American utilities find themselves in an unfamiliar spot. Surging electricity demand from artificial intelligence data centers has them scrambling not just for transformers and transmission lines but for the very fuel that powers the most reliable form of generation available: nuclear.
The pressure comes from hyperscalers like Microsoft, Amazon and Google. They need constant, carbon-free power around the clock. Renewables fall short on that score. Natural gas helps but carries emissions. Nuclear stands out. Yet the fuel to run it faces its own constraints. And the scramble is only intensifying.
Utilities have begun signing long-term contracts for uranium and fabricated fuel assemblies years in advance. Some explore restarting idled reactors or building small modular designs next to data centers. The moves mark a sharp departure from the slow decline nuclear faced for decades. Now it sits at the center of the AI boom.
One recent Reuters report highlights the broader equipment shortages gripping the industry. Lead times for large power transformers now stretch beyond 160 weeks, up from 143 weeks in 2024. Circuit breakers face 125-week waits. Prices climbed 4 to 10 percent. Data centers could claim 40 percent of the U.S. equipment market by 2030, up from 2 percent in 2020. “Equipment availability is becoming the biggest concern for developers as they value time to market so highly,” said Ben Boucher of Wood Mackenzie.
That equipment bottleneck compounds the fuel challenge. Without reliable supply chains for both, new nuclear capacity stays theoretical. Utilities know it. They buy five years ahead. They diversify suppliers. Some ask data center customers to prepay. Smaller utilities struggle most. “They don’t solve everything, particularly for smaller utilities that don’t have the scale,” noted Louis Finkel of the National Rural Electric Cooperative Association.
The original spark for much of this urgency appeared in a Yahoo Finance article that detailed how utilities hunt for every component to keep pace. The piece underscored projections that data center capacity will jump to 110 gigawatts by 2030 from 24 gigawatts now. Those facilities could consume eight times the electricity of electric vehicles. The numbers force hard choices.
Tech giants moved first. Amazon Web Services bought a data center campus next to Talen Energy’s Susquehanna nuclear plant in Pennsylvania for $650 million. The deal gives AWS priority access to up to 960 megawatts. Constellation Energy, the nation’s largest nuclear operator, struck agreements with Microsoft to restart a unit at the shuttered Three Mile Island plant. The goal: deliver power to Microsoft data centers by late 2027 or 2028. Regulators and grid operators still must approve timelines. Delays loom.
Google signed with Kairos Power for small modular reactors expected online by 2030. Oracle’s chief technology officer Larry Ellison revealed plans for three SMRs to support over a gigawatt of AI compute. Oklo, backed by OpenAI’s Sam Altman, received letters of intent for up to 750 megawatts from unnamed data center operators. Its design runs on recycled nuclear waste. AWS also partnered with X-energy on TRISO fuel, a high-integrity form endorsed by the Energy Department.
These arrangements carry risks. A Wall Street Journal investigation revealed tensions in states such as Connecticut, Maryland, New Jersey and Pennsylvania. Removing nuclear output from the shared grid could raise costs for other customers and threaten reliability. American Electric Power and Exelon asked federal regulators to review the Amazon-Talen deal, arguing it could shift $140 million in costs and let the data center act as a “free rider.” Talen called the complaint misguided.
Yet the demand keeps climbing. The International Energy Agency projects global electricity use by data centers will more than double to over 1,000 terawatt-hours by 2030. In the United States, natural gas supplies over 40 percent of data center power today, followed by renewables at 24 percent and nuclear near 20 percent. Nuclear’s share could grow if new projects materialize. But fuel supply must expand first.
Domestic uranium production remains limited. Most fuel comes from abroad. The Trump administration pushed to make 20 metric tons of Cold War-era plutonium from dismantled warheads available for reactor fuel. Five companies, including Oklo, entered advanced talks. Critics cite proliferation risks. Supporters see a way to turn waste into energy while reducing dependence on foreign enrichment.
Centrus Energy plays a pivotal role here. The company operates the only U.S. facility licensed to produce high-assay low-enriched uranium, the specialized fuel needed for many advanced reactors. Recent discussions on X highlight its position as a chokepoint in domestic supply. One post noted that as AI, robotics and data centers drive demand for 24/7 carbon-free power, Centrus sits at the critical juncture.
States act too. Texas, New York, Indiana, Utah and Virginia passed measures to speed nuclear development. Illinois, Kentucky, Montana, West Virginia and Wisconsin lifted bans on new construction. Governors from both parties see nuclear as essential for attracting AI investment. Federal executive orders signed in 2025 directed reforms at the Nuclear Regulatory Commission to accelerate licensing. They also encouraged private investment in advanced reactors for national security and data center use.
Challenges persist. Nuclear projects carry high upfront costs and long timelines. Supply chains for specialized components stay tight. Public acceptance varies, especially near past accident sites. Still, the economics shifted. Tech companies pay premiums for reliable, clean power. Power purchase agreements from Microsoft, Amazon and others provide revenue certainty that traditional utilities rarely offered.
Analysts at Goldman Sachs noted that accelerating power demand from data centers could boost new energy technologies, including nuclear. Uranium producers and SMR developers saw stock surges of 40 percent or more in 2025 on the back of these deals. Constellation’s acquisition of Calpine expanded its reach into Texas and California markets hungry for power.
Even so, natural gas fills much of the immediate gap. Some data center developers build their own gas plants behind the meter to avoid grid queues and regulations. That choice trades carbon goals for speed. It also raises emissions at a time when many hyperscalers tout net-zero commitments.
The nuclear fuel piece remains the least discussed yet vital link. Without secure supplies of uranium, enrichment services and fabricated assemblies, reactors cannot run at full capacity or expand. Utilities now treat fuel contracts with the same urgency once reserved for power purchase agreements. They lock in prices. They seek domestic sources. They explore recycling.
One industry observer put it plainly on X: the AI power cycle reaches far beyond data centers into nuclear fuel supply chains, grid expansion and advanced materials. Fluorspar for chemicals used in semiconductors and batteries enters the picture too. The interconnections multiply.
Utilities that secure fuel and equipment early gain advantage. Those that wait risk watching their data center customers build elsewhere or generate power onsite. The stakes rose with each new hyperscaler announcement. Microsoft alone committed billions. Amazon targets multiple gigawatts. Google and Meta follow similar paths.
Regulators scramble to adapt. The Federal Energy Regulatory Commission fields complaints about cost shifts. Grid operators like PJM Interconnection study interconnection queues that stretch years. Some projects now target co-location at nuclear sites to bypass transmission bottlenecks.
The result is a fragmented but accelerating push. Nuclear restarts. New reactor designs. Fuel supply deals. Equipment orders placed years out. All driven by the insatiable appetite of AI training and inference clusters that can draw hundreds of megawatts apiece.
Whether this delivers on time remains uncertain. History shows nuclear projects often face delays and cost overruns. Yet the private capital now flowing from tech giants changes the equation. It de-risks projects in ways government incentives alone never could.
So utilities press forward. They negotiate fuel contracts. They order transformers. They talk to tech executives about siting the next reactor nearby. The AI era demands it. The old rules no longer apply. And the scramble continues.
WebProNews is an iEntry Publication