In 2024, a cutting house in Surat, India could sell you a one carat, lab grown, VS2 clarity diamond, chemically and optically identical to one pulled from a Botswanan mine, for around eight hundred dollars. The natural version of that same stone was going for four thousand. Nothing about the lab diamond was inferior. Under a jeweler’s loupe, even trained gemologists need a specialized machine to tell the two apart.

Subscribe now

And yet for most of the twentieth century the natural stone commanded a fortune while a synthetic one, chemically superior in some respects since labs can grow diamonds with fewer inclusions than the earth ever managed, was treated as industrial grade junk, fit for saw blades and drill bits, never for a ring finger.

That gap between eight hundred dollars and four thousand is not a story about carbon. It is a story about what happens when a market built entirely on manufactured scarcity meets a technology that manufactures the same thing more honestly.

A monopoly, not a mineral

Start with what actually made diamonds valuable, because it was never geology. Diamonds are not particularly rare. Geologists have long known that gem quality stones exist in quantities large enough that, absent intervention, prices would have collapsed decades ago. What kept them expensive was control of supply, engineered with a precision that would make a game theorist blush. Cecil Rhodes founded De Beers in 1888, and by the mid twentieth century, under the Oppenheimer family, the company controlled roughly eighty to ninety percent of the world’s rough diamond supply at its peak.

That is not a company operating in a market. That is a company that is the market, deciding how many stones reach buyers in any given year and stockpiling the rest in vaults in London to make sure the tap never opened too far. This is the textbook case of a monopolist restricting quantity to keep price above the competitive level, except De Beers did it not for a single country but for the entire planet, for a hundred years, without ever technically calling itself a monopoly in public.

Manufacturing desire

But controlling supply only gets you halfway. You also need people to want the thing badly enough to pay whatever price you set, and that is where De Beers ran one of the most successful demand side operations in the history of consumer marketing. In 1947 the agency N.W. Ayer coined the line “A Diamond Is Forever” for a client that, by its own internal memos, was worried that diamonds had no functional resale market and therefore no durable reason to buy one beyond sentiment.

So the campaign built the sentiment industrially. It planted diamond engagement rings in Hollywood films. It invented the notion that a man should spend a month’s salary, later revised upward to two, on a ring, a number with no basis in anything except that it was large enough to matter and vague enough to feel like tradition rather than instruction.

Subscribe now

This is a masterclass in what economists call a Schelling point, a focal number that a whole society converges on not because it is correct but because everyone assumes everyone else is using it too. Once two months’ salary became the norm, deviating from it did not just mean spending less. It meant signaling something uncomfortable about your relationship, or your income, to everyone who saw the ring.

The diamond stopped being a commodity and became, in Thorstein Veblen’s old formulation, a good whose value rises specifically because it is expensive and visibly so. Nobody buys a diamond for its physical properties. They buy it because it is legible to other people as an expensive, and therefore serious, gesture.

The substitute that broke the spell

That legibility depended entirely on one condition holding indefinitely, that natural diamonds remain the only socially valid diamonds. Lab grown stones broke that condition, and the numbers since have been brutal. Wholesale lab grown prices have fallen somewhere between seventy and ninety six percent since 2018, depending on whose trade data you trust, driven in large part by a more than threefold expansion in cutting and polishing capacity out of India, particularly Surat, which now processes the overwhelming majority of the world’s diamonds, natural and synthetic alike.

That collapse did not stay contained to the lab grown category. It dragged natural prices down with it, because once a nearly identical substitute exists at a fifth of the price, the forever premium becomes very hard to defend to a rational buyer. Average one carat natural stones fell from roughly six thousand dollars in 2021 to around four thousand two hundred by 2025. Anglo American, De Beers’ parent, has written the unit down by 6.8 billion dollars across 2023, 2024 and 2025 alone, and by late 2024 De Beers was sitting on two billion dollars of unsold rough stone, its largest stockpile since the 2008 financial crisis. The company that once decided how much of the world’s most desired good would exist in any given year is now trying to find a buyer for itself.

Subscribe now

What the price was really measuring

What makes this a genuinely interesting economics story, rather than just a satisfying one about comeuppance, is the information asymmetry that made the original cartel possible in the first place. For a century, the average consumer had no reliable way to assess a diamond’s true cost of production or its actual abundance. They relied entirely on the seller’s framing, reinforced by cultural signaling so deep it felt like biology rather than branding. Lab grown diamonds did not just add a cheaper substitute to shelves. They handed consumers a benchmark. Once you can buy an optically perfect diamond for eight hundred dollars, the four thousand dollar version stops being priced by scarcity and starts being priced, transparently, by story. Some buyers still pay for the story.

Pandora, one of the largest jewelry retailers in the world, dropped natural diamonds from its lineup entirely in 2022 and has grown since, betting correctly that a meaningful share of consumers were paying for the diamond, not the myth. Others have not made that switch, and De Beers is now funding an industry wide marketing consortium, the Luanda Accord, explicitly to rebuild the narrative that natural equals authentic, a remarkable admission that the product’s value was always more narrative than mineral.

The next time something in a market seems expensive for reasons that feel more cultural than functional, run the diamond test on it. Ask who controls the supply, who financed the story that made scarcity feel meaningful, and what happens to the price the moment a technically identical alternative shows up without the myth attached. Sometimes the answer is that the price was protecting something real. Diamonds tell you how often it isn’t.

What makes this story genuinely interesting to an economist isn’t the headline. It’s the mechanism. Snacks and semaglutide are about as unrelated as two products can be. One is a bag of chips, the other a weekly injection prescribed by an endocrinologist. And yet demand for one just fell because supply of the other expanded. That is cross-price elasticity playing out in the wild, except it’s operating through a channel economics textbooks rarely discuss: not price, but appetite itself.

Subscribe now

Substitutes without a shared shelf

Normally, cross-price elasticity describes how the price of Coke affects demand for Pepsi, or how a jump in beef prices nudges shoppers toward chicken. Those goods compete because they satisfy the same want through the same channel, taste and hunger. GLP-1 drugs do something stranger. They don’t compete with a Snickers bar on price or flavor. They compete with it by dulling the biological signal that made you want the Snickers bar in the first place. Semaglutide slows gastric emptying and mutes the reward circuitry that makes ultra-processed food so compulsively craveable. In effect, the drug substitutes for the craving, not the product.

This is why a landmark Cornell study using Numerator purchase-panel data, tracking roughly 150,000 households, found something so specific. Spending on savory snacks dropped by about 10 percent among GLP-1 adopters, with similarly steep declines in sweets, baked goods and cookies. Meanwhile spending on yogurt, fresh fruit, nutrition bars and meat snacks actually rose. The same household, same income, same store trip, just a completely different basket. That is what a demand shock through the appetite channel looks like when you can actually observe it at scale.

Walmart noticed early. Back in October 2023, when GLP-1 adoption was still a niche phenomenon, the company’s US CEO John Furner told Bloomberg that shoppers filling GLP-1 prescriptions at Walmart pharmacies were buying “less units, slightly less calories” than comparable shoppers who weren’t. It sounded like a footnote at the time. Two years later it looks like the earliest data point in one of the more unusual demand-destruction stories in modern retail.

