Newsletter 026

Good morning — it’s your Editor, Ryan Duffy. This week, we’re talking about two American giants who aren’t quite what you’d call overnight success stories. No, these two spent decades, and decades, honing their craft. But they’re proof that good things come to those willing to put their head down for a long time, ignore the noise, and emerge with something that actually works.

The first is preparing to take his final bow, having spent six decades turning his company from a dying textile mill into a $1 trillion infrastructure empire, taking on the unglamorous but essential — and valuable! — work that few others wanted to touch.

The other spent nearly three decades building Blue Origin in near-total silence and exhausting vast sums of personal capital, before finally launching two spacecraft to Mars last week and, more importantly, landing New Glenn’s 321-foot booster on the rocket’s second-ever flight (thereby ushering the U.S. into a new era of two affordable, reusable heavy-lift rocket providers).

Today’s Per Aspera newsletter is a good reminder: "Never bet against America".

IN THIS WEEK’S EDITION:
🔭 What we can learn from the Oracle of Omaha
🌊 Project Prometheus & Physical AI
🏅 Friends in high, hard places

Were you forwarded this email? Subscribe to Per Aspera here.

Warren Buffett announced earlier this month that he’s “going quiet.” That means no more marathon Q&A sessions and no traditional annual report letters at Berkshire’s legendary shareholder meetings, though he will still pen his annual Thanksgiving messages to shareholders. After 60 years of dispensing wisdom from Omaha, the Oracle will officially hand off his CEO duties to Greg Abel on January 1, 2026, remaining as chairman.

While the financial press obsesses over what this means for Berkshire’s stock, we see something more important: a masterclass, and working model, for financing American infrastructure. Buffett proved you could compound billions by owning the “boring” stuff: railroads, utilities, the physical backbone everyone needs and nobody wants to finance. And he was no charity case, averaging ~20% compound annual returns for six decades, nearly double the S&P 500’s pace. If America wants to rebuild, it pays to study his playbook.​

Warren Buffett bought a railroad the same year Instagram launched.

It was 2010, and while Silicon Valley chased photo-sharing apps, the Oracle of Omaha was spending $44B to buy BNSF Railway, and with it, 32,500 miles of track, 40,000 employees, and nearly 7,000 locomotives that belched smoke across the Great Plains. He called it an “all-in wager on the economic future of the United States.” Somewhere in Palo Alto, a VC surely recoiled: capital-intensive, union-heavy, regulated to hell, with margins measured in basis points rather than multipliers. Why would anyone buy something so…physical?

So who was right? Both Buffett and this fictional VC. Instagram connected 1B people and changed the world. And Buffett’s railroad hauls the actual world. Today, it throws off $5B+ in profit annually and carries nearly one out of every six tons of American freight: grain from 1,500 elevators, containers from Pacific ports, the physical goods that keep civilization running.

The lesson wasn't that tech is bad and trains are good (though trains are definitely great). It's that somebody has to own and operate the boring, capital-intensive infrastructure that makes everything else possible.

Warren Buffett built his empire through eras defined by offshoring, financial engineering, and software maximalism…and chose, again and again, to plow money into railroads, power lines, factories, and housing. Let’s look at what he’s built and/or stewarded: Berkshire Hathaway Energy operates 10-13 GW of wind power, among the largest U.S. portfolios. Clayton Homes built more than 60,000 homes last year, representing ~5% of new U.S. housing starts. Precision Castparts forges jet engine components. Lubrizol mixes industrial chemicals. Shaw Industries runs carpet mills.

With the U.S. embarking on a generational industrial resurgence, six lessons from Omaha matter more than ever:

001 // Treat capital as a strategic weapon, not a quarterly performance drug: Buffett turned Berkshire into permanent capital, buying outright, removing pressure to “make the quarter,” and empowering managers to invest for decades. Railroads laid track. Utilities buried cable and built wind farms. Manufacturers retooled plants. We need more of this and less flip-and-strip. Semiconductor fabs, reactors, refineries, and shipyards don’t always fit inside fund (or election) cycles. If your capital structure assumes a quick exit, you will underbuild, under-maintain, and under-invest.

