Google’s near-$1 billion nuclear power deal tests how fast AI can decarbonize
Google is close to a roughly $1 billion agreement with Constellation to secure nuclear-generated electricity for its data centers, according to public reports. The prospective deal shows how the race to power AI infrastructure is pulling tech giants into long-term bets on nuclear energy and reshaping energy markets around digital demand.
Google’s cloud business is turning to nuclear power to keep its data centers humming. The company is reported to be nearing a deal worth close to $1 billion with U.S. utility Constellation for nuclear-generated electricity, a sign that the scramble to fuel artificial intelligence infrastructure is starting to redefine who locks in long-term baseload power.
The prospective agreement, disclosed in broad terms on 6 October, would see Google commit to buying a significant volume of nuclear electricity from Constellation to supply its data centers. While exact capacity figures haven’t been made public, the value of the arrangement points to a multi-year, large-scale contract. For Google, the attraction is clear: nuclear plants offer around-the-clock, low-carbon power in a grid increasingly strained by intermittent renewables and rising digital demand.
AI workloads are particularly power hungry. Training and running large machine-learning models requires dense clusters of servers, advanced cooling systems, and resilient backup. As Google competes with Microsoft, Amazon, and others to dominate AI services, securing reliable electricity becomes a strategic priority as important as buying chips. Outages or supply constraints can slow product rollouts and raise operating costs, eroding the advantage of being first to market.
For Constellation and the broader nuclear sector, a long-term arrangement with a marquee technology client provides something utilities crave: predictable revenue and political validation. Nuclear plants in the United States have struggled in recent years to compete with cheap gas and subsidized renewables in wholesale markets. Having a deep-pocketed buyer willing to pay for zero-carbon attributes and long-term stability can help keep reactors online and justify investments in life extensions or advanced designs.
The deal also signals a shift in how climate goals intersect with corporate energy strategy. Google has publicly pledged to run on carbon-free energy around the clock in coming years, a step beyond the widely used model of simply buying enough renewable certificates to offset annual usage. Nuclear power, with its steady output and low direct emissions, helps close the gap between variable wind and solar generation and the 24/7 load of data centers.
Communities near nuclear plants could feel the impact as well. Long-term contracts tied to digital infrastructure can reduce the risk that reactors are shut early for economic reasons, which would otherwise threaten local jobs and tax bases. At the same time, coupling data center growth to nuclear raises questions about water use, transmission build-out, and whether communities see benefits from hosting both reactors and high-density server farms.
Strategically, if this deal is replicated across the sector, it will deepen the link between America’s digital dominance and its nuclear fleet. The United States has the largest number of operating reactors in the world and a tech industry that leads in AI infrastructure; connecting the two more directly could give Washington an edge in both climate diplomacy and technological competition with China, which is rapidly expanding its own nuclear and data center capacity.
There are risks. Locking in large amounts of nuclear power may crowd out some investment in storage and demand-response technologies that could balance renewables. Public opposition to nuclear in some regions, and unresolved questions over waste and new plant economics, could still limit expansion even if data center demand is strong.
One line distills the shift: AI doesn’t live in the cloud, it lives in concrete buildings wired to very real power plants, and those plants are now becoming part of the tech arms race. That makes energy policy decisions about nuclear and grid investment directly relevant to the speed and shape of AI deployment.
The next indicators to watch will be the final structure and term of the Google–Constellation contract, whether other hyperscale cloud providers sign similar nuclear-backed deals, and how regulators treat long-term power purchase agreements that tie digital infrastructure to specific plants. Movement in regional power prices and grid planning documents will show how quickly this new demand changes investment priorities for utilities and transmission operators.
Sources
- OSINT