.avif)
Nuclear power offers “the mirror image” of artificial intelligence’s energy profile, says Joseph Dominguez, President and CEO of Constellation, operator of the nation’s largest producer of emissions-free nuclear energy. AI’s demand for electricity is spurring economic growth, increasing productivity and reshaping the grid. Now nuclear power, long sidelined, has become an essential partner. While most traditional customers use electricity intermittently, AI data centers demand massive amounts of power continuously. Nuclear plants are the answer, says Dominguez, since they deliver consistent power around the clock—reliably, affordably and with zero emissions.
“Most customers don’t use energy 24/7, but AI data centers do, because they are among the most expensive machines ever created and the technology depreciates very quickly,” Dominguez says. That’s why AI data centers never sleep. They work around the clock at unprecedented scale—requiring as much as 1,000 to 5,000 megawatts in single locations, far more than the largest customers in recent years (a Chicago steel mill: 140 megawatts). This energy appetite has led Constellation to forge partnerships with Big Tech companies, Dominguez says, matching AI data centers’ energy use with nuclear power plants that offer scale, run continuously and produce no air pollution.
A 3,000-megawatt data center consumes roughly the same amount of energy as Philadelphia and its surrounding region on a mild day. This creates a fundamental transmission challenge, by channeling power to a single location rather than distributing it across the grid. Because of this, AI data centers can benefit from being located near power sources rather than relying on long-distance transmission.
Maximizing Clean Energy Resources
The AI boom has revealed three levels of scarcity in the nation’s energy infrastructure, Dominguez says. First, there’s the overall scarcity of zero-emission megawatts. Second, and more critically, there’s a scarcity of clean power that reliably operates whenever customers need it. (While renewables plus storage will eventually compete, he says, they’re currently more expensive than nuclear’s consistent output.) Third, there’s scarcity of transmission for heavy concentrated energy users.
Dominguez offers several solutions. He notes the fastest, most impactful and cost-effective things we can do immediately are invest in extending the lives of our existing nuclear fleet and expanding their output through uprates.
Constellation’s revival of Three Mile Island’s Unit 1—now the Crane Clean Energy Center—is another example. While Three Mile Island remains synonymous with the 1979 nuclear accident, Unit 1 (which was not involved) was “the best-running power plant in the nation” before economics and policy failures forced its closure in 2019, Dominguez says.
Crane’s restart is part of Constellation’s four-pronged nuclear strategy: restarting shuttered plants, relicensing existing ones for extended operation, upgrading current facilities for higher output and building next-generation reactors.
Between uprates and restarting Crane, Constellation will add nearly 2,000 megawatts of new clean and reliable capacity – the equivalent of a brand-new dual unit reactor, like the one recently built in Georgia, but achievable much faster than new construction.
Nuclear also provides something AI companies desperately need: certainty. Unlike gas plants that require indexed pricing due to commodity volatility, or renewables that need backup power, nuclear power companies can offer 20-year fixed-price contracts that eliminate regulatory and fuel-cost risks.
AI-Powered Demand Response
Finding ways to accommodate AI demand in the near term while we build out new generation resources is critical, Dominguez says, noting it buys us time and allows the AI industry to grow and compete globally. Many big data centers have their own back-up generation, which can also serve as a release valve on grid demand during peak times.
Another key tool is demand response, which can maximize the use of existing energy sources rather than building new ones. A Duke Energy study suggests that just 1% customer curtailment industry-wide can accommodate projected data center growth. Building on this, Constellation is using AI tools to implement demand response programs that financially incentivize customers to limit usage during peak hours. These systems predict weather patterns and orchestrate real-time demand management across Constellation’s customer base. These growing demand response programs reduce the need for new power plants and lower prices for everyone on the grid, even if they don’t participate.
The Stakes Ahead
Estimates of how much power demand AI will bring to the grid vary, but most energy experts generally agree that it will be significant, and the industry must prepare for it. However, Dominguez says overreaction to rising demand from AI is the bigger risk, as it will increase costs and deter investments in additional energy supply – including relicensing, restarts and new nuclear. He warns against overregulation as more than 20 states consider laws requiring data centers to fund new generation before grid connection, noting that excessive restrictions could drive development overseas and hamper investment in AI technology, which is critical to America’s economic competitiveness and national security.
“This AI race needs to be won from an economic standpoint, but also for geopolitical reasons and the nation’s very defense. Harnessed wisely, the AI–nuclear partnership could not only power the next wave of innovation, but also secure America’s place in a high-stakes global race.”
.avif)
.avif)
