
From SpaceX to Founders Fund to Solving America's Nuclear Fuel Problem
What this covers
Patrick O'Shaughnessy talks with Scott Nolan, an early SpaceX engineer who became a partner at Founders Fund and now runs General Matter. The first stretch is about how he picks problems. He covers what he took from Peter Thiel on the Avoid-Trends Heuristic, why a venture investor should concentrate capital in as few companies as possible, and the danger of falling in love with an idea. He also explains why aerospace stagnated, pointing to cost-plus contracts and deep subcontracting. He cites nuclear companies with around 900 subcontractors, and 30 layers in the space shuttle.
The middle turns to energy. Nolan treats per-capita energy use as the ultimate proxy for prosperity and contrasts flat US generation since the 1990s with China's growth. He forecasts that AI data centers could absorb the whole US grid by 2030, a claim he states firmly but one that is contested. His case for nuclear starts from fuel density, one pellet roughly equal to a ton of coal. He then has to explain why costs have risen anyway, citing 10-to-15-year builds and overruns. He puts public fear down to an acute-versus-chronic bias, and he dismisses solar with storage as mainly peak shaving rather than base load.
The last section is the fuel chain. Of its five steps, the US lacks only enrichment. Nolan walks through three "fuel cliffs": HALEU supply, the 1 January 2028 ban on Russian imports, and the eventual depletion of the Navy's stockpile. He admits that demand for new capacity has been visible for only about 24 months, so General Matter is still betting on growth. The close covers building an engineering-led hardware company, including co-locating manufacturing with design, and the trade-offs between investing and operating.
Nolan argues that the best opportunities sit in neglected, stagnant physical-world industries, and that US uranium enrichment, the missing step in the nuclear fuel chain, is a bottleneck urgent enough that it needed a new private company.
- Stagnant cost-plus industries lack incentives for progress, so new entrants can win by targeting them
- Outsourced manufacturing and subcontracting erode the learning and integration needed for innovation
- US enrichment capacity was lost, and the HALEU and LEU fuel cliffs make it the binding constraint on nuclear and energy growth
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
The enrichment technology is the same whether for civilian reactor fuel or weapons; the difference is application, and international consensus holds that enrichment should stay below 20%, while countries enriching to 60% are suspect because they are approaching weapons grade (well over 90%).
“ultimately it's not different technology, but it's applied in a different way and in a completely non-international consensus setup.”
Public fear of nuclear is driven by an acute-versus-chronic bias: people remember dramatic events like Chernobyl and Three Mile Island, but ignore the chronic background harm from fossil fuels, and data alone does not change minds.
“acute events that are very attention attention-worthy, people remember those. And the things that are just low-level in the background, people don't think about”
Learning-by-doing follows a predictable learning curve across iterations, and because manufacturing creates much of the innovation, offshoring manufacturing means losing the innovation capacity that comes with it.
“this whole learning by doing thing which has now been carefully studied... there's like a a literal predictable curve to these things”
Across countries, GDP per capita and energy consumption per capita are tightly correlated (R-squared likely above 0.8), so energy use and production is effectively the ultimate proxy for human prosperity and economic activity.
“GDP per capita versus energy consumption per capita. And it's almost like the R-squared on it is, you know, probably like certainly over 0.8.”
Nuclear has the lowest deaths-per-energy risk among base-load sources and no carbon emissions, so the high-profile accidents (Chernobyl, Three Mile Island, Fukushima) overstate its actual risk compared with fossil fuel plants that cause known health harm from emissions.
“the actual risk associated with nuclear, despite those high-profile incidents, is still far, far lower than any other form of base load energy.”
The US lost its enrichment capability despite being the world leader in the 1980s, stopped doing it entirely, and as a result enrichment became the bottleneck for all advanced nuclear energy.
“you realized the US had lost the ability despite being the number one in the world during the '80s by far.”
The incumbent aerospace industry stagnated because it rode government cost-plus contracts with layers of subcontractors, and the US assumed space launch would be a nation-state capability that never made commercial sense, so only a new company could change it.
“They were all just riding government cost-plus contracts. The US assumed that space launch was a nation-state capability that would never be a commercially interesting thing to do.”