Subscribe now

When 9 percent of the population moves the whole curve

Here is where the economics gets sharper. A KPMG analysis found that GLP-1 users reduce annual caloric intake by roughly 21 percent and cut monthly grocery spending by about 31 percent. Those are individually dramatic numbers, but the aggregate effect depends entirely on how many people are taking the drug, and that share has been climbing fast. The proportion of US households reporting at least one GLP-1 user rose from about 11 percent in late 2023 to more than 16 percent by mid-2024. Morgan Stanley projects that as much as 9 percent of the entire US population could be on some form of GLP-1 medication by 2035, driving what the bank estimates as a 4 percent long-term decline in soft drink, alcohol and salty snack consumption.

This is a textbook illustration of how a small behavioral shift among a growing minority reshapes an entire industry’s baseline. You don’t need most people to change their eating habits for a food conglomerate’s growth model to break. You just need enough people, spending enough less, consistently enough, that the category’s structural assumption, steady 2 to 4 percent annual growth in salty snacks, stops holding. Piper Sandler now estimates a persistent 20 to 30 basis point drag on confectionery and salty snacks from GLP-1 adoption alone, compounding year after year like a slow leak in a tire nobody wants to admit is losing air.

Subscribe now

Creative destruction dressed as a protein bar

What happens next is where the story becomes genuinely fascinating rather than just gloomy. Faced with a shrinking addressable market, snack makers didn’t simply cut costs and wait. They repositioned. PepsiCo acquired Siete, a plant-based, functional snacking brand, explicitly to move away from commodity salty snacks. Conagra rolled out “On Track” badges on more than two dozen Healthy Choice frozen meals, flagging them as high-protein, low-calorie and fiber-rich, a direct pitch to GLP-1 users trying to preserve muscle mass while eating less. Nestlé launched Vital Pursuit, a range of portion-controlled, high-protein ready meals aimed squarely at the same audience.

The interesting wrinkle is who’s actually buying these products. Roughly 77 percent of Vital Pursuit’s sales come from households with no current GLP-1 user at all. The reformulation wasn’t really a niche accommodation for drug users. It became a broader signal of health positioning that appealed to anyone newly primed to think about calorie density, whether or not they’d ever touched an injection pen. That’s Schumpeterian creative destruction operating in real time, a threat to the old product category becoming the seed of an entirely new one, with the global protein snacks market now projected to nearly double from about 4.9 billion dollars to 10.8 billion dollars by the mid-2030s.

Subscribe now

The effect isn’t confined to the West either. In South Korea, Abbott launched Protality, a nutrition shake specifically engineered to address the nutrient and muscle-mass deficiencies that emerge when GLP-1 users eat dramatically less, a sign that the same substitution logic playing out in American supermarkets is already being priced into product strategy across Asia, even in markets where GLP-1 adoption itself remains far lower than in the US.

A country’s GDP versus a country’s snack aisle

Zoom out far enough and the two sides of this trade start to look almost absurd next to each other. Novo Nordisk, the Danish manufacturer of Ozempic and Wegovy, saw its market capitalization climb past 400 billion dollars, at one point exceeding Denmark’s entire annual GDP. A single pharmaceutical company’s valuation outgrew the output of its home nation, while on the other side of the ledger, snack conglomerates worth tens of billions were writing down growth targets built on decades of stable, almost boring, consumption patterns.

Subscribe now

It’s worth sitting with the asymmetry. The industry losing value employs millions, occupies enormous shelf space, and has spent a century engineering products for maximum craveability. The industry gaining value makes a molecule that, quietly and at a cellular level, undoes that engineering. Nobody planned this collision. Novo Nordisk wasn’t trying to compete with Frito-Lay.

But that’s the thing about genuine substitution effects, the most consequential ones rarely show up on the same shelf, in the same aisle, or in the same industry classification code. They show up in a body’s bloodstream first, and in a balance sheet months later. The next question worth asking isn’t which snack company adapts fastest. It’s what other industries are quietly priced against a craving nobody has mapped yet.

What you are experiencing has a name, and it was discovered not by a game designer but by a psychologist working with pigeons in the 1950s. B.F. Skinner found that animals rewarded on a variable ratio schedule, meaning an unpredictable number of actions before a payout, pressed the lever far more compulsively than animals rewarded every single time or on a fixed schedule. Predictable rewards get boring.

Unpredictable ones get addictive. Slot machines were built on this discovery decades ago, tuned by mathematicians to keep the ratio of near misses and small wins just high enough to keep the lever getting pulled. Gacha games run the identical architecture. A wish is a lever pull. Primogems are chips. The five star character is the jackpot. The only real difference is that one sits inside a casino with a license bolted to the wall, and the other sits inside an app that a nine year old can download for free.That difference is not an accident. It is the entire business model.

Subscribe now

The missing third element

Gambling law in most of the world defines gambling through three elements: you pay something of value, the outcome depends on chance, and you can win something of value in return. Loot boxes and gacha pulls clear the first two without argument. The third is where the industry built its escape hatch. In almost every major market, the characters, weapons, and skins you win cannot be legally cashed out for real money.

There is no redemption window, no chip counter that converts back to cash. Because the prize has no formal exchange value, most regulators simply cannot classify the mechanic as gambling, even though the psychological engine underneath is the one Skinner described seventy years ago. This is regulatory arbitrage in its cleanest form, designing a product to sit one legal technicality away from a category with expensive compliance costs, age restrictions, and licensing regimes, while replicating everything that makes that category profitable.

Subscribe now

One country at a time

Not every regulator accepted the workaround quietly. Belgium’s Gaming Commission investigated loot boxes in four popular titles in 2018 and concluded that several of them did meet the legal definition of games of chance, chance being sufficient even without a formal cash-out mechanism, because the games created a market where items were traded and monetized informally outside the platform.

Rather than fight it, EA simply pulled loot boxes from FIFA’s Ultimate Team mode for Belgian players and left the mechanic untouched everywhere else. That single country carve-out tells you something important about how these companies actually think about regulation. It is not a moral line, it is a cost center to be geofenced around, the same way a factory routes pollution to the jurisdiction with the loosest environmental rule.

Subscribe now

Japan’s scalpel

Japan offers the sharper case study, and it predates the Western panic over loot boxes by half a decade. In 2012, Japan’s Consumer Affairs Agency banned a specific mechanic called kompu gacha, or complete gacha, where players had to collect a full set of randomly drawn items to unlock a bonus prize, a design that GREE’s Dragon Collection and several rivals had built entire economies around. The agency ruled this specific combinatorial structure violated an existing law against disguised prize-based sales going back to 1977.

Billions of yen in market value evaporated from Japanese mobile publishers within days of the announcement. But notice what the ban did not touch: the underlying gacha pull itself, the single random draw for a single item, remained completely legal and remains the dominant monetization mechanic in Japanese mobile gaming today. Regulators closed one specific loophole with surgical precision and left the load-bearing structure standing. This is the pattern across nearly every jurisdiction that has tried to regulate gacha. Rules arrive narrow, publishers adjust the mechanic by a few degrees, and the core variable ratio engine survives intact.

