Big Tech’s dash for nuclear power

Unlock the Editor’s Digest for free

Big Tech is going nuclear. In the past week, Amazon agreed with utilities in Washington state to support development of four next-generation “small modular reactors”, with a similar deal in Virginia, and took a stake in X-energy, an SMR developer. Google agreed to buy power from SMRs to be built by a start-up, Kairos Power. And last month Microsoft agreed a 20-year power purchasing deal that will entail Constellation Energy reopening a unit at the Three Mile Island plant in Pennsylvania that was shuttered in 2019 (not the one closed in 1979 after a partial meltdown).

The tech industry’s dash for nuclear reflects in part the take-off of power-guzzling artificial intelligence; an AI query consumes up to 10 times the energy of a standard Google search. Goldman Sachs reckons power demand from data centres will grow 160 per cent by 2030. In the US, data needs on top of electrification of transport and a manufacturing revival sparked by “reshoring” efforts are forecast to at least double electricity demand growth in the next decade compared with the prior one.

In Europe, Goldman estimates power demand could grow by 40 per cent from 2023 to 2033. The International Energy Agency declared last week that, after the age of coal and age of oil, the world was entering the age of electricity.

Tech companies know that to get data centres built in countries such as the US, they will have to arrange much of their own power. Their net zero pledges require the sources to be green, and they have already invested heavily in wind and solar. Expanding their portfolios to nuclear energy is understandable, but something of a gamble.

Nuclear has a strong claim, in principle, to be part of the climate solution. It is low carbon, delivers lots of power for decades and doesn’t falter when wind or sunlight does. The trouble is large plants are cripplingly costly and time-consuming to build.

SMRs — reactors up to 300 megawatts, compared with 1,000MW for large nuclear plants — claim to offer a cheaper, faster alternative. Largely prefabricated to cookie-cutter designs, their small size in theory means they can be installed close to where power is needed, and on sites such as former coal plants already plugged into the grid.

But they may face similarly hefty costs to larger units to get designs approved by regulators, in a sector where safety is paramount. They might divert critical investment from proven solar, wind and batter power systems. SMRs also remain unproven. Based on the three SMR-style projects in operation and a fourth being built, the Institute for Energy Economics and Financial Analysis calls them “still too expensive, too slow and too risky”.

Bringing in Big Tech’s financial clout and innovative flair may aid SMRs’ development, and accelerate the shift from largely government-led and financed nuclear development to private financing and initiative (look what Elon Musk has done to the economics of space). But finding ways to reopen or extend the life of existing nuclear stations might prove more feasible; as well as Three Mile Island, a plant in Michigan is being recommissioned.

Either way, the surge in AI-driven data demand even before 2030 means Big Tech will probably have to invest still more in wind and solar. Amid the competition for resources, regulators will have to ensure deep-pocketed tech companies do not lock up large parts of the new energy supply. One option might be to insist clean energy projects for data centres are made large enough also to supply the grid or other customers. There is also scope to use AI to improve energy efficiency in factories, offices and across grids. In the new age of electricity, AI must be not just another energy-hungry mouth to feed, but a central part of the green solution.