The nuclear fuel supply chain has five steps (mining uranium, converting it to gas, enriching it, deconverting it back to solid, and forming fuel pellets), and the US has commercial capability in all steps except enrichment, which is the bottleneck.
“the US has good capability across the board except in enrichment. It's the one area that we don't have any commercial scale capability.”
Competition operates at two layers: company-level competition, and theme-level competition where a trend attracts many entrants. Trends should be avoided because new entrants compete profits down toward perfect-competition equilibrium, and because investors who see the trend already price it in, eliminating any edge.
“if there's a trend, inherently you have many companies going after the same trend. You're going to have new entrants... And so, you want to avoid competition on both fronts.”
On first principles, nuclear should be among the cheapest energy sources: nuclear fuel is so energy-dense that a pellet equals roughly a ton of coal, so reactors need far less material, mining, and space, making costs potentially an order of magnitude lower than fossil fuels.
“a pellet of nuclear uh fuel and the amount of coal that that's equivalent to is like a ton of coal... It's just an order magnitude different.”
Nuclear costs have risen because the US built less nuclear starting in the 1970s; large reactors now take 10 to 15 years and can exceed budgets by double, making utilities unable to commit, and nuclear is now more expensive than fossil fuels.
“we did less and less and less nuclear, probably starting in the '70s. And you see the cost of nuclear going up as we do less of it.”
Outsourcing manufacturing abroad gives up the capability to scale rapidly and to develop new manufacturing techniques, so the US risks losing economic position and supply chain control at the front of the value chain.
“I think just purely outsourcing stuff overseas just gives up capability, especially as it gets like new manufacturing techniques”
Enrichment is sold as a tolling service: utilities own the uranium and pay per SWU for each upgrade step, and enrichment is one of the largest cost segments in fuel production (about a third for LEU and more for HALEU).
“it's really a tolling business where you buy they own license to the material and it's book transfers all the way through”
Uranium ore supply is not a constraint: the world has hundreds of years of supply in the ground, the US has active mining, Canada has even larger higher-grade deposits, and Australia and Kazakhstan (from which the US imports) have large deposits.
“certainly hundreds of years of supply in the ground. US has supply. We have active mining in the US. Canada has even more supply, higher grade ores.”
There are three nuclear fuel cliffs: HALEU supply for advanced reactors (no reliable supply in the next few years), the January 1, 2028 US ban on Russian uranium imports, which removes roughly 25% of US enrichment imports since only Europe and Russia enrich at scale, and eventual depletion of the US Navy's enriched uranium stockpile.
“The second nuclear cliff... is 2028 when the US's ban on Russian uranium imports goes into full effect, January 1st, 2028.”
Latent demand for new enrichment capacity only became visible in the last 24 months because of AI data center growth; through the 2010s the US built no new nuclear capacity and little new HALEU demand existed, so the company is still betting on the market growing.
“any large latent demand I think is in the last 24 months linked to the thought that there's going to be a lot more data centers doing AI inference or training jobs.”
Energy supply constraints are binding for AI data centers, which could consume the entire US grid by 2030 at current growth rates, so the US must build capacity proactively rather than assume markets will bring energy online when needed.
“looks like data centers can consume if they were allowed to the entire grid by 2030 at this growth rate”
US electricity generation grew for decades, then stayed essentially flat from the 1990s until today, while China grew rapidly and is now on track to have about triple the US total energy production, so the US must add energy supply to stay economically relevant.
“we grew our grid for a long time, and then starting around the '90s, essentially no growth until today. China was growing a long time”
The US has no commercial-scale enrichment capacity and no competitive R&D capacity relative to Russia and Europe; enrichment is the bottleneck that drove General Matter's product decision.
“There's some R&D capability, but nothing that's commercial scale or commercially competitive with Russia and Europe.”
Nuclear is the right base-load power source because base load must be always on and reliable enough to design industrial processes around, and solar with storage is not typically economical at the storage levels needed to serve as base load, so solar is mainly peak shaving.
“solar with enough storage could be base load, but typically it's not anywhere near that amount of storage. Usually solar is more intermittent, more for peak shaving.”
