IO Digest Desk
Event date: 2024-10-15
Google has signed an agreement with Kairos Power to use electricity from small advanced nuclear reactors for its artificial-intelligence data centers, adding one of the world’s biggest technology companies to a growing industry push toward nuclear energy as AI systems consume more power.
The deal, reported by the BBC, would see Google begin using output from the first Kairos reactor this decade and bring more capacity online by 2035. Neither company disclosed the financial value of the arrangement or the locations where the plants will ultimately be built. Those unknowns leave major practical questions unresolved, but the strategic message is clear: Google expects the next phase of AI growth to require large, stable and low-carbon electricity supplies that the existing grid may struggle to provide on its own.
Michael Terrell, Google’s senior director for energy and climate, said the grid needs new electricity sources to support AI technologies and argued that the agreement could help unlock AI’s full potential while delivering clean and reliable energy. Kairos executive Jeff Olson described the partnership as a way to demonstrate both the technical and commercial viability of advanced nuclear energy. Together, those statements frame the deal as more than a procurement decision. It is also a bet that nuclear power can move from policy discussion into real infrastructure for digital industry.
The companies still face a long regulatory path. The BBC said the projects will require approval from the U.S. Nuclear Regulatory Commission as well as local agencies before they can proceed. That means the agreement is a step toward future generation, not an immediate operating plan. But Kairos enters the partnership with momentum. Last year, U.S. regulators issued the company the first permit in 50 years to build a new type of nuclear reactor, and in July it began construction on a demonstration reactor in Tennessee. Those milestones help explain why a buyer the size of Google would be willing to place an early strategic wager on the technology.
Kairos is developing smaller reactors that use molten fluoride salt as a coolant instead of water, a design intended to differ from traditional large nuclear plants. Supporters of such systems argue they could be deployed more flexibly and safely than conventional reactors, while also providing continuous power without the carbon emissions associated with fossil fuels. That combination has become especially attractive to technology companies trying to reconcile rapid AI expansion with climate commitments.
The timing matters. The BBC noted that global energy use by data centers is expected to more than double by the end of the decade, citing Goldman Sachs. AI infrastructure is unusually power-hungry because it depends on specialized hardware that both consumes significant electricity and generates substantial heat, creating heavy demand not just for computation but also for cooling. Those realities have pushed energy planning into the center of AI strategy.
Google is not alone. The report says Microsoft recently reached a deal to restart operations at Three Mile Island, while Amazon said in March it would buy a nuclear-powered data center in Pennsylvania. Industry analyst Somnath Kansabanik told the BBC that Google’s move marks another major step in tech’s embrace of nuclear energy. In other words, the sector is no longer treating nuclear as a distant possibility. It is beginning to sign contracts around it.
The enthusiasm does not erase longstanding objections. Nuclear energy remains controversial because of accident risk, waste disposal and high development costs. But it also offers around-the-clock generation at a time when companies want reliable clean power and national governments are revisiting nuclear expansion as part of decarbonization plans. The United States joined other countries last year in backing a goal to triple nuclear capacity by 2050.
Google’s agreement with Kairos therefore sits at the intersection of two fast-moving transitions: the expansion of AI and the search for firm low-carbon power. Whether the plants are built on schedule or not, the deal shows that future computing capacity is now being negotiated not only in code and chips, but also in reactor designs, licensing calendars and power markets.