002 // Own boring, regulated, indispensable things, and reinvest ruthlessly: Buffett made a fortune on businesses most investors considered too dull to bother with. They offered “reasonable” returns, not a ten-bagger or 100x, and they shared three traits: essential service, local monopoly or deep moat, and a huge capacity to absorb reinvested cash. He let those businesses retain and reinvest earnings, compounding capacity and resilience, often far beyond what Wall Street would allow.

003 // Decentralize execution, centralize ethos: Omaha sets the philosophy and the capital allocation, then got the hell out of the way. BNSF is run by railroaders, the utilities by power engineers, and Clayton by people who know how to build and finance homes for working families. Buffett didn’t try to “synergize” everything six ways to Sunday. He picked people he trusted, gave them a fortress balance sheet, and expected them to behave like stewards.

004 // Accept that compounding beats disruption: Buffett, of course, opted for compounding advantage inside existing platforms. The railroad doesn’t need to become a hyperloop. It needed more sidings, better signaling, and incremental annual efficiency improvements that look boring until you wake up 20 years later with a fortress. Today we’re feeling the industrial hype: AI factories, fusion reactors, giga-this, mega-that. This is real and necessary. But much of what lifts national capability will look like dull compounding: 3-5% annual efficiency improvements in plants, relentless uptime, steady reduction in defect rates, reducing project overruns. The builders that win will be the ones that show up every day to make their systems 1% better.

005 // Align profit with resilience, not extraction: Buffett was unapologetically capitalist, and his model usually created resilience as a byproduct of profit: better rail networks, more robust grids, preserved manufacturing know-how, affordable housing capacity, dependable insurance. He avoided stripping assets and leaving hollow shells. As we pour money into chips, energy, logistics, and defense, we can build A) brittle systems optimized for IRR screenshots and maximized leverage, or B) slightly less “optimized” systems, with redundant capacity, domestic supply, skilled workforces, and thicker margins of safety.

006 // Stay inside your circle of competence, and expand with humility: The Oracle of Omaha largely avoided tech because he didn’t understand its dynamics well enough to underwrite it responsibly. This humility lost him a couple generational opportunities, to be sure, but it also kept him from doing a lot of dumb deals. When he did tiptoe into tech, it was only once Apple’s cash-generating machine looked more like a consumer staple than a speculative rocket launch company.

As America embarks on the greatest industrial mobilization campaign since World War II, Buffett’s style offers us as good a handbook as any. Where we go from here is simple to state and hard to do: Build long-duration capital pools that think in generations. Point them at the hard, indispensable systems that everything else depends on. Let operators lead instead of financiers. Reinvest relentlessly in capacity, competence, and compounding.

The goal isn’t to copy this model down to every detail. But if even a fraction of U.S. capital follows these principles — patient, physical, permanent — it will help the nation build systems that last, adapt, and sustain.

Jeff Bezos is back in operator mode for the first time since Amazon, with a new startup, a hell of a large seed round, and a thesis: LLMs can’t bend metal.

Bezos’s ~new thing~ is called Project Prometheus, which has raised $6.2B in funding, much from his own pocket. His co-CEO is Vik Bajaj, the acclaimed physicist-chemist who ran Google X’s life sciences division and co-founded Verily. Their first heist: nearly 100 AI researchers poached from OpenAI, DeepMind, and Meta. Their mandate: AI for the physical economy, with early work targeting engineering and manufacturing across compute, cars, and space systems.

Today’s frontier models are linguistic savants who’ve never touched grass. They excel at writing and reasoning about documented problems, but struggle with physical causation: how materials behave under stress, or how a single parameter change cascades through a manufacturing system.

The Rise of Physical AI: A tidal wave of well-capitalized startups has emerged to attack this gap. Periodic Labs, for example, raised a $300M (!) Series A to build “AI scientists” aimed at accelerating discoveries in physics, chemistry and materials science. Robotics companies, meanwhile, are racing to combine world models and simulation to train systems that can operate effectively in our physical world.

