The U.S. Energy Department just handed a $1.9 billion loan to the owner of an Iowa nuclear plant that Google committed to bring back from decommissioning. Frame it as an energy story and it looks like a bipartisan infrastructure win. Frame it correctly and it is a data center story: Google needs power at a scale that solar and wind cannot guarantee around the clock, and nuclear is the only baseload carbon-free source that can serve a hyperscale campus. The federal loan is effectively a subsidy for AI training infrastructure laundered through energy policy language.
The Fossil Fuel Feedback Loop
The timing is uncomfortable. Brent crude is approaching $100 a barrel, oil infrastructure is being struck in the Middle East, and markets are pricing in an inflation bump from energy costs. Into this environment, Google is spending billions to lock in a nuclear supply chain, a hedge against exactly the fossil fuel volatility that the broader economy is absorbing. The company's forward-deployed engineers going into enterprise clients via the Accenture deal need that power to run the models they're selling. Energy is now a competitive moat, not just an operating cost.
Infrastructure as Politics
Nuclear's political rehabilitation is nearly complete. What was a third-rail issue for environmentalists a decade ago is now being embraced by tech companies as a climate solution, with federal loan guarantees following. The same logic is driving European sovereign AI ambitions: control your compute, control your energy, control your data, and you control your policy options. A 2026 Nature commentary on compound climate risks warned that single-point infrastructure dependencies, exactly the kind nuclear avoids by design, are the underappreciated threat of the coming decade. Google, whatever its motivations, is reading that risk correctly. The question is who else gets to plug into the socket.