English
The global electricity demand from AI data centers surged 75% between 2023 and 2024 and is projected to account for over 20% of demand growth in advanced economies by 2030, according to the International Energy Agency. This unprecedented hunger for round-the-clock, carbon-free power has forced technology giants and governments to pursue nuclear energy at a scale not seen in decades. Small modular reactors (SMRs) have emerged as the cornerstone of this shift, with Google signing the world's first corporate SMR purchase agreement and the International Atomic Energy Agency (IAEA) arguing that only nuclear can meet AI's need for reliable baseload electricity. This analysis examines the strategic, financial, and regulatory dimensions of the nuclear-AI energy nexus, assessing whether SMR deployment by 2030 is a realistic path or an overhyped gamble.
The AI Energy Crisis: A Structural Market Shift
Morgan Stanley's 2026 Outlook identifies AI-driven energy demand as a structural force reshaping global markets. Data center electricity consumption reached approximately 415 TWh in 2024 — about 1.5% of global demand — and is expected to double to roughly 945 TWh by 2030. A single large AI data center now consumes as much electricity as 100,000 households. The World Economic Forum's Global Risks Report 2026 ranks energy infrastructure strain among the top short-term threats, warning that grid underinvestment could create power constraints as early as 2027-2028.
Hyperscalers — the largest cloud and AI companies — could spend over $1 trillion on energy infrastructure in 2025-2026 alone, relying heavily on credit markets to finance the buildout. The global energy transition is colliding with the AI revolution, creating what analysts call a "power bottleneck" that traditional renewables alone cannot resolve.
Why Nuclear? The Baseload Imperative
Solar and wind power, while essential for decarbonization, cannot provide the 24/7 reliability that AI data centers require. Nuclear energy offers high-capacity factors exceeding 90%, with zero carbon emissions during operation. The IAEA, in its 2025 SMR Platform Annual Report, emphasizes that advanced nuclear technologies are uniquely positioned to serve large industrial loads with firm, dispatchable power.
Small modular reactors, defined as nuclear fission reactors with a rated electrical output below 300 MWe per module, are designed for factory fabrication and modular assembly. This approach promises lower upfront capital costs, shorter construction timelines, and enhanced safety through passive cooling systems. The nuclear energy policy landscape has shifted dramatically, with the U.S. Nuclear Regulatory Commission (NRC) expected to issue licensing decisions on the first two commercial SMR construction permits in 2026.
Big Tech's Nuclear Bet: The Corporate Off-Take Revolution
Google and Kairos Power: The First Corporate SMR Agreement
In October 2024, Google signed a Master Plant Development Agreement with Kairos Power to purchase energy from multiple SMRs — the first corporate agreement of its kind. The deal targets 500 MW of 24/7 carbon-free power from six to seven reactors, with the first unit online by 2030 and full deployment by 2035. Kairos Power's technology uses a fluoride salt-cooled, high-temperature reactor design with ceramic pebble fuel, operating at low pressure for enhanced safety. The company has already broken ground on its Hermes demonstration reactor at Oak Ridge, Tennessee, expected operational by 2027.
Amazon, Microsoft, and Meta Join the Race
Amazon has invested $500 million in X-energy, backing the development of up to 12 SMRs for data center and industrial use. X-energy priced its IPO at $43 per share for a $3.1 billion Nasdaq debut in April 2026, signaling strong investor appetite. Microsoft signed a 20-year power deal with Constellation Energy to restart Three Mile Island Unit 1, providing 835 MW of nuclear power exclusively for AI data centers. The $1.6 billion refurbishment, accelerated by a $1 billion federal loan, is expected to be operational by 2027. Meta has partnered with TerraPower — Bill Gates' advanced nuclear venture — to explore up to eight Natrium sodium-cooled reactors. Oracle has announced plans for a gigawatt-scale data center powered by three SMRs.