Who is actually funding the arena

The economics of who actually funds this engine explain why publishers defend it so fiercely. Across mobile gaming broadly, the top one percent of paying users generate somewhere around a quarter to a third of total revenue, and in gacha specifically, five to ten percent of players convert into payers who together account for the large majority of the take. HoYoverse’s three flagship titles, Genshin Impact, Honkai Star Rail, and Zenless Zone Zero, pulled in a combined 1.77 billion dollars across iOS and Android in 2024 alone, and Genshin’s China revenue crossed 5 billion dollars by October of that year through a domestic payment channel that Western app store trackers don’t even see.

Subscribe now

This is price discrimination taken to its logical extreme. The free player is not the customer, the free player is inventory, a body in the ecosystem that makes banners feel populated and social and worth chasing.

The actual product being sold is a carefully calibrated psychological hook aimed at the small fraction of players whose loss aversion and sunk cost thinking will not let them stop three pulls short of a pity guarantee.

When the state blinks

China’s own regulators nearly moved to shut this down entirely. In December 2023, the National Press and Publication Administration published a sweeping draft rule that would have capped monthly spending, banned daily login rewards, and prohibited gacha mechanics for minors outright. Within weeks, the announcement had wiped tens of billions of dollars off the market value of listed Chinese game publishers including Tencent.

The rule was quietly withdrawn the following month and has not been reissued. It is one of the clearest recent examples of an industry using its own systemic financial weight as leverage against the state trying to regulate it, a live demonstration that when a monetization mechanic becomes large enough, the deterrent to reforming it stops being ethical and becomes macroeconomic.

Disclosure without a cure

Disclosure laws, now active in China, South Korea, and elsewhere, are the current regulatory frontier, forcing publishers to publish the actual odds behind every pull. It is a real improvement over the opacity of a decade ago. But disclosure solves an information asymmetry problem, and the thing keeping players pulling was never really a lack of information. It was Skinner’s lever, working exactly as designed, in full view of everyone, odds published in eight point font right below the button.

The next frontier of this mechanic is already running quietly in the background of the games you play, publishers using machine learning to personalize which banners you see and when, tuning the exact moment a targeted offer appears to the moment your own spending data says you’re most likely to say yes. Which raises a genuinely uncomfortable question for the next decade of regulation: what happens once the slot machine doesn’t just know the odds, but knows you better than you know yourself?

That company is TSMC, and by late 2025 it controlled somewhere around 70 percent of the global foundry market and roughly 90 percent of the world’s most advanced chip production. One company. One island. Ninety percent of the frontier.

The Physics of an Unbeatable Lead

Start with why that number is even possible, because it isn’t an accident of geography or cheap labor. It’s what happens when a business becomes so capital intensive that scale stops being an advantage and starts being the entire game. Building a leading edge fab now costs somewhere north of twenty billion dollars, and every new process node, from 7 nanometers down to 3 and now 2, demands another round of investment that only makes sense if you can spread it across enormous volume.

Economists call this increasing returns to scale, and it behaves very differently from the diminishing returns you learn about in an introductory course. In a normal industry, a bigger competitor eventually runs into higher costs per unit. In advanced semiconductors, the biggest player gets cheaper and better at once, because more volume funds more research, which produces higher yields, which funds the next round of volume.

Subscribe now

Samsung has poured tens of billions into catching up and still struggles with yield on its most advanced nodes. Intel, once chipmaking’s undisputed leader, needed the American government to take a direct equity stake in the company just to keep its foundry ambitions alive. When the cost of staying in the race grows faster than the number of players who can afford it, you don’t get competition. You get a natural monopoly that happens to trade on a stock exchange.

Why Nobody Can Just Leave

The second force at work is subtler, and it explains why even TSMC’s biggest customers can’t simply walk away. When Apple designs a chip, its engineers don’t write generic instructions and hand them to whichever factory has an opening. They design around TSMC’s specific process and packaging techniques, down to details that would need rebuilding almost from scratch elsewhere. Economists call this asset specificity, a concept associated with Oliver Williamson’s work on why firms choose to depend on a single partner rather than shop around.

Every iPhone processor since 2016 has been made exclusively by TSMC, and that exclusivity is the product of years of co-engineering that makes switching enormously expensive for both sides. Nvidia’s most advanced AI chips, the ones training the large models everyone is suddenly obsessed with, are locked into the same arrangement. This is why “just build another fab somewhere else” sounds simple in a headline and is almost meaningless in practice. You cannot copy and paste twenty years of relationship-specific engineering.

Subscribe now

A Peace Treaty Written in Balance Sheets

Then there’s the part of the story that reads less like an economics textbook and more like a geopolitical thriller, except the thriller is actually a fairly cold blooded equilibrium. Taiwan’s dominance in chipmaking has a name in policy circles: the silicon shield. Taiwan is thought to be safer from invasion precisely because the global economy, including China’s own tech industry, depends so heavily on its fabs that any disruption would be catastrophic for everyone, aggressor included, a strange modern version of mutually assured destruction run through balance sheets instead of missiles.

But that shield only holds as a Nash equilibrium as long as no other actor can credibly replace the capacity being threatened, which is why the chokepoints run deeper than Taiwan alone. Every leading edge fab in the world, TSMC’s included, depends on extreme ultraviolet lithography machines made by exactly one company, ASML, a Dutch firm built on decades of nearly unreproducible engineering knowledge. When the United States wanted to slow China’s chip ambitions, it didn’t sanction Chinese chip designers directly.

Subscribe now

It leaned on the Dutch government to restrict ASML’s export licenses, because controlling the machine that makes the machines is a more efficient chokehold than controlling the end product. Whoever holds the narrowest point in a chain gets leverage disproportionate to its share of the value created, which is why geopolitics increasingly runs through export lists rather than tariffs.

The Chokepoint Has Already Been Pulled

We’ve actually watched this chokepoint logic get used in anger before, and it didn’t involve Taiwan at all. In 2019, Japan restricted exports of three obscure chemicals, including hydrogen fluoride and photoresist, to South Korea, ostensibly over a historical trade dispute but widely read as retaliation in a diplomatic fight.

Samsung and SK Hynix, which depended on Japanese suppliers for these inputs, suddenly faced real production risk over compounds most people had never heard of, a preview in miniature of the leverage a concentrated supply chain hands to whoever sits at its narrowest point.

The Insurance Nobody Wanted to Buy

What makes all of this worse, economically speaking, is that the global chip supply chain was optimized for exactly the wrong thing. Just in time manufacturing minimizes cost by holding almost no buffer inventory anywhere in the system, which is brilliant until the system hits a shock. When the pandemic disrupted chip supply in 2021, most Western automakers, having spent decades stripping out inventory to boost efficiency, were forced to idle plants and lost tens of billions in revenue. Toyota barely stumbled, because after the 2011 tsunami it had rebuilt its supply chain around deliberate stockpiles and mapped supplier risk down to individual factories.

Toyota was, in effect, paying an insurance premium that looked like wasted money for a decade, until the shock came and it paid for itself many times over. Concentrated, hyperefficient supply chains generate an externality that never shows up in anyone’s unit cost: systemic fragility the market has no mechanism to price, because no single firm captures the benefit of building in slack that protects everyone else too.

