Episode Summary
Executive Summary: The episode argues that AI’s exploding electricity demand is reviving nuclear power: big tech firms, facing grid limits and net-zero pledges, are funding reopened plants and small modular reactors. It traces nuclear’s long decline after major accidents, explains why renewables alone are hard to scale for 24/7 data centers, and concludes that tech capital is now the main force pushing a nuclear comeback.
Main Topics: Big Tech’s AI electricity problem (Priority: 5/5): Microsoft, Amazon, and Google need reliable low-carbon power for energy-hungry AI data centers and can’t depend on strained utility grids alone. The nuclear power resurgence (Priority: 5/5): Shuttered plants like Three Mile Island are being restarted, while major tech firms are backing new nuclear capacity to secure baseload electricity. Why nuclear stalled for decades (Priority: 4/5): The transcript reviews how Three Mile Island, Chernobyl, Fukushima, public fear, regulation, financing, and fuel constraints caused cancellations and stagnation. Renewables, grid limits, and baseload (Priority: 4/5): Wind and solar are presented as intermittent and location-dependent, with storage and grid-following inverter limitations making them insufficient for round-the-clock industrial demand. Small modular reactors as the new bet (Priority: 5/5): SMRs are portrayed as a potentially cheaper, factory-built option that could be deployed in smaller increments, though still largely unproven at scale. Policy, finance, and climate alignment (Priority: 4/5): Governments, banks, and funding programs are increasingly supportive of nuclear as a decarbonization tool, aligning climate goals with industrial power needs. Fusion as a speculative long-shot (Priority: 2/5): The episode briefly covers fusion funding and hype, but treats it as far from commercial viability compared with fission-based nuclear options.
Key Arguments: AI data centers require so much power that existing utility grids cannot supply them, pushing tech firms to seek their own generation assets. Big tech’s net-zero pledges make nuclear attractive because it provides steady, low-carbon baseload power unlike intermittent renewables. The nuclear industry’s decline was driven not only by fear after accidents, but also by economics, regulation, financing, and weak electricity-demand growth in advanced economies. Electrification of transport, reshoring manufacturing, and AI are combining to reverse decades of flat demand growth in developed markets. SMRs could lower costs through mass production and modular deployment, but the sector lacks scale and proven commercial models. Tech firms’ profitability and access to cheap capital make them the only customers likely to finance first-of-a-kind nuclear projects. The broader energy transition may require far more electricity than current planning assumes, making nuclear a practical complement to renewables rather than a replacement for them. Fusion remains highly uncertain in the near term, so it cannot solve current AI power needs.
Data Points: Global nuclear generation peak: 2006 - The transcript says nuclear power output globally peaked in 2006. Global electricity demand growth in 2023: 2.2% - Worldwide electricity demand grew slightly less than the prior year. Global electricity demand growth in 2022: 2.4% - Referenced as the prior-year comparison. Microsoft expected electricity use increase: 5 to 6 times - Microsoft says AI has made its planned electricity consumption far higher than when it made its carbon-negative pledge. Microsoft emissions increase since 2020: 30% - Microsoft disclosed emissions growth, putting it off track for its climate goals. Power for Three Mile Island restart: More than 800 megawatts - Constellation’s reopened Unit 1 is sold to Microsoft under a long-term contract. Microsoft power purchase agreement term: 20 years - Contract with Constellation Energy for Three Mile Island output. Three Mile Island restart date: 2028 - Plant is expected to come online in 2028. Three Mile Island expected operating horizon: Until at least 2054 - Projected run length of the recommissioned plant. Jobs from Three Mile Island deal: More than 3,000 - Constellation’s estimate of employment impact. Pennsylvania GDP impact: $16 billion - Projected economic boost from the restarted plant. State and federal tax revenue: At least $3 billion - Expected tax contribution from the Three Mile Island deal. Amazon investment in SMRs: More than $500 million - Amazon is supporting development of four next-generation small modular reactors via X-energy. Google SMR purchase: 500 megawatts - Google agreed to buy power from six to seven SMRs built by Kairos Power. DOE definition of SMRs: Up to 300 megawatts per unit - The transcript cites the U.S. Department of Energy’s SMR capacity definition. Rolls-Royce SMR capacity: 470 megawatts - Example of a design larger than some first-generation UK reactors. Data center power demand growth forecast: 160% by 2030 - Goldman Sachs estimate for data center power demand growth. U.S. corporate factory investment since 2021: More than $500 billion - Plans for semiconductors, batteries, solar panels, and more. People without electricity: Over 700 million - Used to argue that global demand for power will keep rising. Operational SMRs in the world: 3 - One each in China, Russia, and India. SMRs under construction: 3 - Current count cited in the transcript. SMRs still in design stage: 65 - Early-stage projects worldwide. France electricity from nuclear: About 70% - France is highlighted as the world leader by share of electricity from nuclear. Used nuclear fuel fit: Could fit on an American football field at less than three yards deep - U.S. Office of Nuclear Energy framing of spent fuel volume. Coal radiation comparison: 100 times more radiation - Fly ash from coal power is said to emit far more radiation than nuclear generation. Wind drought in Europe: 2021 - Low wind generation caused power prices to spike. Britain nuclear cost comparison: Six times more per megawatt than South Korean equivalents - Hinkley Point C cost comparison cited in the transcript. France/Finland cost comparison: About 50% less per megawatt than UK - Same reactor design built more cheaply in those countries. Global nuclear power tripling target: By 2050 - Officials at UN Climate Week reaffirmed this goal. AI infrastructure fund: $30 billion - Microsoft and BlackRock launch a fund for AI infrastructure. Potential total investment mobilized: Up to $100 billion - Expected total with debt financing included. DOE Loan Programmes Office allocation: $600 billion - Can be used to fund renewables, advanced manufacturing, and new nuclear projects. Fusion investment: More than $7 billion - Amount of VC capital attracted by fusion startups.
Pivotal Quotes: "the only realistic option is nuclear power" — Narrator: Summarizing the argument that data-center demand and net-zero goals make nuclear the most workable zero-carbon baseload source. "the world is on track to replace all fossil fuel power plants by 2050, [but] this would not be good enough" — Solomon Goldstein-Rose: Used to argue that electrification and development will require much more total electricity than current plans assume. "this volume is equivalent to 18 Olympic-sized swears. Swimming pools filled with Teslas." — Narrator via Microsoft Copilot: Humorous aside translating nuclear waste volume into a more visual comparison.
Implications: AI is becoming an industrial-scale electricity shock, accelerating investment in nuclear, especially for firms that need 24/7 clean power. Expect more long-term power contracts, SMR bets, and policy support—but also high costs, long timelines, and uncertainty around whether new nuclear can scale fast enough.
About Patrick Boyle on Finance
This podcast is all about quantitative finance and financial history. Subscribe to hear about financial markets, derivatives, and how investors use quantitative tools from statistics and corporate finance theory. Included are interviews with some of the most interesting thinkers in finance. Occasional longer form financial documentaries, open up fascinating elements of financial markets history. Patrick Boyle is a quantitative hedge fund manager, a university professor, and a former investment banker. To contact Patrick visit http://onfinance.org Find Patrick on YouTube at: https://www.youtube.com/c/PatrickBoyleOnFinance