Why does Resource Extraction Produce Such Concentrated Wealth?

In this issue:

• Why does Resource Extraction Produce Such Concentrated Wealth?—The business of getting metals and minerals out of the ground has produced some big fortunes, and those are disproportionate to its share of global market cap. There are many reasons for this.
• RLHF—One source of high-quality training data: what kinds of rules tend to make sense over centuries, especially among people who aren't sharing notes?
• When Will AI Show Up in Labor Market Data?—You can look at official employment data for software engineers if you want, but it's going to be an increasingly noisy metric.
• Aligning Incentives—Credit markets don't always take equity markets seriously, but they do care when credit people are willing to take equity risk.
• Local Volatility—Making money on commodities trading by being closer to where people are desperate to trade.
• FDE Shortage—Tech M&A is often a matter of buying time.

The Diff May 11th 2026

0:00

/807.810612

1×

Talk to this post with ReadHaus.

Why does Resource Extraction Produce Such Concentrated Wealth?

I’ve been spending some time at Anomaly talking to some early-stage companies selling into the mining sector. Which is a reminder of a minor economic puzzle. There's an old finance joke that a mine is "a hole in the ground with a liar on top." Investors in various resource-extraction sectors, and in particular companies, have reason to believe this. 3.1% of the Forbes billionaires list made their money in metals and mining, but it's about 1.9% of global listed equities market cap. If you take the hundred biggest countries by GDP, and look up the source of wealth for the richest person in each of them, you'll see that mining is responsible for eight of them.

And yet, mining is also a respectable business with plenty of responsible large-cap companies that operate in a conventional way, return capital to shareholders, etc. Since the launch of the S&P Metals & Mining ETF, its returns have been 6.4% annualized, compared to 11.4% for SPY over the same timeframe, but over the last ten years it's been better to own metals and mining (19.7%) than a broad equity index (15.5%). Shareholders have been rewarded for backing mining from time to time, but in the aggregate it's been a better deal to own the other 98% of the market instead.

The first driver of this is the classic problem that mining is cyclical, and it absorbs the most capital at the top of the cycle. A cyclical business that produces average returns over a full cycle will give investors worse total returns, because at the peak it won't return quite enough capital, and because the equity slice of the capital structure can't always survive the worst part of the cycle. The equity in some high-cost miner might 100x during a good cycle, but that 100x return may be enjoyed mostly by former bondholders who got most or all of the equity when it reorganized: shareholders in Warrior Met Coal have seen their holdings compound at 33% since it went public in 2017. But "Warrior Met" was a new holding company formed by the first lien debtholders of Walter Energy, whose shareholders were zeroed in 2016.

Another issue, related to the cycle, is Engineering Ego: people who go into the business of solving complex technical challenges around getting ore out of the ground are going to be excited if there's a uniquely extreme way to do it. Investors are understandably skeptical of these big projects, but a given set of estimates will have some wiggle-room, and if someone really wants to set a record, they'll be able to justify it. This is a cyclical phenomenon, where sometimes it's in the opposite direction; after the Mukluk disaster, a billion-dollar dry hole in 1982, there was probably some reluctance to do any big-ticket speculative projects. This kind of ego-driven situation is not purely a bad thing. In some industries, ego is a continuous wealth transfer to, rather than from, investors: prestige jobs pay less than what would otherwise be the market value for their skills, and people at prestige companies will feel a healthy level of impostor syndrome.^[1]

There's also the fractal element of luck. There's the obvious piece that some fortunes, personal and national, start with the accident of who owns which plot of land, or which promising initial find either does or doesn't pay off. But within that, there are accidents of timing: if you were betting on a spike in demand for oil for idiosyncratic reasons, and suddenly bombs start falling on Tehran, you'll make money (and, incidentally, mitigate the supply crunch), then even though you had no idea what was coming next, you'll make money.^[2] And even independent of that, mines have complex supply chains, and sometimes have n-of-1 equipment—in 2009, a mine responsible for 10% of global uranium production was mostly knocked offline for about seven months. These kinds of incidents are rare, but they're more likely to happen when equipment is at full utilization, spare part inventories have been drawn down, workers are tired, etc.—i.e. it's more likely to happen when there's a lot of money to be made pulling particular kinds of ore out of the ground.