Subscribe now

Paying to Rebuild What the Market Wouldn’t

Governments have started trying to buy that insurance directly, since firms won’t do it on their own. The American CHIPS Act helped fund TSMC’s Arizona buildout, now a project north of one hundred and sixty billion dollars, with equipment installation for advanced nodes accelerating into 2026 because Washington wants a version of TSMC’s capability that doesn’t sit across a contested strait.

It’s an expensive way to manufacture redundancy the market, left alone, would never have built, because redundancy is a public good and markets are notoriously bad at supplying those on their own.

A Second Monopoly, Built to Hedge the First

Here’s the part that should keep you up a little later than the headlines suggest. Every dollar spent diversifying away from Taiwan still runs through the same handful of equipment makers and the same scarce, physics-defying engineering knowledge that made TSMC dominant in the first place. Arizona’s fabs are TSMC fabs.

The insurance policy is being underwritten by the very monopoly it’s meant to hedge against. So the real question isn’t whether the world can reduce its dependence on a single company in a single place. It’s whether concentration of this kind can ever actually be undone, or whether we’re just building a second, more expensive version of the same chokepoint and calling it resilience.

Now picture the American equivalent happening at the same moment, a few time zones away. A five star high school quarterback is deciding between three college programs. Each school has a booster funded “collective” quietly assembling a package, texted and emailed in fragments, never disclosed to the other bidders, never benchmarked against a public number anyone can point to.

Subscribe now

Nobody in that negotiation, not the recruit, not his family, not the coaches, not the compliance officer down the hall, actually knows what a quarterback of his caliber is worth. They are all guessing, and they are all guessing differently.

The Monopsony That Lost Its Grip

This is the strange state of college athlete pay in 2026. The total NIL market has grown from roughly nine hundred million dollars in its first year, 2021 to 2022, to something north of two billion dollars now, more than doubling in just a few seasons. That is a genuine labor market appearing almost from nothing. And yet it is a labor market with almost none of the infrastructure that normally makes a market function, which is what makes it such a strange and revealing case study in what economists actually mean when they talk about price discovery.

For nearly a century, the NCAA ran something close to a textbook monopsony. There was effectively one buyer of major college athletic talent, the cartel of member schools acting in concert, and thousands of sellers with nowhere else to sell their labor if they wanted to play at the highest level while still enrolled in school. Wages were fixed at zero, dressed up as scholarships, and justified with the language of amateurism.

Subscribe now

When the Supreme Court finally took this apart in NCAA v. Alston in 2021, the majority opinion stated plainly that the NCAA held monopsony power capable of depressing wages below competitive levels. Justice Kavanaugh’s concurrence went further, arguing that no other American industry would be allowed to fix labor prices this way and call it tradition. Economists like Roger Noll had been making some version of this argument for decades. The Court just finally agreed.

Price Discovery, or the Absence of It

What nobody quite anticipated is what happens the instant a monopsony loses its wage fixing power without a replacement pricing mechanism sliding into place. A functioning market needs more than willing buyers and sellers. It needs some way for information about value to travel, so that today’s price becomes tomorrow’s reference point. The IPL auction does this by design.

Every bid is public, every past sale price becomes public data that shapes next year’s base price, and a right to match rule lets a team retain a departing player by matching the top offer, which anchors expectations even further. Prices still move, sometimes wildly, but they move in front of everyone, which is what economists mean by price discovery.

Signals Without Prices

College sports built almost the opposite structure. Collectives, the donor funded groups now responsible for something like eighty percent of total NIL spending according to Opendorse’s annual market reports, negotiate bilaterally and privately. There is no auction floor, no published comparable, no shared ledger of what the last player at the same position and skill level actually received. Into that vacuum rush proxies.

On3’s valuation index, built mostly from social media following and engagement, has become a de facto price signal even though it estimates market value rather than recording an actual transaction. LSU gymnast Livvy Dunne, whose four million dollar valuation makes her the highest earning woman in college sports, testified at the hearing over the 2025 House settlement that the valuation process used to determine athlete payouts was, in her words, badly flawed. When even the people being valued do not trust the number, you are watching a market that has signals but no real prices.

Subscribe now

Flying Blind on Both Sides

This is where information asymmetry becomes the whole story rather than a footnote to it. A Merrill Lynch survey found that only about eight percent of high potential young athletes work with a financial advisor, and sixty one percent admitted they had not handled their NIL deals well. Athletes are being asked to negotiate in a market where even sophisticated buyers, university compliance departments with lawyers on staff, cannot reliably tell what a fair price looks like.

Both sides are flying instruments in fog. That is a recipe for exactly the pattern economists predict when true value is unobservable, a market that swings between wild overpayment and quiet underpayment depending on who happens to have better information in any given negotiation.

The Winner’s Curse of the Blind Bid

The overpayment side shows up as something close to a winner’s curse. Arch Manning and Cooper Flagg were valued in the seven figures before either had played a single competitive college snap, priced almost entirely on potential, the college sports equivalent of bidding on an oil field before anyone has drilled.

Add the transfer portal, which now moves thousands of athletes a year and functions like free agency without multi year contracts, and schools face a market where this year’s winning bid can walk next spring with no penalty at all. Rational actors respond to that risk by bidding aggressively short term, exactly the instability you would predict from a market with no enforceable contracts and no shared pricing data.

Administering a Market That Just Learned to Run Wild

The 2025 House settlement was an attempt to reintroduce some administered structure into this, letting schools share revenue directly with athletes up to a cap of twenty and a half million dollars a year, rising four percent annually, with deals over six hundred dollars now supposed to run through a clearinghouse and meet a fair market value standard.

It is, in effect, an attempt to bolt a salary cap and a compliance office onto a market that spent four years learning to operate without either. Whether that produces something resembling the IPL’s transparent, self correcting auction, or simply relocates the guesswork one level up, from collectives to compliance officers interpreting what fair market value even means, is still an open question nobody involved can answer with confidence.

What a Fair Price Would Even Look Like

What is fascinating, and a little unsettling, is that we are watching a market invent its own pricing infrastructure in real time, under litigation pressure, with billions of dollars already moving before anyone agreed on the rules. Most labor markets built their pricing conventions over generations. This one is building them during live broadcasts.

The question worth sitting with is not whether college athletes deserve to be paid, that argument is settled. It is whether a market can ever price something fairly when it skipped the century of trial and error that usually teaches everyone else how.

Chelsea pocketed roughly £6 million in profit on a player nobody in west London had watched kick a ball in a Chelsea shirt. On paper that looks like a scandal, or at least a strange accident. It is neither. It is the system working exactly as intended, and understanding why requires borrowing a concept from finance rather than football.

The right, not the obligation

The concept is called a real option, and it explains a huge amount of what looks, on the surface, like reckless spending on unproven kids. A real option gives you the right, but not the obligation, to make a future investment once more information arrives. You pay a relatively small amount now to secure that right, and later, once you have learned more, you decide whether to invest further, sit tight, or walk away.

Oil companies use the framework to decide whether to drill a well. Pharmaceutical firms use it to decide whether to fund the next phase of a drug trial. Football clubs, whether or not their sporting directors would ever use the term, are now using it to decide whether to sign a seventeen year old. The transfer fee is the option premium. First team minutes, loan spells and contract extensions are the exercise decisions. Ângelo Gabriel wasn’t a failed investment. He was an option that Chelsea chose not to exercise, and they still sold it at a profit before it expired.