A venture investor should concentrate capital into the fewest companies possible, because diluting returns through indexing across many deals dilutes the few outcomes that actually matter.
“you should just concentrate into the fewest number of companies possible. And don't dilute your average returns by indexing.”
Industries that stagnate are typically cost-plus industries with little incentive to reduce costs, which leads to a fixed market size, oligopoly status, and a pricing equilibrium that extracts margin rather than growing the market, so such industries are prime targets for disruption (space launch, defense, infrastructure).
“a cost-plus industry, where there's very little incentive for progress, not much incentive to to bring the cost structure down”
Enrichment in the US has seen little change since the 1990s, and building new facilities is extremely hard because they are million-square-foot, multibillion-dollar industrial projects; standing up a large industrial facility is underappreciated and is the real hard part, more than the initial technology.
“I think it's extremely underappreciated how much goes into standing up a huge industrial facility. And I think that's actually the hard part.”
Deep subcontracting, with some nuclear companies having around 900 subcontractors and 30 layers in systems like the space shuttle, locks interfaces in place and calcifies architecture, while bringing design and manufacturing in-house lets engineers trade off interface requirements across the whole system.
“every one of those interfaces that crosses another company is typically a fixed interface that's not going to move very quickly... you end up with a really um calcified architecture of your system.”
Co-locating first-of-a-kind manufacturing with engineering matters because designers can adjust designs to manufacturing feasibility, enabling choices like using a laser cutter instead of a six-axis CNC to achieve roughly 10x throughput at a tenth of the cost.
“you at least have your first of a kind manufacturing co-located with engineering under the same roof”
At SpaceX, the operating method was to define the goal, pick a good-enough solution quickly, test extensively, and iterate, rather than deliberating for months; the aim is to get to roughly 90 to 99.9% and become operational, since a company that never launches cannot succeed.
“Let's not deliberate for months and months and months over things. Let's just decide and move forward.”
Data centers can go behind the meter by building their own on-site power, a fenced island that never touches the grid, and this can also benefit host communities: if a data center adds about 100 MW of power to a gigawatt-scale campus, it can lower local utility prices, so 'bring your own energy' is a no-brainer for hyperscalers seeking to be welcomed.
“Imagine a fence around a few hundred acres, you got a data center and you've got its nuclear power plant there... It does not even touch the grid.”
Up rounds are often under-priced because investors anchor on the past round's price; the steeper the up round, the greater the undervaluation, so a 2x up round may really warrant 4x, since returns are paid on exit price rather than proximity to the last round.
“the steeper the up round the greater the undervaluation.”
Founders are not often passionate about the domain they are addressing: they need to be in love with the idea, because starting a company is irrational compared with easier ways to make money, but investors who fall in love with an idea make compromises on the team and may put good money after bad.
“From the investor seat, I don't think you want to be in love with an idea... But I think on the company side you have to be in love with the idea.”
Energy demand from AI data centers arrived faster than the US can build permitted physical energy infrastructure, so the energy system is caught off guard: permitting and construction timelines for physical assets take far longer than market demand can shift.
“physical world stuff has a timeline with it and then in the US it can be very hard to get stuff permitted to go do things.”
Reactor designs fall into three size buckets with different markets: gigawatt-scale reactors (e.g., Westinghouse AP1000) for the grid, roughly 100 to 300 MW small modular reactors (SMRs) likely to find a niche behind the meter at data centers over 5 to 10 years, and roughly 1 MW microreactors for remote communities and military installations that displace diesel generators.
“micro reactors... Think yeah, 1 megawatt scale, not gigawatt, megawatt. And then you have the middle, which is 100 to 300 total megawatts, SMR.”
The biggest risk to advanced reactors is that they lack fuel to operate, a showstopper problem, ahead of cost; the physics is considered likely to work, so the fuel supply is the constraint that the company is trying to eliminate.
“I'd say number one is they don't have fuel to operate. That's definitely a showstopper.”
Operating a company is far more demanding than investing: the feedback loop is fast and tasks cannot be deferred, whereas investing allows skipping deals with the cost of a bad outcome only visible years later, though a great investor sees far fewer companies that matter than many meetings would imply.
