$250 Billion in Assets. Here's What They're Buying Now. First Eagle Global Value

ELI5 / TLDR

Two fund managers from First Eagle — a 160-year-old firm running $250 billion — explain how they pick stocks. The short version: buy good businesses that are genuinely hard to copy, only when they’re on sale by at least 30%, hold them for a decade, spread bets across 100-plus names, and keep some gold around as a fire extinguisher. They walk through a few things they’re buying right now: a French engineering-software company, the Mexican arm of Walmart, and some copper. Along the way they argue the US market looks like the dot-com era again, and that the AI spending boom may not pay off the way fiber and railroads did.

The Full Story

The firm, briefly

First Eagle started in 1864 as a private bank in Germany, run by two families for six or seven generations. It survived German hyperinflation (it happened to own a brewery and some gold), fled the Nazis in the 1930s, and rebuilt in the US. George Soros started his career there in the 1960s. Today it manages about $250 billion and runs the First Eagle Global Fund, which has a track record going back to 1979. The guests, Christian Heck and Julien Albertini, describe it as one of the last “temples of global fundamental long-term value investing.”

The North Star: don’t lose money

Everything flows from one phrase the team repeats: resilient wealth creation.

“We really concern ourselves about avoiding the permanent impairment of capital for our clients. Think about it as participating in the march of man but avoiding the potholes along the way.”

In practice that means mostly equities, held a long time (the portfolio turns over 10-15% a year, so the average holding period is 7-10 years), plus a slug of gold as insurance, plus whatever cash happens to be left over when they can’t find anything cheap.

What counts as a “good business”

Value investing splits into two camps. The Benjamin Graham camp screens the whole world for the statistically cheapest stocks — cigar butts with one puff left. The Warren Buffett camp starts by hunting for good businesses and then waits for a fair price. First Eagle sits closer to Buffett.

Their definition of “good” hangs on one word: scarcity — owning something hard to replicate. It comes in two flavors.

Tangible scarcity is a physical asset you simply can’t recreate. They own Grupo México, which owns the main rail network of Mexico. As Heck puts it: “Good luck replicating that. You’re not going to lay more track.”

Intangible scarcity is a brand, a customer lock-in, or a dominated niche. Their example is Colgate, which holds 40-45% of the global toothpaste market. One in two people on Earth brushes with their product. You don’t get there by accident.

Once a business clears the scarcity bar, they want a clean balance sheet (no leverage — leverage is how you hit a pothole) and owners worth standing beside. That’s why the portfolio is stuffed with family-controlled and founder-led companies: a family that owns half the shares tends to run the business for the next decade, not the next quarter.

Only then do they ask what a rational buyer would pay for the whole thing — and they demand a discount of 30% or more before buying. That gap is the margin of safety.

Diversification as humility

This is the part that cuts against fashionable “high conviction, ten best ideas” investing. First Eagle owns 100-120 names. Their reasoning is disarmingly honest:

“Ex ante, I don’t know which one turns out to be my best idea. If I knew, I could run a 10 or 15 stock portfolio. But the real world is difficult, is messy.”

Albertini reaches for behavioral research to back this up. He cites Tetlock’s Superforecasting and a CIA manual on analysis: as you pile on more information, the thing that rises isn’t accuracy — it’s confidence. Everyone thinks they’re above average, which is statistically impossible. So diversification, for them, is a confession of fallibility rather than a hedge against laziness. Positions get planted small (around 0.5%) and grow into 2.5-3% only if they get cheaper or better understood. They deliberately resist averaging down hard, because a 1% bet that becomes a 4% loss through repeated buying is how value investors blow themselves up.

Three things they’re buying now

Dassault Systèmes (Paris) — the software engineers use to design planes, cars, and iPhones. Boeing and Airbus both build on it; so do more than ten of the top fifteen carmakers. It’s a near-duopoly with Siemens, and it’s extraordinarily sticky: a competitor once revealed that of its 500 biggest customers 25 years ago, all but six are still customers. Three reasons nobody switches — risk aversion (if your design tool fails, the plane falls out of the sky), retraining cost (imagine being told to abandon Excel), and a network effect (students learn it because employers use it). The whole software sector sold off on AI fears — the “SaaSocalypse” — and Heck argues mission-critical systems-of-record like this got thrown out with the bathwater. It now trades under 4x sales while industry deals happen at 8-10x.

Walmart de México (ticker WALMEX) — the trick here is geographic arbitrage. US Walmart trades at 40x earnings and barely opens new stores. Its Mexican subsidiary — same model, same people, 71-72% owned by Walmart itself — trades at 15x because it’s listed in Mexico. And Mexico is where US Walmart was 20-25 years ago: less than half the retail market is “modern,” Walmart has 350 supercenters versus 3,500 in the US, and same-store sales are still growing mid-to-high single digits. Same business, much earlier, big discount.

Copper, via Grupo México — their pick-and-shovel play on the AI build-out. Every data center needs enormous amounts of copper, and the cheapest, highest-quality copper sits in northern Chile’s Atacama Desert, where Grupo México owns a giant low-cost open-pit mine. As Albertini frames it, copper is “a toll road” on the trillions being spent — you collect regardless of which AI company wins.

