Summary

The core call: America enriches zero commercial uranium, and enrichment is the bottleneck to all of nuclear energy if you believe in advanced nuclear energy. Nolan spent 2023 hunting for a company to fix the one broken step of the five-step fuel cycle — mine, convert, enrich, deconvert, pelletize — found nothing for a year, and concluded “if this is going to exist, it’s got to be a new company.” The US was “number one in the world during the ’80s by far” and stopped entirely; today enrichment comes only from Russia and Europe.
Three fuel cliffs frame the trade: HALEU (~20% enriched fuel advanced reactors need, with no reliable supply beyond small DOE allotments), LEU (3–5% fuel for the 94 grid reactors — a “$2–2.5 billion US market”), and eventually the Navy propulsion stockpile. The hard catalyst is January 1, 2028, when the congressional ban on Russian imports removes ~25% of U.S. imports immediately.
Fuel economics make enrichment the choke point: for advanced reactors, fuel can be “more than half of their energy production cost,” and Nolan thinks enrichment will long-term be half of the fuel cost. General Matter’s North Star is dollars per kilogram-SWU — the direct analog of SpaceX’s cost-per-kilo-to-orbit.
The energy macro is stark: GDP per capita vs. energy per capita clusters every country on a line with R² “certainly over 0.8”; the US grid has been flat since the ’90s while China — neck-and-neck in 2010 — will be triple US energy production this year. At current growth rates, data centers “can consume, if they were allowed to, the entire grid by 2030,” and the stranded wind/flare-gas sites that used to absorb them are gone.
SMRs go behind the meter with data centers in the next 5–10 years — Nolan’s “BYOE” (bring your own energy) framing: a fully islanded data-center-plus-reactor campus, with the no-brainer upgrade of oversizing power ~10% (only a 5% project cost) to feed an extra 100MW to the host community and “plummet utility prices” — buying speed-to-power and a welcome mat at once.
The investing playbook, distilled: avoid trends because competition kills you twice — companies compete profits to equilibrium and investors price the theme up. “Trying to find value deals in venture is a dumb idea,” concentrate rather than “dilute your average returns by indexing,” and remember Peter’s line that “the steeper the up round, the greater the undervaluation” — investors anchor on last round when only exit price matters.
Where underpriced companies hide: stagnated, cost-plus, oligopolistic industries (space launch, defense, infrastructure) where fixed interfaces and 30-layers-deep subcontracting calcify architecture. The founder tell: they don’t answer to bounce back to the surface — “here’s the answer, here’s the next question you’re going to ask, and let’s take you all the way down the rabbit hole.”
Nuclear’s unlock is cost, not safety data. It’s already the safest, cleanest base load, but people don’t find data compelling; on first principles it should be potentially much more affordable than fossil fuels (one fuel pellet ≈ a ton of coal). The pitch that lands: “What if your utility bill got cut in half?”

Deep dive

1. The filter for what to work on: important, unsolved, and yours to solve

Nolan’s framework across SpaceX, Founders Fund, and General Matter is one question: “What important problem is there that’s not going to get solved otherwise that somehow I can contribute to?” He joined SpaceX as employee ~35 out of undergrad after interning at Boeing showed him incumbents “riding government cost-plus contracts” would change nothing — and believed even at 30 people that SpaceX “is going to ultimately own the entire space launch industry.”
The path to Founders Fund ran through Stanford business school, where he was “quickly voted most likely to drop out” and nearly left for Square before sitting in on Peter’s law-school class on technology, sovereignty, and globalization. His 2011 thesis: physical-world and hardware companies were the huge ignored opportunity — he’d just left a SpaceX the world didn’t yet appreciate, four years before it was landing rockets.

2. Peter’s imprint: avoid trends, because competition kills you twice

The two-layer logic: a trend means many companies chasing one theme, so “how is it not the case that they compete profits down to economic equilibrium?” — and simultaneously many investors pricing it up, so “where’s your advantage?” Avoid competition on both fronts.
Peter’s method on any deal was orthogonality — “instead of just doing a spreadsheet,” ask why you’re even seeing this investment, through “layers of abstraction that were many layers,” until you hold a genuinely different view that yields alpha. Asked how he affected Peter in return, Nolan’s dry answer: “some of our hardware investments turned out to be pretty good” — nobody, “maybe Elon” aside, predicted SpaceX’s outcome this quickly.