“when you're operating, you can see the feedback cycle very quickly. As an investor... you might not realize that you're not doing well until 5 years from now”
General Matter treats every problem as an engineering problem to be solved quickly, avoiding multi-year R&D and deliberation, and it builds its own in-house EPC (engineering, procurement, construction) capability instead of hiring a general contractor, following the Tesla playbook to retain schedule and cost control.
“Don't hire a GC and outsource everything. You lose total control. The schedule could be it's not in your hands anymore.”
The most persuasive way to build public support for nuclear is to make it cheaper, not to argue about safety data: if a utility bill were cut in half, people would find a nearby plant compelling even if it were tens of miles away.
“the way to make nuclear really compelling to people is actually to check the final box, which is cost.”
Peter Thiel's 'bits versus atoms' observation: the world made enormous progress in bits (software and phones) but almost none in atoms, with little physical change in 50 years, which motivated a search for hardware companies.
“we'd made all this progress in the world of bits, but not in the world of atoms. And you could be on your cell phone and it was interesting and then you look around and nothing's changed in 50 years.”
After a full year of searching in 2023, Nolan found no existing company doing enrichment to debottleneck nuclear, which led him to conclude that enrichment had to be built by a new private company rather than an incumbent or government.
“looked at trying to find a company in the enrichment step to debottleneck nuclear ... and could not find anything for an entire year and then finally decided if this is going to exist, it's got to be a new company.”
The team you build is the company you build, so General Matter's early hires determine its culture, and the interview's final screen tests whether a candidate cares about the problem and will work late nights, weekends, and accept lower pay for the mission of delivering before the end of the decade.
“the early team you build is the company you build”
Founders Fund's firm-level selection is for investors who do not want to be entrepreneurs, and founding a company should happen only in extreme exceptions; Nolan started General Matter because nuclear enrichment was an important problem no one was solving and his background aligned with it.
“Founders Fund recruits for people that want to be investors that don't want to be entrepreneurs.”
Early in his career, an investor should take many meetings and rely on intuition, which is usually correct; later, excessive analysis can lead them astray, so the goal is to get better at asking questions that test intuition rather than relying on analysis for every deal.
“the intuition was like all that you had to go on and it was and I think it's usually correct. ... That might actually lead you astray, I think.”
Low valuation is usually a warning sign in true venture investing: a deal priced below market often means the company cannot raise capital at market prices, so 'value deals' in venture are a dangerous strategy except for very small funds or unusual cases.
“either it says something about the company's ability to raise capital that they're unable to at market prices.”
The hardest part of the nuclear fuel problem is not licensing but getting fuel: advanced reactor developers report that the NRC is not the main obstacle, and fuel supply from Russia (the only country supplying the needed fuel) was a constraint even before the Russia ban.
“Everyone thinks the NRC is impossible to get through, but no, it's not that. They told us it's actually we cannot get the fuel.”
Founders pursuing under-appreciated ideas can be identified in meetings: they are not giving superficial answers but take the investor down the rabbit hole, explaining why everyone else thinks the problem is impossible or is attacking it wrongly.
“They're like, 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.”
General Matter is structured as an engineering-driven company rather than a science project: nuclear engineers are a single-digit percentage of the team, so the company was located in Southern California near hardware and aerospace engineers instead of where nuclear engineers are scattered.
“the number of the percent of the team that's nuclear engineers is single digits... you go to where all the other engineers are, specifically hardware, aerospace, and that's Southern California.”
General Matter treats its work as engineering rather than science: no new physics and no multi-year uncertain R&D; everything from cost, performance, cost of capital, and schedule is to be engineered.
“We're not doing a science project. No new physics. No going down a multi-year R&D path that's uncertain. ... This is an engineering problem.”
Vertical integration of engineering is mandatory for hardware companies because teams sitting side by side can trade interface requirements (for example, one team giving up margin in exchange for another's breathing room) that cannot be negotiated across separate companies.
“you can have those conversations much more easily when two people are sitting side by side at a desk than when you have two separate companies”
Fuel may be half the total cost of advanced reactors, and long-term enrichment costs may be roughly half of the fuel cost, so the enrichment step is central to whether advanced reactors can be economically viable.