These corporate renewable energy procurement strategies have evolved from wind and solar PPAs to direct nuclear investments, reflecting the unique demands of AI workloads.
Regulatory and Financial Hurdles
Despite the momentum, SMR deployment faces significant challenges. The cancellation of NuScale's flagship Idaho project in 2023 due to cost overruns and insufficient subscription remains a cautionary tale. High upfront capital requirements for factory production, unresolved nuclear waste management issues, and the risk of construction delays all threaten the 2030 timeline.
The OECD Nuclear Energy Agency launched RegLab in February 2025, a regulatory sandbox initiative to reduce uncertainties around AI use in nuclear licensing. The NRC is working to streamline approvals for advanced reactor designs, while the U.S. Department of Energy awarded $800 million to the Tennessee Valley Authority and Holtec for SMR deployment in Tennessee and Michigan. Policy support has also come from the "One Big Beautiful Bill Act" and executive orders directing faster NRC approvals.
The energy infrastructure investment trends of 2025-2026 show that private capital is flowing, but the gap between announced plans and operational reactors remains wide.
Expert Perspectives
"The AI industry's energy demand is not a temporary spike — it is a structural shift that will require doubling global electricity generation by 2050," said Dr. Fatih Birol, Executive Director of the IEA, in a January 2026 address. "Nuclear power, particularly SMRs, must be part of the solution alongside renewables and grid modernization."
William D. Magwood, IV, Director-General of the NEA, stated at the World Governments Summit in February 2025: "We are at a turning point. The convergence of AI and nuclear energy presents both unprecedented challenges and opportunities. RegLab is designed to ensure that innovation can proceed safely and efficiently."
However, critics warn of overhyped expectations. Dr. M.V. Ramana, a nuclear policy scholar at the University of British Columbia, noted: "SMRs have been 'five years away' for two decades. The fundamental economics of nuclear power — high capital costs, long construction times, and waste liabilities — have not been resolved by making reactors smaller."
FAQ: AI and Small Modular Reactors
What is a small modular reactor (SMR)?
An SMR is a nuclear fission reactor with an electrical output below 300 MWe per module, designed for factory fabrication and modular assembly. This approach aims to reduce construction costs and timelines compared to traditional large reactors.
Why do AI data centers need nuclear power?
AI data centers require 24/7 reliable, carbon-free electricity. Solar and wind are intermittent, while nuclear provides baseload power with over 90% capacity factor. SMRs can be sited near data centers, reducing transmission losses and grid congestion.
Which tech companies are investing in SMRs?
Google (Kairos Power, 500 MW), Amazon (X-energy, up to 12 reactors), Microsoft (Three Mile Island restart, 835 MW), Meta (TerraPower, up to 8 reactors), and Oracle (3 SMRs for gigawatt-scale data center) have all announced nuclear deals.
When will the first SMRs power AI data centers?
Google's first Kairos Power SMR is expected online by 2030, with full fleet deployment by 2035. Microsoft's Three Mile Island restart is targeted for 2027. The NRC is expected to issue the first commercial SMR construction permits in 2026.
What are the main risks of SMR deployment?
Key risks include high upfront capital costs, construction delays, unresolved nuclear waste disposal, regulatory uncertainties, and the potential for cost overruns as seen with NuScale's cancelled Idaho project.
Outlook: Realistic Path or Overhyped Gamble?
The convergence of AI energy demand and nuclear renaissance is unprecedented in speed and scale. With 71+ reactors under construction globally, over $3.7 billion raised by advanced nuclear developers in 2025 alone, and corporate off-take agreements accelerating, the foundations for SMR deployment are being laid. However, the gap between announced plans and operational reactors remains wide. The next 24 months will be critical: NRC licensing decisions, factory production milestones, and the first SMR concrete pours will determine whether 2030 is a breakthrough year or another missed deadline.
One thing is certain: the future of AI infrastructure is inextricably linked to the future of nuclear energy. The race is on.
Follow Discussion