The US government is investing $4.4 billion to restart three retired nuclear power plants to support the massive electricity needs of AI data centers. Tech giants like Microsoft and Google are anchoring the project through power purchase agreements. While this highlights the critical link between AI infrastructure and energy, the restart process faces technical and regulatory hurdles that could impact project timelines.
The US Department of Energy is launching a major $4.4 billion initiative to bring three retired nuclear power plants back online. This move is designed to solve a specific problem: the staggering amount of electricity required by AI data centers. The facilities being targeted are the Palisades Power Plant in Michigan, the Crane Clean Energy Centre in Pennsylvania, and the Duane Arnold Energy Centre in Iowa. By returning these to the grid, the government aims to add 2.25 gigawatts of carbon-free electricity to meet surging demand.
This is a shift for the energy sector. Traditionally, companies preferred building new facilities to increase capacity. However, the urgency of AI demand has made restarting old ones faster and, in some cases, more cost-effective. The plan relies heavily on private sector involvement. Tech giants like Microsoft and Google are signing long-term agreements to buy the power, which provides the financial security these restart projects need to move forward. For example, Constellation Energy is leading a $1.6 billion effort at the Crane site, while NextEra Energy is working with Google to revive the Duane Arnold facility.
For investors, this shift highlights the direct link between artificial intelligence growth and utility capacity. As AI infrastructure expands, energy availability has become the primary bottleneck. This has turned nuclear energy, once seen as a declining industry, into a critical component of the AI supply chain. The ability of tech companies to secure stable, 24/7 power is now a key factor in their operational success.
However, the path to restarting these plants is not simple. These sites were not merely idled; they were in the process of being dismantled. This creates significant technical and safety hurdles. The Palisades plant, which is the most advanced in its restart journey, recently faced a temporary delay due to a fuel alignment issue. Operators are finding that restarting a facility is as complex as building one, requiring strict safety upgrades to meet modern standards for plants that were originally built decades ago.
Critics have also pointed to the age of these facilities, with some built between the 1970s and 1990s. Questions remain about whether the cost of upgrading these old structures will increase, or if they will face further regulatory challenges. Investors looking at the energy-AI space should track the execution timelines of these projects and the actual cost of repairs. The final outcome will determine whether this strategy is a repeatable model for solving the AI energy crisis or a one-off solution that comes with high maintenance risks.