Godfathers, godmothers, & a new frontier: Yann LeCun, one of the “godfathers” of deep learning, is reportedly leaving his role as Meta’s chief AI scientist and will launch his own world model startup. Fei-Fei Li, the “godmother” of modern AI, launched World Labs last year with $230M in funding. Last week, she published a manifesto on why spatial intelligence is AI’s next frontier:

A vibe shift? LLMs work brilliantly at their job, but they hit a wall when reasoning about physics. Some of the smartest minds in AI increasingly see this not as a scaling problem, but a local maximum. You can’t text-predict your way into understanding turbulence, nor interpolate between descriptions to discover new materials. The only way through, then, is with systems that learn how the world actually works, not just how it’s described in text or abstracted in software…

The bottom line (literally): The $16T digital economy (all of ecommerce, SaaS, cloud) represents ~15% of global GDP. Even though the physical economy is growing slower, it is much larger. Manufacturing alone is $16T globally. Add construction ($13T), agriculture ($4T), transportation, aerospace, energy, and you’re looking at $40-$50T that depends on understanding and manipulating atoms.

For those fleeing the transformer monastery — all of you scientists, financiers, and researchers willing to trade the comfort of pure software for the resistance of real matter — your decade has arrived. Progress demands hard pursuits in the physical world, not just our digital realms. Welcome to the Renaissance, where the hard problems are the only ones worth solving.

Toyota Motor Corp has opened its $13.9B battery plant in Liberty, NC, with the 30 GWh facility supporting the carmaker’s hybrid-heavy lineup and future EV programs. The plant is designed to supply hundreds of thousands of hybrid vehicles annually while also laying the domestic foundation for Toyota’s upcoming three-row BEV and future full-electric programs.

This is Toyota’s first U.S. cell-manufacturing footprint, giving the company local control over pack assembly, thermal-management integration, and compliance with tightening U.S. content rules.

More strategically, it represents a structural shift:

Last week: We noted that nuclear startup Valar Atomics closed their $130M Series A with participation from Anduril founder Palmer Luckey and Palantir CTO Shyam Sankar (plus others).

Yesterday: In partnership with Los Alamos National Laboratory, Valar Atomics’ NOVA Core — a TRISO-fueled, graphite-moderated assembly — achieved criticality not once, not twice, but three times.

This makes them first privately funded startup to split the atom (!!!). It also marks the first major physics-validation step within the Department of Energy’s Advanced Reactor Pilot Program, which aims to demonstrate new reactors on aggressive timelines.

Though not a commercial energy output, this is a real milestone for the builders as well as state/startup partnerships.

Castelion has chosen Sandoval County, New Mexico for Project Ranger, its new 1,000-acre, $100M+ hypersonic rocket motor manufacturing campus expected to create 300 long-term jobs and generate $650M in economic output over the next decade. The facility will help restore America’s ability to produce advanced hypersonic missiles domestically, countering overseas rivals and ensuring the U.S. stays competitive in a key arena of future military technology.

CesiumAstro just inked an MOU with TCI Aircraft Interiors — a global provider of aircraft cabin systems and in-flight connectivity solutions. At the heart of this collaboration is the evaluation of CesiumAstro’s Skylark antenna system, which leverages the company’s modular phased-array architecture to deliver simultaneous, electronically steered multi-beam connectivity — departing significantly from legacy mechanically steered antennas. 🙌 to the future of in-flight connectivity!

Lastly, our beloved Dan Goldin’s speech on “Where Are My Asteroid Mines” caught attention on the X timeline, and we were highlighted on TBPN last Friday — thanks guys!

Meta’s 2Africa and Google’s Blue-Raman subsea cables have hit snags in the southern Red Sea due to security risks // GM instructs thousands of suppliers to remove Chinese parts from North American supply chains // Why Solarpunk is already happening in Africa: decentralized energy, decreasing hardware costs, and innovation in logistics/finance (credit scoring, distribution, mobile money) // World’s oldest RNA extracted from Ice Age woolly mammoth // France announces nearly $5B in new military space funding // Forecasters at US National Hurricane Center increasingly leaning on Google/DeepMind’s new prediction model.