Subscribe now

Why volatility is the product

This is where the logic runs against everything an ordinary investor is taught. In most of finance, uncertainty is the enemy, something to be hedged away. In option pricing, uncertainty is the entire source of value. The more volatile the possible outcomes for an asset, the more valuable the option on it becomes, because your downside is capped at the premium you paid while your upside is theoretically unlimited. A calm, predictable career trajectory produces a cheap option because there is little to bet on.

A wildly uncertain one, where a player might become the best in the world or might flame out by twenty, produces an expensive one to buy outright but a hugely lucrative one to acquire early and cheaply. This is precisely why Brighton paid Independiente del Valle in Ecuador just £4.5 million for Moisés Caicedo in 2021. Nobody, including Brighton, knew whether an unheralded midfielder from a club few European fans could locate on a map would become a Premier League regular.

Subscribe now

That uncertainty was the whole point. Two and a half years later, once his volatility had resolved in his favour and Liverpool and Chelsea were locked in a bidding war for him, Chelsea paid £115 million for the same player. Brighton hadn’t predicted greatness. They had bought cheap optionality on a wide range of outcomes and let the market do the rest.

Buying a forward contract on a sixteen year old

Real Madrid’s deal for Endrick makes the option structure almost literal. In December 2022, when the Brazilian striker was sixteen and had made just seven senior appearances for Palmeiras, Madrid agreed to pay a guaranteed fee with the transfer only completing eighteen months later, once he turned eighteen and could legally move.

On top of the guaranteed base, Palmeiras negotiated add on clauses tied to goals scored while he finished out his development at home. That is not a conventional purchase. It is a forward contract with a built in call option, priced now, exercised later, with the final payout partly contingent on performance in the interim. Madrid didn’t buy a footballer in 2022. They bought the right to receive one in 2024, at a price that reflected what he was worth as an uncertain teenager rather than what he might become.

Subscribe now

The auction problem

None of this happens in a competitive vacuum, and that’s where a second idea, the winner’s curse, becomes essential. When several wealthy bidders chase the same proven, established star, the winning bid tends to reflect the most optimistic valuation in the room rather than the player’s true worth, because the buyer who overestimates the most is the one who wins the auction.

Neymar’s world record £198 million move to Paris Saint-Germain is the canonical case: a fully known quantity, bid up by a small pool of clubs who could all see exactly what they were buying. Teenagers largely escape this trap, simply because far fewer clubs are watching them closely enough to bid. Premier League clubs spent more than £300 million on teenage signings in the summer of 2024, compared with just £58 million on players in their thirties, according to transfer data compiled that year, and the gap wasn’t really about ability. It was about how many buyers show up to the auction, and how thin that market still is compared with the market for the finished product.

Subscribe now

When the rulebook writes the strategy

There is also a quieter force at work, one written into football’s own financial rulebook. Under Financial Fair Play and its successor, Profit and Sustainability Rules, clubs are allowed to amortise a transfer fee, spreading its cost across the length of a player’s contract for accounting purposes. Chelsea’s ownership, under Todd Boehly and Clearlake Capital, spotted that a longer contract meant a smaller annual hit to the books, and began signing young players to deals stretching seven, eight, even nine years, before regulators capped amortisation at five in 2023.

The incentive structure of the rules themselves had quietly made youth investment the financially rational choice, independent of whether any individual player turned out to be good. It is a reminder that markets don’t just reflect talent. They reflect whatever the rulebook rewards.

Subscribe now

Someone has to grow the asset

Underneath all of this sits a simple division of labour. Clubs like Independiente del Valle, or Portugal’s Sporting and Benfica, or Borussia Dortmund, whose player trading has generated more than £800 million in profit over the past decade according to the CIES Football Observatory, have built entire institutions around originating the underlying asset rather than exercising the option on it.

Big clubs increasingly behave like venture capital, buying equity in potential rather than building the product themselves. It’s a strange thing to admit about a sport built on loyalty and boyhood dreams, but somewhere along the way the industry built a fully functioning derivatives market out of adolescence itself.

Which raises the question nobody in the boardroom seems eager to ask. If a sixteen year old’s future is now priced, structured and traded the way a call option is, at what point does his footballing life stop being a career and start being someone else’s balance sheet entry, and what does it do to a kid to grow up knowing the answer.

This is not an isolated flourish. It is a pattern, and once you see it you cannot unsee it. Abu Dhabi’s Mubadala spent 2025 pouring money into AI infrastructure and digital assets at a pace that outstripped every other sovereign investor on earth. It pumped $4.9 billion into pure AI projects and $12.9 billion into digital assets more broadly last year, making it the world’s most active sovereign wealth fund for the second consecutive year.

Subscribe now

Qatar’s fund has quietly built stakes in European logistics and North American real estate. Kuwait is chasing digital infrastructure with a fervor that would surprise anyone who still pictures Gulf money as a passive parking spot for petrodollars. Together, the seven largest Gulf funds now account for 43 percent of all capital deployed by sovereign investors worldwide, a historical high. Khaleej TimesYahoo!

Saving for a Life After Oil

The easy read is that these countries are simply diversifying, the way any sensible investor with too much exposure to one asset class would. That is true but incomplete. What is actually happening is a textbook, almost pedagogically clean, example of intertemporal optimization, the economic idea that rational actors do not just ask what maximizes wealth today but what maximizes the present value of wealth across their entire time horizon.

For an individual this shows up as choosing to save for retirement instead of spending every paycheck. For a state sitting on a finite pool of crude, it shows up as something far more dramatic: converting a depleting physical asset into a permanent, self-renewing one before the depletion becomes anyone else’s problem to worry about.

Subscribe now

This is really Hotelling’s rule wearing a sovereign wealth fund’s clothing. The economist Harold Hotelling argued back in 1931 that the owner of a non-renewable resource should treat the resource itself as a store of value competing against every other investment available to them. If oil in the ground is expected to keep rising in value net of extraction costs, you leave it in the ground. If financial assets or equity stakes in growth industries offer a better long run return, you extract, sell, and reinvest.

Gulf states have effectively concluded that oil’s future upside is capped, whether by climate policy, by the relentless cost curve of solar and battery storage, or simply by the fact that global demand growth is flattening even as supply from the US, Guyana, and elsewhere expands. Money sitting in the ground earns nothing. Money converted into an equity stake in Anthropic’s competitors, in Newcastle United, in a chain of ports and warehouses, keeps compounding long after the last tanker leaves Ras Tanura.

Racing for a Seat at the AI Table

There is a second layer here, and it involves the kind of game theory usually reserved for arms races rather than sovereign portfolios. AI infrastructure right now looks a lot like a winner take most market, the sort of environment where being an early, well capitalized entrant confers durable advantages through data, talent, and compute access that late movers cannot easily replicate.

Gulf funds are not just diversifying away from oil, they are trying to secure a seat at the table of whatever industry ends up organizing the next fifty years of global growth. Abu Dhabi’s G42, built with Mubadala’s backing, has positioned itself as a genuine AI infrastructure player rather than a mere investor, striking partnerships that put it inside the compute supply chain rather than simply owning shares of it. That is a strategic bet dressed up as a financial one: better to be a stakeholder in the new order than a spectator hoping the old one holds.

