In previous lectures, we mainly focused on actual AI, now we look at the systems that power it

Scott Nolan is CEO of General Matter. He says that our current supply chain cannot meet the future energy demands from AI. Thus, GM is a uranium enrichment company for nuclear energy.

origin

Scott did a mechanical engineer undergrad, then did a masters in aerospace, then later in his career he went for a business masters (at Stanford, which is hard to get into but he regrets).

After his masters, he worked with Space X, then after his business masters he joined a VC firm because he was interested in nuclear.

His interest in nuclear started when he was a VC. He talked with firms for 10 years, and he realized there is no way to get nuclear fuel in the US as we buy it from Russia. There is one missing ‘step’ within this pipeline of producing nuclear energy which is General Matter works on.

why is energy important?

Sam Altman, Elon, and Jensen all agree: eventually the cost of intelligence (the cost of AI) converges to AI.

China produces more energy, while the USA is stagnated. He predicts in 10 years USA will just start buying from China since it can’t compete.

In the past, we could get by on stranded energy. stranded energy is supply without demand. Think dams, and wind. This excess energy was usually used by bitcoin mining.

• note: most of the ‘good’ stranded energy sites have been taken up.

Since these stranded energy is gone, only new production moves the needle:

• Think Crusoe and other companies building energy infra in the ocean and space
• But for this lecture we focus on land energy

what do data centers care about?

Ok so AI needs energy, and that energy must be used in data centers.

The biggest factor is high uptime.

• other energy sources like wind, solar aren’t reliable. Storing it in batteries takes a lot of money
• natural gas is a good energy source, but turbines take years to order and are constantly in low stock
• nuclear is safe, has less carbon emissions, and only takes 5-10 years to build.

Big companies (microsoft, kairos power, meta) are looking at nuclear in the long term, and battery storage for the short term.

but nuclear is still a really broad field, what does General Matter do?

Every nuclear reactor needs nuclear fuel refill every 1-2 or 5-10 years. To get nuclear fuel involves 5 steps: mining, conversion, enrichment, deconversion (back to fuel pellet), fuel fabrication.

Enrichment

Currently in enrichment the market share is 40/40/15 russia, europe, china. US only 0.15%

Enrichment is the largest cost input of reacter fuel (nearly 33%)

General matter’s approach.

Enrichment is just refining a fuel / of uranium based on its isotope Then get fissile isotope in enough concentration to run a reactor If you enrich, you can make fuel, 5% fuel, different grades that these reactors run on. You can supply to whoever you want: existing reactors or special advanced reactors

what should college student work on?

Be better off working on something immediately necessary not 20 years into the future Focus on fundamentals - whats important problem thats not getting solved, that won’t be solved by someone else, and align with your skill set

In 2023, we heavily focused on what we wanna solve Yes we wanna solve nuclear. then we picked our team and get deals Then they spent 100 hours weeks solving what they are doing, then picking a site Haloo is advanced reactor fuel, but newer is LEU which is what grid currently consumes We are currently based in LA, building in West Kentucky (He says his approach is the same as SpaceX. Instead of building nuclear from scratch, he focuses purely on enrichment)

we are scaling nuclear FAST

The department of energy for uranium enrichment was set in 2034 but General Matter aims for 2029.

In the next few years we expect 10s of small modular reactors to deploy, with gigawatt-scale reactors taking 5 to 10 years to build in the US.

SpaceX

Scott thinks SpaceX as the only company currently capable of lanching sun synchronous data center satellites into orbit. This would bypass the land energy limitations.

Scott also liked SpaceX approach to the industry. They zeroed in on what mattered instead of trying to solve every problem. They asked “How can we reduce commercializing space into a fundamental piece?” The answer is launch capacity. We measure success from dollars per kilo to orbit

During his time at Space X, Scott worked on propulsion systems and building scrappy test stands. He regrets leaving to get a business masters since he learned more when his company scaled from 100 to 1000 employees.

how the US lost enrichment

Historical Dominance: In the 1980s, the U.S. held 86% of the world’s enrichment capacity.

Market Decline: Following the Cold War, the U.S. began trading with Russia through the “megatons to megawatts” program, down-blending Russian warheads for reactor fuel. This led to U.S. enrichment facilities becoming uneconomical, resulting in the closing of last domestic plant in 2013.