Skip to content
Stanford OnlineStanford Online

Stanford CS153 Frontier Systems | Scott Nolan from General Matter on Energy Bottlenecks

For more information about Stanford's online Artificial Intelligence programs, visit: https://stanford.io/ai Follow along with the course schedule and syllabus, visit: https://cs153.stanford.edu/ In a CS153 Frontier Systems lecture, the class zooms out from AI model labs to examine energy and electricity as upstream bottlenecks to compute and data center growth, intensified since ChatGPT’s 2022 breakout and renewed enterprise demand after Claude 4.6. Guest Scott Nolan, CEO of General Matter, argues that uptime requirements and turbine shortages make baseload power crucial, pushing hyperscalers toward nuclear for its low carbon emissions and safety record. He explains nuclear’s fuel supply chain and identifies uranium enrichment as the key missing U.S. capability, with the U.S. holding under 0.1% enrichment market share and relying on Europe and Russia. Nolan describes founding General Matter in 2024, winning a $900M DOE contract, building a Kentucky facility, and hiring toward hundreds to thousands of roles. Guest Speaker: Scott Nolan is the co-founder and CEO of General Matter, a company working to reshore U.S. uranium enrichment capabilities and revive American nuclear fuel production. He founded General Matter after spending over a year searching for an American enrichment company to invest in and finding none existed. General Matter is sometimes described as the third in a trilogy of companies incubated at Founders Fund, following Palantir and Anduril. He is also a Partner at Founders Fund (since 2011), where he focuses on companies rearchitecting industries — usually with hard engineering at the foundation. He works with mission-driven founders across biotech, crypto, energy, infrastructure, manufacturing, and transportation, including Synthego, Collective Health, Modern Animal, Branch, Nubank, and others. Prior to Founders Fund, he was an early employee at SpaceX, where he helped develop the Merlin and Draco propulsion systems used on the Falcon and Dragon vehicles and was responsible for the Dragon capsule's thermal and environmental subsystems. After SpaceX, he spent time at Bain & Company, evaluating potential investments and driving portfolio company strategy for private equity clients. He also previously worked as a Systems Engineer at Boeing. He serves on the boards of ISEE, Collective Health, Invisibly, and Synthego, and previously served as a Board Observer at Ayar Labs. Follow the playlist: https://youtube.com/playlist?list=PLoROMvodv4rN447WKQ5oz_YdYbS74M5IA&si=DOJ5amlyRdyMJBhG

Scott Nolanguest
May 11, 20261h 0mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

Energy and nuclear fuel supply chains may bottleneck AI scaling

  1. AI capability progress depends on upstream physical systems—especially reliable electricity—so power availability can halt data center deployment even when chips and buildings are ready.
  2. Industry leaders (Altman, Huang, Musk) increasingly frame energy cost and supply as the fundamental constraint because model operation ultimately consumes electricity.
  3. Short-term growth has leaned on stranded energy and natural-gas turbines, but stranded sites are being claimed and turbine lead times are stretching, making the next few years especially tight.
  4. Nuclear is positioned as the best long-run baseload option for uptime, safety, and low carbon, but scaling nuclear requires a steady fuel supply chain with a critical weakness in enrichment.
  5. The U.S. has <0.1% enrichment market share and still relies on Europe and Russia, so restoring domestic enrichment (General Matter’s focus) is framed as a bottleneck to nuclear expansion and therefore AI scaling.

IDEAS WORTH REMEMBERING

5 ideas

Power delivery can be more urgent than compute availability.

The lecture emphasizes that a ready data center is useless without timely interconnection and generation; electricity constraints can stop training and inference regardless of chip supply.

Electricity is the “universal denominator” of AI cost.

Citing testimony and public remarks, Nolan argues chips and model costs may fall, but energy remains the irreducible input for running models—making it the long-term cost floor.

Stranded energy was a bridge strategy, not a permanent solution.

Early builds (often Bitcoin mining) exploited isolated hydro/geothermal/wind or flared gas, but many of the best sites are now claimed and their scale is insufficient for projected AI demand.

Uptime requirements steer data centers toward firm baseload sources.

Wind/solar can power compute only with substantial storage; with today’s grid-scale batteries, achieving data-center-grade uptime tends to be costly, pushing operators toward gas now and nuclear later.

Natural-gas turbines are a near-term constraint with multi-year lead times.

Even the stopgap solution faces supply limits: turbine scarcity and long lead times, plus delayed availability of grid power-electronics equipment, tighten the next 1–3 years.

WORDS WORTH SAVING

5 quotes

Everything is going to converge to the cost of energy, to the cost of electricity.

— Scott Nolan

Even if you have a data center ready to go, if you can't get power to it, doesn't matter. It, it's over. You can't train models.

— Host

We have to go from almost a complete standstill on grid expansion to nearly vertical.

— Scott Nolan

The U.S. has less than 0.1% market share today of enrichment, which is the middle step.

— Scott Nolan

I wouldn't worry so much about what the, what the public narrative of it is or what very surface-level treatment of it tells you. I would go a lot of clicks deeper, like just go all the way to the bottom and figure out, okay, well, what are we actually solving h- for here?

— Scott Nolan

AI “factory” systems view and upstream bottlenecksElectricity demand growth vs. slow grid expansionStranded energy and Bitcoin mining as an infrastructure rehearsalData center uptime requirements and limits of wind/solar plus batteriesNatural gas turbines and supply-chain lead timesNuclear safety and emissions comparisonsUranium fuel cycle and U.S. enrichment dependencyGeneral Matter’s DOE contract and Kentucky facilityPublic perception, policy support, and bipartisan alignmentJobs and industrial renaissance driven by AI-era infrastructure

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

iOSAndroidClaudeChrome