Buying Credibility

Then there is the sports and soft power layer, which looks frivolous until you notice the signaling economics underneath it. PIF’s ownership of Newcastle United and its bankrolling of LIV Golf are not primarily about golf or football economics. They are costly signals, in the strict game theoretic sense, meaning they are expensive precisely because that expense makes them credible. Anyone can announce a country is modernizing.

Buying a Premier League club and rebuilding a global golf tour is the kind of investment cheap talk cannot fake, and it reshapes how the rest of the world perceives Saudi Arabia’s trajectory, which in turn lowers the cost of everything else the kingdom wants to do internationally, from attracting foreign direct investment to hosting a World Cup. The signal is the point almost as much as the return.

Hedging the Resource Curse

None of this erases a genuine structural asymmetry that sovereign wealth funds are trying to correct through diversification. Extractive economies are exposed to the classic resource curse dynamics, currency appreciation that strangles other export sectors, price volatility that makes fiscal planning brutal, and a labor market that never quite develops beyond the resource itself.

By recycling oil revenue into equity stakes in AI, logistics, and entertainment, these states are attempting an intergenerational hedge against the very commodity that built their wealth in the first place. It is the fiscal equivalent of an oil executive quietly buying solar stocks, except at a scale of trillions rather than millions. The five largest Gulf funds alone managed roughly $4.1 trillion as of 2023, projected to climb toward $7.6 trillion by 2030, which gives some sense of just how much capital is being rerouted through this logic every year. Chronograph

Subscribe now

What the Insiders Know

What makes this genuinely fascinating rather than merely large is the information asymmetry embedded in the whole exercise. These funds know more about their own long run oil price expectations than almost anyone else on earth, because they set production policy and see the reserve data up close.

When PIF or Mubadala moves aggressively into AI or logistics while simultaneously managing OPEC Plus supply decisions, they are effectively trading on a private signal about where they think fossil fuel economics are heading, one the rest of the market can only infer after the fact from where the money goes.

So here is the uncomfortable question worth sitting with. If the entities with the best information about oil’s future are the ones moving fastest to build a post-oil portfolio, what exactly does that tell the rest of us about how much time the current energy order actually has left?

Moneyball turns twenty-three this year, and the joke inside football recruitment circles is that most clubs are still catching up to a baseball book from 2003. But the number that actually stops people is this one: when Liverpool sold Philippe Coutinho to Barcelona for a reported 142 million euros in January 2018, the club’s data science team, led by a Cambridge-trained physicist named Ian Graham, had already flagged his eventual replacements before the transfer window even opened.

Mohamed Salah cost 42 million. Sadio Mane cost 34 million. Together they cost less than a third of what Coutinho fetched, and they helped Liverpool win a Champions League and a Premier League title within two years. Graham’s department at Liverpool was, by most accounts, a handful of people with laptops. It punched vastly above its payroll.

Subscribe now

This is the puzzle worth sitting with. Sports teams operate in markets where a single elite player can transform a season, yet the smartest organizations increasingly spend their marginal dollar not on that player but on the person who can tell them which player, which formation, which contract structure actually deserves the money.

Manchester City’s analytics staff has grown from a handful of people a decade ago to dozens today. The Golden State Warriors built a front office around player-tracking data before most rivals knew what SportVU cameras were. The economics behind this shift is not really about basketball or football at all. It is about where returns hide when everyone is chasing the same visible prize.

The diminishing returns of stardom

Start with the idea of diminishing marginal returns, which sounds dry until you apply it to a transfer fee. A club with a mediocre midfield gets enormous value from signing one good midfielder. But once a squad is already stacked with talent, each additional star contributes less incremental improvement while costing proportionally more, because elite players are auctioned in a market where every rival club can see the same statistics you can.

Real Madrid’s Galactico strategy in the early 2000s, when the club spent lavishly on Figo, Zidane, Ronaldo, and Beckham while quietly neglecting its defensive structure and academy pipeline, produced spectacular individual talent and a surprisingly mediocre run of results relative to the wage bill. The marginal superstar was expensive and the marginal minute of tactical coherence was cheap, and the club got the ratio backwards for years.

Subscribe now

Where the real information asymmetry lives

That mispricing is exactly where information economics enters. Talent is a relatively observable good. Everyone with a television and an opinion can see that a striker scores goals. Decision quality, on the other hand, is close to invisible in real time. Nobody claps when a recruitment analyst correctly predicts that a 22 year old playing in Belgium’s second division will outperform a proven striker from a bigger league, not because the model is flashy but because it corrects for the fact that goals scored in a weaker league systematically overstate ability once you control for opposition strength.

This is the essence of information asymmetry working in reverse of how it usually gets taught. Instead of the seller knowing more than the buyer, in transfer markets the selling club and its buyer’s rivals often share almost identical information, all scouted from the same match footage and the same underlying statistics. The team that wins is the one that processes that shared information better, not the one that has access to secret information nobody else has.

The winner’s curse in the transfer window

Which brings in auction theory, because a transfer window is, structurally, an auction, and auctions have a well known pathology called the winner’s curse. When several bidders compete for the same asset under uncertainty about its true value, the winning bid tends to be the one made by whoever was most optimistic, which means winning bidders systematically overpay relative to the asset’s actual worth. This is precisely why clubs like Brentford, promoted to the Premier League in 2021 under owner Matthew Benham, built their entire recruitment philosophy around avoiding bidding wars altogether.

Benham, who made his fortune in sports betting, applies expected value modeling to recruitment the way a professional gambler prices a football match, looking for players whose statistical output suggests they are undervalued by the wider market’s consensus, then buying before that consensus catches up. Brentford’s wage bill in the 2022-23 season was among the lowest in the league, yet they finished ninth, comfortably above several clubs spending three or four times as much. They are not smarter about football in some mystical sense. They are better at not participating in auctions they are likely to lose on price.

Subscribe now

Cricket’s transparent auction and the escalation problem

The same logic runs through the Indian Premier League, where cricket franchises now employ full analytics units to bid in an auction system that is, unusually, transparent and public, with every franchise watching every other franchise’s bids in real time. This transparency should, in theory, eliminate mispricing, yet the IPL still produces wild valuation swings, because psychological momentum inside a live auction room creates a different distortion entirely.

In game theory terms, this is closer to a coordination problem than a pure information problem. When Mumbai Indians and Chennai Super Kings both chase the same all-rounder, prices escalate not because either franchise has revised its estimate of the player’s worth but because bidding becomes a signal of resolve, a phenomenon economists studying escalation of commitment have documented for decades in contexts ranging from corporate takeovers to actual wars.

Subscribe now

The 2022 IPL mega auction saw several players sold for figures well beyond what pre-auction analytical models had projected, and post-season performance data suggested a real gap between auction price and output, the kind of gap a good data team exists specifically to exploit the following year.

Analysts as a check on the principal-agent problem

There is also a quieter principal-agent problem buried inside all of this. Coaches and managers are judged, fairly or not, on short-term results, which creates an incentive to favor familiar, high-status signings that reassure fans and boards even when the numbers suggest a cheaper alternative would perform just as well. An analytics department, positioned slightly outside that pressure, can push decisions back toward what the data actually supports rather than what looks impressive on a press release.

