purchased a nuclear-powered data center in Pennsylvania last month.
For all his speak, Altman’s less-than-$20 million funding in Exowatt is a drop within the bucket for the multibillionaire. Exowatt’s expertise has but to be carried out on a business scale, and the power grid is managing to maintain up with AI demand, at the very least in the meanwhile.
“[$20 million] is a pittance in terms of solving things,” Martynek stated. “I’m sure those funds are going to be directed towards trying to create a proof of concept for that type of energy storage and distribution that they could then scale sometime in the future.”
Exowatt CEO Parvizian advised Fortune that Exowatt expects to have photo voltaic modules on-line this yr.
“All these things…are years in the future. I don’t anticipate this having any kind of real impact in the next three to five years,” Martynek stated. “But certainly, over the longer term, you know, 10 years from now, solutions like…small nuclear reactors and other kinds of innovative ideas will probably be in the marketplace.”