The gold logic

About 10-15% of the portfolio is gold, held as physical bullion (in an HSBC vault in New York), not derivatives. They’re not gold bugs and have no price target. The case is structural: developed-world money supply grows 7-8% a year while gold supply grows about 1%, so over time gold should hold its purchasing power and drift up by that gap. Unlike a credit-default-swap hedge, where you bleed a premium every year, gold is a hedge you get paid to hold. It also tends to shine exactly when equities don’t — the 1970s and 2000s were lost decades for stocks and great ones for gold. They size it between 5% (below which it’s too small to matter) and 15% (above which they’d be making a directional bet). Gold’s recent run has actually forced them to keep selling it to stay inside the band.

The market read: back to the periodic table

The hosts frame today as a K-shaped market echoing 1999-2000. The guests agree, and SpaceX’s rumored $1.8 trillion IPO gets nominated as the possible “AOL Time Warner moment” — one team member called reading its S-1 “like reading a science fiction book.” Their broader thesis is cyclical: when financial-economy darlings (Nifty Fifty, dot-coms, mega-cap internet) get overpriced, the market eventually swings “back to the periodic table” — real assets, commodities, physical structures — which favors international and emerging markets. They have about a third of the portfolio in real assets versus 5-10% for the S&P 500.

On the AI capex boom, Heck makes the sharpest point of the episode. The hyperscalers historically spent ~20% of operating cash flow on capex while growing fast — a beautiful model. Now they’re spending nearly 100%. “You can only do that once.” That’s why Google is raising $80 billion externally (with Berkshire taking $10 billion). And unlike railroads or fiber, which became lasting consumer surplus, GPUs have short useful lives — so this build-out might destroy capital rather than gift it to consumers.

How far out they underwrite

They read the last 15-20 annual reports and rebuild 20 years of financials by hand — not for the model, but to force themselves through the footnotes and see how a business behaved across past cycles. They explicitly don’t try to forecast the next quarter. Their edge is time arbitrage: most investors fight over the next 3-12 months, a crowded and probably pointless game. First Eagle buys during soft patches (Chinese internet under regulatory fire, Hong Kong real estate mid-COVID) when they can’t say what next quarter looks like but are confident the business will still matter in five years. Patience is scarce; a loyal, long-term client base lets them supply it.

Key Takeaways

Resilient wealth creation = avoid permanent loss of capital first, chase returns second. Everything else is downstream of this.
Scarcity is the core filter — own things that are hard to replicate, either tangible (rail networks, low-cost mines, prime real estate) or intangible (dominant brands, customer lock-in, niche monopolies).
Margin of safety of 30%+ before deploying capital; value the whole enterprise as a rational acquirer would.
Diversification as humility, not hedging — 100-120 names because you can’t know your best idea in advance; piling information raises confidence, not accuracy.
Position discipline — seed at ~0.5%, let winners grow to ~3%, rarely buy in above 1% at cost to avoid the averaging-down death spiral.
Gold as a paid hedge — money supply grows 7-8%/yr vs gold’s 1%, so it holds purchasing power; held as physical bullion, sized 5-15%, no price target.
Geographic arbitrage — Walmart de México is US Walmart 25 years ago at 15x earnings vs 40x, same owners and model.
AI capex math — hyperscalers went from 20% to ~100% of operating cash flow on capex; “you can only do that once,” forcing external fundraising.
GPUs depreciate fast unlike railroads/fiber, so the AI build-out may destroy capital rather than become consumer surplus; copper is the toll-road play.
Time arbitrage — refuse to compete on next-quarter forecasts; buy out-of-favor quality, underwrite 5+ years out, hold 7-10.
“Back to the periodic table” — real assets/commodities/EM tend to outperform after financial-economy bubbles deflate.

Claude’s Take

This is a clean, articulate statement of orthodox global value investing from people who clearly live it — no gimmicks, no torture of the framework to justify a hot stock. The most quotable insight isn’t an investment pick; it’s Heck’s observation that hyperscaler capex going from 20% to 100% of cash flow “can only happen once.” That reframes the AI-spending debate better than most dedicated tech commentary, and the GPU-versus-railroad depreciation point is genuinely useful.

Where to keep your guard up: this is, at the end of the day, a fund-marketing appearance. Every business they name is something they own, presented at its most flattering angle, and you’re hearing the thesis without the counter-case. The Walmex pitch is elegant but the multiple gap partly reflects real risks (Mexican political and currency exposure) that get a one-line wave. The “back to the periodic table” cycle theory is a tidy narrative that’s easy to assert and hard to disprove — value managers have been calling the international/real-assets turn for over a decade. And the firm’s actual long-run performance versus a simple global index never comes up, which is the question that would have made this rigorous rather than merely articulate.

Scoring it a 7: high signal-to-noise, several ideas worth keeping (scarcity framing, diversification-as-humility, the capex insight, time arbitrage), let down only by being a one-sided showcase. Jake Taylor’s Roman-Empire-as-customer-acquisition-cost segment is a fun standalone bonus — the flywheel that runs out of cheap conquests is a clean mental model for any growth business.