3. Where the unworked-but-important quadrant lives: stagnated cost-plus oligopolies

Nolan’s structural answer to Patrick’s puzzle — why would anything important go unworked? — is cost-plus industries: little incentive for progress or cost reduction, oligopoly stalemate, max-profit pricing at the breaking point, “this fixed market size that never takes off.” Space launch, defense (Anduril breaking it), infrastructure (the Boring Company’s “prime thesis”).
The founder tell from many meetings: great ones explain why everyone else thinks the problem is impossible or is attacking it wrong, and when questioned, “they’re not just giving me an answer and trying to bounce back to the surface… here’s the next question you’re going to ask, and let’s take you all the way down the rabbit hole.”
Outside cost-plus, the pattern can be deep personal obsession held for 5–10 years: Sean Parker’s Spotify memo was “so well reasoned” because of the Napster history — years of waiting for the company with the right geography and licensing formula.

4. Judgment over analysis — and the anchoring error in up rounds

Nolan’s arc as an investor: start knowing nothing and do the work; get better at analysis — which “might actually lead you astray,” because you already know which ones you like; then concentrate into the fewest companies possible and “don’t dilute your average returns by indexing.”
Airbnb was the one his gut didn’t flash yes on immediately. Founders Fund angel-checked the air-mattress era; the reimbursement stance after a guest destroyed a home may have marked the turn mainstream. Then they sliced every market’s share and marketing spend: “you analyze the data and it was like, they’re winning. This is it’s over. They’re going to win.” The big check followed.
On price: “trying to find value deals in venture is a dumb idea” — a cheap round usually signals a company that can’t raise at market. And Peter’s guidance, “the steeper the up round, the greater the undervaluation”: investors anchor on last round’s price when a 2x up round should maybe be 4x — “you don’t get paid based on how close you were to last round’s price. It’s ultimately against exit price.”

5. Investors shouldn’t love ideas; founders must — and the road to General Matter

From the investor seat, loving an idea is “really dangerous”: you compromise on team, and “90% of the time it doesn’t work” — Founders Fund never believed in swapping the jockey. But on the company side “you have to be in love with the idea… it’s not that rational to start a company.”
Nolan never cared about enrichment per se — nuclear and space were “the two industries we were supposed to have” from ’60s sci-fi. The path ran Planet Labs → Crusoe (stranded flare-gas supply) → Radiant (stranded demand: microreactors for Alaskan villages and army bases paying “crazy rates” for diesel). Then every advanced reactor company said the same thing: the bottleneck isn’t the NRC — “it’s actually we cannot get the fuel. The fuel comes from Russia.” After a year of searching for a company to back, “this was the important company that nobody was building.”

6. SpaceX lessons: government true believers, and the vertical-integration mandate

On Dragon and NASA’s COTS program — docking with “the most expensive asset mankind had ever developed” — Nolan found agencies staffed by “true believers” who’d stayed through decades of stagnation because they loved it, and were “super collaborative” with any company that took safety as seriously as they did. The hard problems weren’t collision but thermal and pressure interface requirements.
His anti-subcontracting case is architectural, not just patriotic: the space shuttle may have had “30 layers of subcontractors in this one system,” and a nuclear company recently bragged on an analyst call about ~900 subcontractors needing regional organizers. Every cross-company interface is fixed, so “you end up with a really calcified architecture” nobody can optimize across.
The fix: pull engineering in-house so the electrical and mechanical teams can trade interface requirements “sitting side by side at a desk,” and co-locate first-of-a-kind manufacturing with engineering — the conversation that turns a six-axis CNC part into a laser-cut one at “10x throughput at 1/10 the cost.”

7. Energy is the proxy for prosperity — and the US got caught flat-footed

Nolan’s two favorite charts: GDP per capita vs. energy consumption per capita across every country — R² “certainly over 0.8,” “just so predictive… energy use and production is the ultimate proxy for human prosperity” — and the US grid, which stopped growing around the ’90s while China went from neck-and-neck in 2010 to “triple us on total energy production” this year.
Against the ’90s–2020 view that markets would supply energy when demanded, his rebuttal: “physical world stuff has a timeline,” permitting is slow, and unexpected demand catches you flat-footed — which is exactly now. Data centers “can consume, if they were allowed to, the entire grid by 2030 at this growth rate”; the stranded wind in West Texas and flare gas in the Dakotas that used to host them “are gone,” and gas turbines are backordered.
On what’s really the bottleneck to reshoring: labor responds in a few years (electricians “can make more than people who went and got a master’s”), so infrastructure and permitting are the binding constraints — and merely abundant-but-expensive energy “won’t bring back certain industries.”