This is uncomfortable inside a sport built on romance and reputation, which is exactly why so many clubs resisted it for so long and why the ones who embraced it early captured most of the advantage before their rivals adjusted.

None of this means analysts will replace players, any more than better auditors replaced factories. What is changing is where the scarce resource actually sits. For decades that scarce resource was talent itself, visible, measurable in goals and points, and fought over accordingly. Now the scarcer resource is the judgment required to price that talent correctly before everyone else does. The Coutinho trade worked because Liverpool understood something obvious in hindsight and invisible in the moment, that the club with the best model of the market beats the club with the biggest checkbook.

The harder question is what happens once every team has that model. If information advantage is the new source of edge, and information tends to diffuse, what replaces analytics as the next undervalued asset once everyone has a laptop and a data scientist sitting where the star striker used to sit?

Sarah Kunst tells a story that’s become a kind of parable in tech circles. An engineer at a major software company decamped from San Francisco to Boise during the pandemic, expecting nothing but freedom and a lower mortgage. Instead, HR called. His salary was being cut by eighteen percent, effective the following quarter. Same code, same manager, same laptop. Different zip code, different price.

This is not a glitch in the system. It is the system working exactly as designed, and it reveals something uncomfortable about the so called global labor market that remote work was supposed to deliver. We were promised that talent would finally be priced like a commodity, floating freely to wherever it could earn the most. Instead, five years into the remote work experiment, most companies have built elaborate machinery to make sure your address still determines your paycheck almost as much as your résumé does.

Google’s internal relocation calculator can show an employee a pay cut of up to twenty five percent for moving away from a Tier 1 metro. Meta has been explicit since Mark Zuckerberg first addressed the policy that pay tracks the local labor market, not the employee’s output. Stripe went further and turned the logic into a public spectacle, offering remote employees twenty thousand dollars to leave the Bay Area while quietly attaching a permanent ten percent cut to their base salary. By 2025, roughly two thirds of fully remote companies had adopted some version of location based pay, up sharply from just over four in ten in 2022.

Why density pays

The economics behind this starts with agglomeration, an idea Alfred Marshall sketched over a century ago and that economists like Edward Glaeser and Enrico Moretti have spent careers proving with data. Cities do not just contain workers. They generate productivity that individual workers could never produce alone, through three linked mechanisms. Dense labor markets let firms find precisely the right specialist and let workers find precisely the right job, cutting search cost.

Subscribe now

Proximity lets knowledge leak between firms almost accidentally, over lunch, at a conference, when a friend at a rival company mentions a technique that reshapes how you think about a problem. And shared infrastructure, from fiber networks to specialized law firms to venture capital offices, lowers the cost of doing business for everyone clustered nearby. None of this shows up on an individual’s résumé. It shows up in the wage.

Moretti’s research found that a college educated worker in a dense innovation hub does not just earn more because they are surrounded by other high earners bidding up rents. They are measurably more productive, because the ecosystem itself does work that no single employer is paying for directly. Remote work optimists imagined agglomeration as a coordination problem, solvable with a video call. It is closer to a production function, one that depends on unplanned, undocumented interaction that a scheduled Zoom meeting structurally cannot replicate.

The matching machine

There is also a matching problem hiding underneath the geography question, and it belongs to search theory rather than agglomeration proper. A thick labor market, one with many buyers and many sellers physically close together, resolves mismatches faster and more precisely than a thin one. If you are a machine learning engineer with a narrow specialty in reinforcement learning for robotics, the Bay Area does not just pay you more because rents are high.

It pays you more because enough companies within a fifteen mile radius do exactly that work that both sides can find a near perfect match quickly, and near perfect matches are worth paying a premium to secure. Move that same engineer to a metro with one relevant employer, and the wage should, in theory, be discoverable through a screen. In practice it collapses, because the outside option that disciplined the negotiation has vanished.

Subscribe now

Pricing the information gap

Information asymmetry compounds this. When an employer sets a location adjusted salary, it is making a bet about your local cost of living and local market rate using aggregated data, not your actual circumstances. The employer knows its full compensation philosophy. You typically see one number.

Companies that have gone the other direction, like GitLab, have tried to solve this by publishing their entire location factor formula, built from Radford benchmarking data cross referenced against Numbeo cost of living indices, producing more than ten distinct global pay zones. Transparency does not eliminate the geographic wage gap. It just makes the mechanism legible, which is its own quiet admission that geography, not skill, is doing most of the pricing work.

Bengaluru’s own geography

The pattern holds well outside Silicon Valley, and looking at India makes the mechanism even sharper because the labor market there scaled up almost entirely after broadband made remote coordination trivial. You would expect geography to matter least in a market built on the assumption of distributed work. Instead, Bengaluru software engineers now earn something like thirty five to fifty percent more than counterparts with equivalent experience in Tier 2 Indian cities such as Jaipur or Coimbatore, according to 2025 and 2026 compensation surveys from Michael Page and Deloitte with Nasscom.

The premium exists because Bengaluru concentrates the global capability centers, the funded startups, and the specialized recruiters that make a thick, fast moving market possible. An engineer in a smaller city with identical certifications simply cannot access that density, and the market prices the difference ruthlessly.

The holdouts

A handful of companies have tried to prove the whole framework wrong. Airbnb scrapped US geographic pay tiers entirely under its Live and Work Anywhere policy. Basecamp goes further still, paying every employee at the ninetieth percentile of San Francisco rates no matter where they live, treating location adjustment as a form of unjustified price discrimination rather than sound economics. These experiments are real, and they are still the exception rather than the emerging norm, which itself tells you something about how expensive it is to fight agglomeration effects rather than simply price around them.

What remote work actually did was unbundle geography from labor supply while leaving it firmly attached to labor demand. You can live anywhere. The value your specific combination of skills creates in a specific market, next to a specific density of collaborators and competitors and customers, still cannot be shipped over a fiber line. The next serious test of this arrangement will not come from more remote workers moving to Boise. It will come from cities that manage to digitize agglomeration itself, building dense virtual labor markets with the same matching speed and knowledge spillover that physical proximity has monopolized for a century. Nobody has built that yet. The question worth sitting with is whether it is even possible, or whether the thing that makes a city valuable is precisely the thing no network can replicate.

Tokyo never forgot the feeling. Over the following decade, Japan poured money into rare earth recycling, financed a mine in Australia through trading house Sojitz, and built stockpiles that officials still refuse to fully disclose. None of it made economic sense on a spreadsheet. All of it made sense once you stopped treating factories as profit centers and started treating them as insurance policies.

A machine that keeps running at a loss

That distinction is the key to understanding something that looks bizarre from the outside: China’s own industrial machine keeps expanding even as prices collapse and factories bleed money. Solar panel makers Longi, JA Solar and Tongwei together lost close to three billion dollars in the first half of 2025 alone, and module prices sank so low that Chinese-made panels were routinely selling for seven to nine cents a watt, below what it cost to produce them.

China’s national capacity utilization rate, a rough measure of how much of its industrial machinery is actually running productively, slid from 76.2 percent in the final quarter of 2024 to under 74 percent by the middle of 2025, its lowest reading since the pandemic. By any normal accounting, this is a country building factories nobody needs to make products nobody can profitably sell. Economists call it overcapacity. The more useful question is overcapacity relative to what.