One piece of luck is political. The returns from owning land that turns out to contain valuable mineral and metal deposits is a pretty good example of an economic rent, and these tend to distort economic behavior no matter what form they take. A common default is for the host government to own the resources, and to charge royalties to mining companies, with those royalties ideally capturing most of the gap between that mine's production cost and the marginal cost of production elsewhere. But, particularly in poor countries, this creates a perverse situation where most of the wealth produced in that country each year has a distribution that's up for negotiation. A pleasant social democracy will probably divert most of that money into the government's budget, and will pay for a generous safety net while building up a rainy-day fund. But in a place with weaker institutions, the default is that the main use of revenue from a resource bonanza is a mix of keeping existing elites in charge and rewarding them for their loyalty, and that the original owner of those resources is not necessarily the one who reaps the benefit.^[3] Combine this with the fact that resource exports put upward pressure on currency, making other exporters less competitive and making foreign luxury goods more affordable for the people in charge. So that wealth can end up being very efficiently allocated to the local Wabenzi, while everyone else starves. This is a recipe for an extremely uneven distribution of wealth: Forbes might list the relevant source of wealth as "mining," but it's more apt to say that these rich people are bagmen. In a setup like that, there's enough cash flow that they don't need to tap public markets, and keeping things private also means not having to produce detailed financial statements or details on executive compensation practices. But it also means that natural resource wealth in countries with weak institutions is very tenuous. Even if the total upside from being less corrupt eventually swamps the near-term flow of bribes and kickbacks, it's hard to move in that direction: the constituency for remaining corrupt has more concentrated interests than the constituency for getting less so.

This regulatory risk doesn't just show up at the level of which military and legal system controls the physical territory where the resources are. It applies to where they're traded, too. Commodity exchanges have different incentives than other exchanges. Their transactions are anchored to real-world transfers of physical products, and that means that mining companies, and commodity consumers, are an important constituency. So there's some commodity price volatility that can't quite be hedged, because a losing counterparty is simply too powerful. This happened in the 70s with potatoes and in 2022 when the London Metals Exchange simply canceled a day's trading when a major nickel producer faced a short squeeze. These were both events that led to more wealth concentration for the parties on the winning side, but they also explain why commodities extraction is such a bumpy ride.

The last big issue is information asymmetry between management and investors. It's just a little too easy to fudge the details. There are periodic headlines about various places having trillions of dollars worth of resources, like Afghanistan and the ocean and asteroids. This number is very different from a quantity like "reserves," which is an estimate of how much material it's economically viable to extract. There are companies that have an SIC code that says they're mining, but that primarily manufacture press releases and capital raises. And someone who's spent a while in an industry, hobnobs with other CEOs, and pays close attention to commodity trader psychology might have an edge in timing the cycle, which only has a practical effect if it means that the broader investing public is relatively overweight their company, and they, as executives, are increasingly underweight it, when profits peak.

All of these factors together mean that there are many ways for outside investors to have a very different financial experience from a given natural resource company than the people who started it. But it also means that companies in the industry can, if they choose, be in the business of providing outsourced corporate governance services and due diligence on behalf of investors. The total value of a mine is going to be higher when it doesn't have to have a line item for bribes, and the value of a mining company is higher if it doesn't periodically engage in performative drilling just to impress shareholders. The airline sector demonstrates that an industry can graduate to having a smaller valuation discount if it can credibly signal that it's built to last more than one cycle. The resources are going to be there regardless, and, over time, they'll end up in the hands of whoever has the lowest cost of capital.

1. For a while, some big tech companies seemed to use a strategy where they'd devote a small fraction of their resources to moonshots, partly as a way to recruit people who'd end up mostly improving ad targeting, but now that one of those moonshots is the technology that defines the current era in tech, they can't really do that. The former moonshots are now the main source of long-term revenue growth! ↩︎

2. Unless the oil you decided to produce has to get shipped through the Strait of Hormuz. ↩︎

3. A good proxy for how strong the rule of law is in a given country is: at what value of reserves does "I found valuable resource deposits on land I own" flip from being an asset to a potentially fatal liability? By that standard, the US is a relatively new member of the strong rule of law club. ↩︎

Diff Jobs

Companies in the Diff network are actively looking for talent. See a sampling of current open roles below:

• Series A startup building multi-agent simulations to predict the behavior of hard to sample human populations is looking for a founding recruiter who’s able to attract and close the best research and engineering talent in the world. Experience building high-quality teams as a former founder, VC, or operator a plus. No formal experience in a “recruiting” function required. If you have experience communicating and persuading smart, disagreeable counterparties of your vision, this is for you. (NYC)
• Well-funded, frontier AI neolab working on video pretraining and computer action models as the path to general intelligence is looking for researchers who are excited about creating machines that learn from experience, not text. Ideally you have zero-to-one pre-training experience and/or are a high-slope generalist who’s frustrated that the big labs aren't doing this. (SF)
• A Fortune 500 cybersecurity company with decades of proprietary security data is running an internal incubation with a pre-seed startup mentality and a mandate to build something new in AI. They are looking for a founding product engineer who can ship fast, an engineer with a security background who’d be excited to contribute to OpenClaw’s security efforts, and a generalist (ex-banking/consulting/PE background preferred) who wants to wear a bunch of different hats. Comp is FAANG+ and cash heavy. If you want to build something new in AI, but also need runway, this is for you. (SF/Peninsula) 
• High-growth startup building dev tools that help highly technical organizations autonomously test and debug complex codebases is looking for senior product managers who enjoy defining developer-facing APIs and abstractions. Experience with fuzzing or property-based testing a plus! (London, D.C.)
• A leading AI transformation & PE investment firm (think private equity meets Palantir) that’s been focused on investing in and transforming businesses with AI long before ChatGPT (100+ successful portfolio company AI transformations since 2019) is hiring experienced forward deployed AI engineers to design, implement, test, and maintain cutting edge AI products that solve complex problems in a variety of sector areas. If you have 3+ years of experience across the development lifecycle and enjoy working with clients to solve concrete problems please reach out. Experience managing engineering teams is a plus.