“It's the fuel can be half of their total cost and I think long-term enrichment will be half of the fuel cost.”
General Matter chose to focus first on HALEU enrichment because advanced reactor developers urgently need it and incumbents are unlikely to serve that small emergent market on the same timeline, then move to LEU production for the 94 existing US reactors, a roughly $2 to $2.5 billion US market.
“the most urgent cliff that we need to address as soon as possible is HALEU for advanced reactors... phase two is going to LEU production”
Airbnb's mid-stage inflection came when the company took a hard stance after a damaged-home controversy, reimbursing hosts and professionalizing the platform, which signaled it would go mainstream.
“And then the company took a hard stance on that and said, We will reimburse the hosts and we are professionalizing this. And I think that was like maybe the moment that you could tell it was going to go mainstream.”
Data centers used to locate at stranded power sources such as West Texas wind or South and North Dakota flare gas, but those stranded assets are now used up, so new capacity must be built, and data centers are backed up on natural gas turbines while nuclear takes years to install.
“those stranded assets are gone and now it's time to build new capacity and data centers are completely backed up on getting natural gas turbines”
General Matter's business is robust even if nuclear growth stalls, because its LEU enrichment serves an existing roughly $2 billion-plus US market and a similar allied market, so time is on its side for when advanced reactor demand grows.
“worst case, there's an existing market and we can start a business there and time is then on our side for when and if nuclear grows”
Venture investing does not require much labor because only a few companies per vintage matter; the job is one of ideas and thinking rather than labor, so an investor can focus deeply on a handful of companies rather than meeting everyone.
“all that it ultimately matters is a few companies per vintage and you just have to be into those. ... It's really a job of ideas and thinking.”
Labor responds to energy shortages within a few years, as shown by the electrician shortage where wages exceed those of master's degree holders, but infrastructure build-out takes longer, so infrastructure and permitting, not labor, are the binding bottlenecks.
“I think labor responds. It probably takes a few years, but it's faster than building infrastructure. And so, infrastructure is probably the bottleneck.”
The strategy is to sell the highest-cost segment of the workflow to a small but emerging customer segment that the company can own, driving all value through the narrowest possible choke point, which is how a great business is built.
“you sort of have found the highest cost segment of the workflow sold to the highest percentage of the cost um new customer that's emerging that's small that you can go own.”
Founders Fund's early deals show that large opportunities often start as small markets: the firm's thesis is to start with a small market you can own and grow from there, rather than worrying about total addressable market, as seen in Crusoe's flare-gas business and Radiant's microreactors for stranded demand.
“start with a really small market and grow into a bigger market. Like don't don't worry about your TAM, worry about owning that market and then grow from there.”
Peter Thiel's influence on investing was to avoid herd thinking and trends, and to take an orthogonal view of companies by building many layers of abstraction about why an investment is being seen at all, rather than simply building a spreadsheet.
“Instead of just doing a spreadsheet and trying to analyze this investment, why don't we think about like why are we even seeing this investment?”
The US fails to build infrastructure because energy is not just abundant but cheap is what matters; higher-cost energy alone will not bring back industries that have left the US.
“just making more high-cost energy will also won't bring back certain industries. It won't cause us to start doing things we used to do.”
General Matter's North Star metric is dollars per kilogram of separative work units (SWU), because enrichment is sold as a tolling service priced per SWU, and for advanced reactors that input cost matters a great deal, while for large LEU light-water reactors fuel is a small share of cost.
“our North Star metric is dollars per kilo SWU.”
Hardware and physical-world companies (biotech, computer chips, satellites, space launch, transportation, infrastructure, and nearly anything not digital) were a huge, widely ignored investment opportunity in the early 2010s, while the world focused on software.
“my thesis was hey, I think there's a huge set of you know, physical world companies, hardware companies that could be really valuable”
Airbnb looked like a niche backpacking and air-mattress business, but analysis of demographics, market share by market, and marketing spend showed a shifting, older customer base and dominant share gains in every market, signaling that it would win.
