A solar-thermal startup backed by OpenAI CEO Sam Altman is betting billions of hot rocks can solve AI's power crisis. Exowatt claims its thermal battery technology can deliver electricity at just one cent per kilowatt-hour - roughly 90% cheaper than current grid rates. The ambitious target comes as AI data centers strain power grids worldwide, with Microsoft, Google, and Amazon all scrambling for clean energy solutions.
Exowatt just landed backing from OpenAI CEO Sam Altman for what might be the most unconventional solution to AI's energy crisis yet - billions of superheated rocks stored in modular units that can power data centers around the clock.
The solar-thermal startup's approach couldn't come at a better time. AI workloads are pushing data center power consumption through the roof, with Google's energy use jumping 48% year-over-year largely due to AI training and inference. Microsoft recently signed agreements for multiple gigawatts of clean power, while Amazon is investing billions in nuclear energy projects.
Exowatt's technology works by using concentrated solar power to heat ceramic or rock materials to over 1,200°C during sunny periods. These thermal batteries then release stored heat through steam turbines to generate electricity when the sun isn't shining - essentially turning rocks into a 24/7 power plant. The company claims this approach can deliver electricity for as little as one cent per kilowatt-hour, compared to industrial electricity rates that typically range from 8-15 cents per kWh across major markets.
"The math is compelling when you consider the scale AI companies need," according to industry analysis from Wood Mackenzie. Data centers already consume about 1% of global electricity, but AI workloads could push that figure to 3-4% by 2030. Traditional renewable sources like wind and solar can't match AI's constant power demands without massive battery storage - which remains prohibitively expensive at scale.
The challenge isn't just technical - it's manufacturing. Exowatt needs to ramp production to one million thermal battery units annually to hit its cost targets through economies of scale. That's a massive leap for a company that's still in early commercial phases. Similar thermal storage companies like Malta and Antora Energy are wrestling with the same scaling hurdles.
Altman's investment signals serious Silicon Valley interest in solving AI's power bottleneck. The OpenAI chief has been vocal about energy constraints limiting AI development, recently telling investors that breakthrough AI models require exponentially more compute - and therefore power. His backing through venture firm MVP Ventures suggests Exowatt could become a key supplier for future OpenAI data centers.
The timing aligns with broader industry shifts toward on-site power generation. Meta is exploring small modular reactors for its facilities, while Google recently announced partnerships with nuclear startups. These moves reflect growing skepticism that traditional grids can handle AI's voracious appetite for electricity, especially as models scale toward artificial general intelligence.
Exowatt's thermal approach offers several advantages over lithium-ion batteries for grid-scale storage. Rock-based systems can maintain heat for days without degradation, unlike chemical batteries that lose capacity over thousands of cycles. The materials are abundant and cheap - essentially industrial ceramics rather than rare earth elements. And crucially for data centers, thermal batteries can provide the consistent baseload power that AI training runs demand.
But questions remain about real-world performance and grid integration. Solar-thermal projects have historically struggled with higher-than-expected maintenance costs and efficiency losses. Exowatt will need to prove its systems can operate reliably in diverse climates while maintaining the ultra-low costs that make the business case work. The company hasn't disclosed specific deployment timelines or named data center customers yet - critical details for validating commercial traction.
Exowatt's hot rock technology represents a bold bet on solving AI's power crisis through thermal storage rather than traditional batteries. With Altman's backing and AI companies desperate for reliable, cheap electricity, the startup has both capital and market timing on its side. Success hinges on executing massive scale-up plans while proving the technology works reliably in real data center environments. If Exowatt can hit its ambitious cost and production targets, it could reshape how the industry thinks about powering the next generation of AI infrastructure.