Even if you don't see an exact match for your skills and interests right now, we're happy to talk early so we can let you know if a good opportunity comes up.

If you’re at a company that's looking for talent, we should talk! Diff Jobs works with companies across fintech, hard tech, consumer software, enterprise software, and other areas—any company where finding unusually effective people is a top priority.

And: we're now actively deploying capital into early-stage companies through Anomaly. Our focus is on defense, logistics, robotics, and energy. If you'd like to chat, please reach out.

Elsewhere

RLHF

Most people at AI companies, and most people writing regulations for them, would like chatbots to be good at helping people to be good, and worthless for helping them to be bad. It's rare to get this kind of consensus on such a high-profile issue, so it's important to take advantage of that by quickly figuring out a universally agreed-upon definition of The Good. The old approach was that The Good was, basically, the norms of well-educated people living in large coastal cities, though they've largely loosened up in the last year or so. One of the lessons of AI is that you want to use as much high-quality training data as you can, and, from a purely evolutionary point of view, polling religious authorities is a good way to do that. Regardless of truth value, religions that survive have to, in some sense, actually work: they have to promote behaviors that help the ingroup, and they have to be able to handle edge cases, quirky personalities, social upheaval, and the like. If a religion survives the upheaval of centuries and millennia, it's probably doing something right. And religions tend to exhibit some convergent evolution, with some absolutes (murder, etc.) and some cases where they tend to compromise (there are very old religions with a tradition of tithing, and with some concept of a person who renounces worldly possessions, but it's hard to find one that's been able to thrive for a long time while making everybody renounce their possessions). And, importantly, they don't have convergent evolution on matters that are more specific to religion rather than other moral systems: they don't all settle on the same prophet, have mutually-contradictory dietary taboos, vary widely in where they land on assorted social issues, etc. So this probably leads to better moral calibration on human universals, and a better-calibrated wide range of possibilities on more transient issues.

When Will AI Show Up in Labor Market Data?

The WSJ notes that software engineering unemployment ticked up last month, to 3.8%, but also that it's down from 7% (!) in September and October ($, WSJ). Which is another way of saying that whatever it's measuring is very hard to track through that methodology. For one thing, as workers spend a smaller fraction of their time actually typing out code, and as the technical requirement for building simple prototypes drops, there will be people who produce more lines of code per year than the average software developer did in a lifetime and whose job title will be in marketing, or HR, or some other function. It wouldn't be the first time it took a while to find the right metrics; "eyeballs" were a topic throughout the 90s, "registered users" and “pageviews” prevailed in the early 2000s, and it took the rise of hyper-addictive social media services for anyone to have a daily active user number worth bragging about. But, for now, it's a good idea to be very skeptical that existing metrics will track what they're supposed to.

Aligning Incentives

Equity people tend to think of equity as ownership of the upside, but when credit people talk about equity, they'll sometimes talk about it as the "first-loss" tranche—it does get a better return when things go well, but its most important feature is that equity holders take losses before creditors do. That means that private credit managers can signal credibility by buying equity in their investment vehicles. Asset managers who focus on one asset class tend to limit their creativity, but if they're looking across the capital structure, they can find opportunities to get a good return from taking that first-loss risk themselves, while also giving everyone else room to keep those assets marked close to 100 cents on the dollar.

Local Volatility

Big oil companies achieve financial stability the way big banks do: the same circumstances that lead to higher uncertainty and sometimes lower margins in their core business lead to gains in their trading business. And this happens to be true on a geographical basis: European oil companies, who are more exposed to disruptions in Middle Eastern oil supply, are also making more money trading oil ($, FT). It's an advantage for commodities traders to be as close to the fundamentals as possible, and there's no better way to do that than to be part of the org chart that includes people who are turning high-level fundamentals (i.e. less oil supply traveling via the Strait of Hormuz) into specific reactions.

FDE Shortage

The Diff recently speculated about a shortage of forward-deployed engineers ($). OpenAI is acting as if it's already here: they're acquiring a company called Tomoro, an AI implementation consulting company. Part of what they're paying for is headcount (not that they aren't hiring them directly). But they're also paying to rush through the first few months of finding a prospect, negotiating terms, and actually getting an engagement going. Once there's a bottleneck, there's an enormous premium on anything that saves time.