“the demographics were shifting to slightly older crowd, not just backpackers out of college... you could see like all these markets that they were in, they were just taking share”
Unlike the advanced reactor market, the company's enrichment capacity would apply to natural uranium refined to any level, so enrichment services can serve LEU at 3 to 5% and HALEU at about 20% from the same underlying capability, measured in separative work units.
“the fundamental unit of enrichment capability is it's just the ability to do this refining. It's measured as something called separative work units”
Small modular reactors could be cheaper than large gigawatt construction projects for the grid, but whether advanced reactors can deliver energy at a low enough cost is still an open question.
“SMRs I think ultimately have the plan of being cheaper than the gigawatt scale construction projects for the grid. And the question's going to be is their energy production cost low enough?”
NASA's COTS program worked because agency staff who had worked there for decades were true believers in space who welcomed credible companies that cared as much about safety and performance, producing very collaborative relationships.
“in industries where you have true believers who are in the government agencies and who have been doing this for a long time... there's a lot of openness and excitement to working together”
Nuclear demand has repeatedly failed to materialize in past renaissances, so industry players argue they will build capacity only when need is proven, creating a chicken-and-egg problem for enrichment.
“They've been through nuclear renaissances before that didn't pan out, and I think their point of view would be we'll build it when we know that there's a need.”
Nolan spent a decade investing, then left because the company existed for the mission rather than for returns; he was willing to give up quality of life, and is screening hires for commitment to the mission over salary or predictability.
“I'm screening for is does this person not just have the skills... but do they have the attitude of you know, they care about this problem.”
The biggest hurdle General Matter faced when Peter Thiel pushed back was the meta question of why the US has built no new nuclear in a long time and whether regulation is designed to block it, forcing the team to develop answers about SMR and HALEU demand dependence.
“Is nuclear just regulated to death? Is the regulation a thing that's meant to stop it”
A personal framework for choosing work is to do something useful that makes a real contribution, asking what important problem will not get solved otherwise and whether one's talents can help solve it.
“My framework has always been just do something that's useful. Do something that you feel like you're making a real contribution”
Nolan has no inherent love for uranium enrichment; he was always excited about nuclear energy and space, two technologies he thought America was supposed to have, so enrichment became the bottleneck he chose to attack.
“Never had a specific interest in uranium enrichment, but then through the course of investing in Founders Fund”
Founders Fund's Spotify investment (led by Sean Parker) was well reasoned because it built on deep music-industry knowledge from the Napster era and years of searching for the right licensing and market-entry formula.
“the internal memo or thread on the Spotify investment was just so well reasoned. And it was because of this history of like understanding music and doing the Napster thing”
Peter Thiel's board service, support for the transition, and challenge to the idea were the most meaningful support Nolan received, and the hardest challenge was the meta question of why nuclear had not been built, whether regulation was designed to block it, and whether the company depended on every SMR succeeding.
“Obviously, Peter like tried to beat up the idea and make make sure it's good”
Founders Fund's late hardware investments (Crusoe Energy using stranded flare gas, Radiant using stranded demand for microreactors in remote Alaskan villages and army bases paying high diesel prices) led Nolan to the nuclear bottleneck.
“invested in Crusoe Energy and understood the whole stranded supply of flare gas ... invested in company called Radiant which was the inverse stranded demand.”
Nolan believed SpaceX would ultimately own the entire space launch industry even when it had only about 30 people, which made joining it an obvious choice right after college.
“there's this new company SpaceX that is going to ultimately own the entire space launch industry, which I believed even when it was 30 people.”
The early conversations about General Matter with Thiel took place in 2023, before the AI data-center boom, and since that boom the need for the company has become extremely obvious.
“a lot of these conversations that I had with Peter were 2023 before the AI data center boom.”
Founders Fund was founded in 2005 on a founder-friendly premise that founders should keep control of their companies, because the most successful companies were founder-run to the end; by the early 2010s its focus had become contrarian investing in what important companies no one was funding.
“The premise was let's give founders back control of their companies and unilaterally support them in building that. That was 2005 genesis of Founders Fund”
Procedural remark about the host's question: 'Scott, this has been so much fun, thank you for your time.'
“Scott, this has been so much fun. Thank you so much for your time.”