Artificial intelligence has triggered an unexpected energy crisis that is forcing America to reconsider nuclear power as a critical infrastructure solution. The Department of Energy awarded $94 million in federal funding to eight companies to accelerate the deployment of small modular reactors (SMRs), advanced light-water nuclear systems designed to come online in the 2030s and expand U.S. power grid capacity. Simultaneously, Microsoft has signed a 20-year power agreement to restart a reactor at Three Mile Island, the site of America's worst nuclear accident in 1979, with operator Constellation Energy planning to spend roughly $1.6 billion to bring the dormant facility back online by 2028.

Why Is AI Creating Such Massive Energy Demand?

The surge in electricity consumption stems from the unprecedented scale of modern AI infrastructure. A single hyperscale, AI-focused data center can consume 100 megawatts of power annually, enough energy to power approximately 100,000 households. This demand is climbing far faster than utilities anticipated. Millions of users querying AI models continuously every day have created a second wave of energy demand that analysts warn could soon rival the electricity consumption of entire towns or small cities. The International Energy Agency and Department of Energy both project that without a significant boost to the nation's energy grid, U.S. power demand could outpace supply within the next five years.

The scale of this challenge has fundamentally altered how technology companies approach infrastructure planning. Meta recently signed a nuclear energy deal to power its Prometheus AI supercluster, while Google has signed agreements to power its data centers through advanced nuclear plants by 2030. In essence, AI companies have shifted from competing for office campuses and software talent to competing directly for electricity itself.

What Makes Small Modular Reactors the Solution?

The DOE's $94 million investment targets eight companies developing SMR technology under the Generation III+ SMR Pathway to Deployment Program, a $900 million initiative launched to accelerate nuclear energy development. The companies receiving funding include:

• Constellation SMR Development and Nebraska Public Power District: Collectively awarded more than $45 million to obtain U.S. Nuclear Regulatory Commission early site permits for future SMR deployments
• BWXT Nuclear Energy: Received funding to support supply chain development for SMR components
• Container Technologies Industries: Developing supply chain infrastructure for modular reactor deployment
• Framatome U.S. Government Solutions: Supporting supply chain advancement for SMR technology
• Global Nuclear Fuel Americas: Contributing to fuel supply chain development
• North American Forgemasters Company: Providing manufacturing support for reactor components
• Scot Forge Company: Supporting industrial manufacturing for SMR deployment

SMRs offer advantages over traditional large reactors because they can be deployed more flexibly across the grid and constructed faster. Several SMR designs have already received U.S. Nuclear Regulatory Commission approval, and a few operational examples exist globally, making them more proven than experimental alternatives like orbital data centers. NuScale Power, the only U.S. company with an approved SMR design, has several megaprojects in its pipeline, including a 6-gigawatt project with the Tennessee Valley Authority, with first projects expected to come online in the early 2030s.

How Are Tech Giants Securing Nuclear Power?

The nuclear energy race reflects a broader shift in how AI infrastructure companies approach long-term planning. Microsoft's commitment to restart Three Mile Island represents the most symbolically significant move, reviving a reactor once viewed as politically radioactive and economically obsolete. The decision signals that tech giants now view stable baseload power as essential to their competitive advantage in AI development.

Energy Secretary Chris Wright stated the administration's commitment to this strategy, saying:

"President Trump has made clear that America is going to build more energy, not less, and nuclear is central to that mission. Advanced light-water SMRs will give our nation the reliable, round-the-clock power we need to fuel the President's manufacturing boom, support data centers and AI growth, and reinforce a stronger, more secure electric grid."

Chris Wright, Energy Secretary, Department of Energy

The federal investment aligns with executive orders signed by President Trump in May 2025 that aim to accelerate nuclear energy development by fourfold over current production levels within the next 25 years. This represents a dramatic policy shift, as nuclear energy had faced decades of political and economic headwinds before AI's energy demands reshaped the conversation.

What Does This Mean for Global AI Competition?

The nuclear energy race is becoming a geopolitical competition. Countries dominating AI development may no longer be determined solely by machine learning breakthroughs but by their capacity to sustain massive, stable energy systems. China continues expanding its nuclear capacity aggressively, Russia operates fast breeder reactors commercially, and India recently pushed its long-delayed Kalpakkam breeder reactor to criticality while linking nuclear growth to its broader development ambitions. Governments worldwide are recognizing that AI leadership depends as much on megawatts as on algorithmic innovation.

The infrastructure transformation extends beyond reactors. Technology companies are increasingly reshaping national energy policy, securing long-term power partnerships, and influencing grid development simply to sustain their operations. This represents a fundamental shift from the era when tech companies competed primarily through software innovation and talent acquisition.

How to Understand the Real Cost of AI Infrastructure

The revival of nuclear plants exposes a critical reality that has been obscured by the abstract language surrounding AI technology:

• Physical Resource Intensity: Every AI model response, image generation, and search query runs through physical systems consuming land, water, minerals, and enormous quantities of electricity, making AI fundamentally dependent on industrial infrastructure rather than existing as a weightless digital service
• Supply Chain Complexity: Deploying nuclear reactors requires developing entire supply chains for specialized components, manufacturing capabilities, and regulatory compliance, creating multi-year timelines and billions in capital investment
• Grid Infrastructure Limitations: Current electricity grids were not designed to handle the concentrated power demands of AI data centers, necessitating fundamental upgrades to transmission systems, substations, and generation capacity across regions
• Long-Term Energy Partnerships: Technology companies must now secure 20-year power agreements and influence national energy policy to guarantee the stable baseload power required for continuous AI operations

The symbolism of restarting Three Mile Island underscores this reality. For decades, the site represented nuclear fear and the limits of industrial ambition. Now it represents the material foundation that AI infrastructure requires. The cloud, as the saying goes, still runs on steel, concrete, and electricity. What began as a software revolution has become an industrial one, with technology companies increasingly resembling energy companies themselves in their strategic priorities and long-term planning horizons.