Google is reportedly close to a multi-year nuclear power agreement worth $1 billion or more with Constellation Energy, a move that shows how the AI boom is increasingly becoming an energy-infrastructure race. Reuters reported on October 6, citing Bloomberg News and people familiar with the matter, that Alphabet is nearing a deal to procure a substantial supply of nuclear electricity for its operations.
The reported agreement is not yet a confirmed contract from Google or Constellation. That distinction matters: the current reporting points to negotiations nearing completion, rather than a formally announced power-purchase agreement. Still, the scale of the reported deal highlights how aggressively hyperscalers are looking for dependable electricity as new AI infrastructure pushes data-center power requirements higher.
Why Google is looking at nuclear power
AI data centers are fundamentally different from many conventional cloud facilities. Large numbers of accelerated servers, particularly those using GPUs and other AI processors, can create far denser and more electricity-intensive workloads. The International Energy Agency estimates that global data-center electricity consumption could roughly double from about 415 TWh in 2024 to around 945 TWh in 2030, with AI-focused accelerated servers accounting for a large share of the increase.
That makes electricity availability a technology bottleneck of its own. A company can order more accelerators and build more server halls, but those facilities still need a stable grid connection and enough generation capacity to operate them at scale.
Nuclear power is attractive in that equation because it can provide firm electricity around the clock without depending on weather conditions. Google has already been investing in several forms of clean, firm power, including advanced nuclear, geothermal and other technologies. In its 2026 Environmental Report, Google said its AI infrastructure buildout is accelerating faster than many power systems can expand and that grid-connection delays, supply-chain constraints and regulatory bottlenecks are becoming increasingly important.
This is bigger than one Google contract
The reported Constellation agreement fits into a much broader shift across the technology industry. Microsoft has backed the restart of the former Three Mile Island Unit 1 reactor, now known as the Crane Clean Energy Center, through a long-term power agreement with Constellation. Other technology companies are pursuing advanced reactors, geothermal systems, dedicated generation and grid-flexibility projects.
The common thread is simple: AI infrastructure is turning electricity into a strategic computing input. In the early cloud era, the discussion around data centers focused heavily on servers, networking, storage and cooling. In the AI era, the question increasingly starts with whether enough power can reach the building in the first place.
That pressure is also affecting semiconductor deployment. Reuters reported this week that Morgan Stanley estimates U.S. data-center developers face a potential net power shortfall through 2028 even after accounting for measures such as behind-the-meter generation and fuel cells. The bank warned that delays to AI deployments could eventually ripple into suppliers of memory, optical and other supporting components.
Why nuclear fits the AI workload
For AI operators, power quality and reliability can matter just as much as total generation. Training and large-scale inference workloads can require enormous amounts of computing power for long periods, while modern data centers increasingly use high-density infrastructure that leaves less room for inefficiency.
Nuclear plants offer a predictable baseload supply, which can complement variable renewable generation. That does not make nuclear a universal answer: projects can face long regulatory timelines, significant capital requirements, transmission constraints and local opposition. But for hyperscalers trying to secure electricity decades into the future, a long-term nuclear arrangement can provide a type of certainty that spot electricity markets cannot.
Google already has experience using long-term clean-energy agreements as part of its infrastructure strategy. The company has also backed advanced nuclear technology through a deal with Kairos Power that is intended to support up to 500 MW of new 24/7 carbon-free electricity for U.S. grids.
India has the same problem coming from a different angle
The story matters to India because the country’s data-center and AI buildout is expanding at the same time that electricity demand is rising quickly. The IEA expects India’s electricity demand to grow by an average 6.4% per year through 2030, adding more than 570 TWh to annual consumption over five years.
India is also pursuing increasingly large digital-infrastructure projects. Google has announced a $15 billion AI hub investment in Visakhapatnam, including a gigawatt-scale data-center campus, new energy infrastructure and expanded fiber connectivity. These projects make the availability, cost and reliability of electricity an increasingly important part of India’s AI competitiveness.
India does not need to copy the U.S. power strategy exactly. The country’s electricity system has a different generation mix, transmission structure and policy environment. But the underlying lesson is similar: AI capacity cannot scale faster than the physical infrastructure that powers it.
The deal is still only a report
There is an important caveat to the headline. Reuters’ October 6 report says Google and Constellation are close to a deal; neither company had publicly confirmed a final agreement at the time of publication. That means the final price, duration, power volume and specific nuclear asset should not be treated as settled details.
What is already clear is the direction of the market. The largest AI companies are no longer treating electricity procurement as a routine facilities issue. They are increasingly making long-term energy commitments alongside decisions on GPUs, networking, cooling and data-center locations.
That is an important change for the technology industry. The next phase of the AI race will depend not only on who has the best models or the fastest chips, but also on who can secure the electricity needed to run them at scale.
For the latest reported details, see Reuters’ report on the Google–Constellation talks and the International Energy Agency’s analysis of AI and data-center electricity demand.
