The tech industry has a power problem. Training a single frontier AI model can consume more electricity than some small countries use in a year. Running inference at scale, the actual deployment of these models to hundreds of millions of users, multiplies that demand continuously. And the grid, as currently constituted, cannot keep up.
So Big Tech is doing what it does best: writing very large checks to make the problem go away. Microsoft, Google, and Amazon have collectively committed billions of dollars to nuclear power projects over the past year, signaling a dramatic shift in how the industry thinks about energy infrastructure.
The Deals Piling Up
Microsoft made headlines last year with its agreement to restart a unit at the Three Mile Island plant in Pennsylvania, a symbolic choice given the site's association with America's most famous nuclear accident. The company has also signed letters of intent with multiple small modular reactor developers, betting that next-generation designs will prove more economical and faster to deploy than traditional nuclear plants.
Google followed with its own SMR deal, partnering with Kairos Power to bring multiple reactors online by the early 2030s. Amazon, through its cloud computing arm AWS, has taken stakes in nuclear development companies and secured power purchase agreements with existing plants.
The pattern is unmistakable. Companies that once championed 100% renewable energy targets are now hedging those commitments with baseload nuclear capacity. The reason is straightforward: solar and wind are intermittent, and AI workloads are not.
Why Renewables Aren't Enough
A large language model training run doesn't pause when clouds roll in. Data centers require consistent, reliable power around the clock. Battery storage can smooth out short-term fluctuations, but storing enough energy to run hyperscale facilities through extended low-generation periods remains prohibitively expensive.
Nuclear offers what grid operators call firm capacity. It runs continuously regardless of weather conditions, producing zero carbon emissions during operation. For tech companies facing mounting pressure over their environmental footprint, this combination is increasingly attractive.
The math has also shifted. New reactor designs promise construction timelines measured in years rather than decades. Small modular reactors can be factory-built and shipped to sites, theoretically avoiding the cost overruns that plagued projects like Vogtle in Georgia.
The Skeptics Have Points
Whether these timelines prove realistic is another matter. The nuclear industry has a long history of overpromising and underdelivering. No SMR design has yet achieved commercial operation in the United States, and regulatory approval processes remain lengthy.
There's also the question of whether tech companies are simply offloading their energy problems onto a technology that carries its own risks. Nuclear waste storage remains politically contentious. Reactor incidents, however rare, can render large areas uninhabitable.
Some environmental groups argue that the billions being directed toward nuclear would be better spent on grid-scale storage, transmission upgrades, and demand response programs that could make renewables more viable for baseload applications.
The Broader Implications
What's happening here extends beyond corporate energy procurement. Tech giants are effectively becoming major players in energy infrastructure, a domain traditionally dominated by utilities and governments. Their financial resources dwarf those of many power companies, giving them unusual leverage over which technologies get built.
This has geopolitical dimensions as well. Countries with streamlined nuclear regulations may attract AI investment that would otherwise flow elsewhere. The competition between the U.S., China, and emerging players in the Gulf states increasingly involves energy infrastructure as a strategic asset.
For now, the deals keep coming. Whether they represent a genuine renaissance for nuclear power or another round of optimistic projections that will collide with reality remains to be seen. What's certain is that AI's appetite for power is reshaping energy markets in ways that seemed implausible just five years ago.