8. Why nuclear: safety and cleanliness, but not cost

The case as Nolan builds it: the economy needs base load — “something you can actually design an industrial process around” — and nuclear is statistically the safest and cleanest form of it despite Chernobyl/Three Mile Island/Fukushima, whose actual risk “is still far, far lower than any other form of base load.”
Cost is the unchecked box: decades of doing less nuclear drove costs above fossil fuels, with 10–15-year builds running double budget — “that’s really hard for a utility to stand up to.” But on first principles it should be among the cheapest: one fuel pellet equals roughly a ton of coal, so reactors need less material, less mined product — “just an order of magnitude different.”
His answer to the fear question sidesteps data, which “most people don’t find compelling” against acute, memorable accidents: check the final box instead. “What if your utility bill got cut in half? I think people would suddenly find that extremely compelling.”

9. The reactor taxonomy, BYOE, and the fuel cliffs General Matter attacks

Size maps to market: gigawatt scale (AP1000-class) battles it out on the grid; ~1MW microreactors beat diesel in stranded communities and bases; the 100–300MW SMR middle “is going to find its niche behind the meter with data centers over the next 5 to 10 years” — the ultimate version being a fenced, fully islanded campus. His BYOE add-on: oversize power 10% (a 5% project increase on a 50/50 compute/power build) and give a community 100MW — “a complete no-brainer for everybody” when the real fight is being invited in.
The two handicaps on advanced reactors, in his ranking: “number one is they don’t have fuel to operate” — the showstopper General Matter exists to remove — and number two, cost, where fuel can exceed half of energy production cost and enrichment will long-term be half of fuel cost.
The three cliffs: HALEU (about 20% U-235, with no reliable supply beyond small DOE amounts); LEU when the Russian import ban takes full effect January 1, 2028, instantly removing ~25% of U.S. imports; and eventually the Navy’s enriched stockpile. Strategy: serve the small, urgent, incumbent-ignored HALEU market first — pure Founders Fund “own the small market, then grow” — then phase two into LEU for the 94 grid reactors, “a two, two-and-a-half billion dollar market” in the US alone. On weapons: same technology, but the international consensus line is 20%, weapons grade is “well over 90%,” and countries at 60% claiming energy use are “pretty suspect.”
The North Star metric, mapped straight from SpaceX’s cost-per-kilo-to-orbit: dollars per kilogram-SWU (separative work unit) — priced as a tolling service where utilities own or license the material and pay per upgrade step through the supply chain.

10. Building the company — and why almost nobody makes this jump

Team DNA drove everything, per Naval’s line “the team you build is the company you build” — sharpened to the early team. Nuclear engineers are single digits of headcount (“there’s actually no nuclear reactions in our process”) and scattered nationwide, so General Matter went where the hardware and aerospace engineers are: Southern California. The mandate: “no new physics,” no science project — “we’re engineering cost out of the system, we’re engineering performance up, we’re engineering cost of capital down,” including building an in-house EPC firm rather than hiring a GC, “taken from the Tesla playbook,” because million-square-foot facilities on schedule are the truly underappreciated hard part.
Nolan is the last interview step, screening for people who know “this is going to be a lot of late nights, weekends” against a before-end-of-decade delivery target — and who’d take that over more salary elsewhere.
Why don’t more investors do this? Founders Fund “explicitly selects for people that want to be investors,” and companies get started only as extreme exceptions — here, because “if I don’t do this… it’s actually wrong to not go do that.” His honest comparison: operator quality of life “is not as good,” while VC properly understood “is not a job of labor. It’s really a job of ideas and thinking” — only a few companies per vintage matter.
The kindest-thing answer doubles as the stress test: Peter “tried to beat up the idea” through meta-level questions — is nuclear regulated to death, why grow now? The surviving answer: even if SMRs disappoint, the technology makes LEU too, so worst case is a $2B+ existing US market plus allied demand — “time is then on our side.” Those conversations were 2023, pre-AI-datacenter boom; “now it’s extremely obvious why we need this.”

The Early Days of Anthropic & How 21 of 22 VCs Rejected It | The Four Bottlenecks in AI | Anj Midha Market Overview — April 14, 2026