Subscribe now

Trade theory for industries where size wins

Ordinary trade theory says a country should specialize in what it does best and let comparative advantage sort out the rest. If Vietnam can sew a shirt more cheaply, buy the shirt from Vietnam. But strategic trade theory, developed in the 1980s by economists like Paul Krugman and Barbara Spence to explain why governments subsidize aircraft makers and semiconductor firms, argues that in industries with huge fixed costs, steep learning curves, and only a handful of viable global players, the first mover captures returns that never show up in a simple cost comparison.

Whoever controls the factory controls the knowledge embedded in running it, the supplier networks around it, and eventually the pricing power that comes from being indispensable. China’s leadership reads solar, batteries, shipbuilding, and rare earth processing through exactly this lens. A factory that loses money is still a factory that keeps the workforce skilled, the supply chain intact, and the off switch in Chinese hands rather than someone else’s.

Capacity as an option you hope never to exercise

This only becomes rational once you fold in a second idea: capacity as a real option rather than a static asset. In finance, an option is valuable even if you never exercise it, because it gives you the right to act later under conditions you can’t fully predict today. A shipyard that can build both commercial tankers and naval vessels is worth more to the state than its order book suggests, because it preserves the ability to pivot the moment geopolitics demands it. South Korea understood this instinct from the other direction.

When Japan tightened export controls on photoresist and hydrogen fluoride in 2019 after a wartime labor dispute, Samsung and SK Hynix suddenly discovered how exposed their chip fabs were to a single foreign supplier. Seoul responded not by shrinking its semiconductor ambitions but by expanding domestic materials production and diversifying suppliers, accepting years of higher costs to buy back strategic autonomy. Both Beijing and Seoul were pricing in a scenario, not a spreadsheet.

Subscribe now

Why no province wants to be the one that blinks

There’s a game theory layer sitting underneath all of this too, and it looks a lot like a multiplayer prisoner’s dilemma played out across provinces. Local Chinese governments compete for investment, jobs, and tax revenue, and each has every incentive to subsidize land, electricity, and credit for a new battery or solar plant even when national demand is already saturated. No single province wants to be the one that stands down while its neighbor keeps building, because standing down means losing the factory, the jobs, and the political credit that comes with them.

The result is a coordination failure that produces exactly the kind of collective overbuilding you’d predict from a group of players who cannot credibly commit to restraint. Beijing’s own Ministry of Industry and Information Technology now openly complains about what it calls involutionary competition, a domestic acknowledgment that decentralized incentives have generated more capacity than any central planner would have chosen alone.

Who actually pays for the losses

None of this would be financially sustainable without a fourth piece of the puzzle: the soft budget constraint. State-linked banks in China have historically extended credit to strategic industries on terms that would bankrupt a private lender’s balance sheet, because the loans are ultimately backstopped by the state’s tolerance for losses in pursuit of a policy goal. A private American or European firm facing Longi’s margins would have shut down production lines years ago.

A firm operating inside China’s industrial policy apparatus can keep running at a loss because the relevant cost being managed isn’t quarterly profit, it’s the externality of falling behind in a technology the state considers non-negotiable. Externalities usually refer to pollution or congestion, costs that spill onto people who didn’t choose them. Here the externality runs the other way: the benefit of maintained industrial capability spills onto the whole country’s strategic position, and no individual factory owner would rationally pay for that benefit out of their own pocket, so the state pays for it instead.

Subscribe now

A world learning the same lesson twice

What makes this moment different from previous industrial gluts is that everyone watching now understands the game being played. The European Union has opened investigations into Chinese electric vehicle subsidies. Washington has restricted advanced chip equipment sales in the other direction, essentially running its own version of the Japan to China rare earth playbook against Beijing. Both sides are learning the same lesson Japan learned in 2010, that specialization is only a virtue until the supplier decides it’s a weapon.

The genuinely uncomfortable question isn’t whether China is behaving irrationally by building factories it doesn’t need. It’s whether every country now racing to duplicate that capacity, in batteries, in chips, in shipbuilding, is quietly signing up for a world with far more redundant factories and far less of the cheap efficiency that made globalization feel like a free lunch in the first place. Somebody has to pay for all that insurance. The only question left is whose consumers end up holding the bill.

The surge came in the aftermath of tension in the Middle East, when traders scrambled for alternatives to dollar-clearing routes that suddenly felt politically exposed. And yet, in the very same quarter, the US dollar’s share of global central bank reserves actually rose, ticking up to 57.13 percent according to the IMF’s latest COFER data. Two headlines, both true, both pointing in opposite directions. This is not a contradiction. It is a clue.

The Succession Story We Keep Telling

For decades the conversation about the dollar’s future has been framed as a succession story. Sterling gave way to the dollar after the Second World War, so surely something will eventually give way to the dollar too, whether the euro, the yuan, or some digital successor nobody has invented yet. This framing borrows, often unconsciously, from a very old idea in market design: that markets with strong network effects tend toward a single winner.

The more people who use a currency, the more useful it becomes to everyone else who might use it, which pulls in still more users, until one currency captures almost the entire market the way one keyboard layout or one social network eventually does. Economists have a term for this kind of self-reinforcing equilibrium, borrowed from Thomas Schelling. A dominant reserve currency is a focal point, a solution that everyone converges on not because it is objectively best but because everyone expects everyone else to converge on it too.

Money Is Not One Market

The trouble is that money is not actually one market. It is at least three, and they behave very differently. There is the market for holding wealth, which is what COFER measures. There is the market for invoicing trade, which is what determines the currency your export contract is written in. And there is the market for clearing and settling the actual payment, the plumbing that moves the money once the invoice is due. A currency can dominate one of these markets while losing ground in another, and that is precisely what the data now shows.

The dollar’s share of global foreign exchange turnover sat at 89 percent as of the Bank for International Settlements’ 2025 survey, a level that has barely budged in twenty years. Its share of reserves, at 57 percent, has been sliding gently since the early 2000s, when it stood above 70 percent. And in the narrower world of cross-border payment messaging, the yuan is growing fast from a tiny base, moving past 3 percent of SWIFT volumes while CIPS itself processed roughly 180 trillion yuan, about 25 trillion dollars, over the course of 2025. These are not three data points on the same trend line.

Subscribe now

They are three separate markets, each with its own switching costs, its own incumbents, and its own logic for why change happens slowly in one and quickly in another.

The Rupee Dirham Experiment

This is where the idea of specialized payment networks becomes useful, and where India’s recent experiment with the UAE offers a small but telling case study. In 2023 the Reserve Bank of India and the Central Bank of the UAE signed a memorandum to settle bilateral trade directly in rupees and dirhams, bypassing the dollar as an intermediary currency entirely. By early 2025 the RBI had gone further, amending its foreign exchange rules to let Indian exporters hold foreign currency accounts abroad and receive advance payments more flexibly, explicitly to make rupee invoicing easier.

Indian refiners buying discounted Russian crude through Dubai-based traders had already begun settling in dirhams rather than dollars. None of this threatens the dollar’s role as the world’s reserve asset. Central banks are not about to swap their Treasury holdings for rupees. But it chips away, corridor by corridor, at the dollar’s role as the default settlement currency for a specific bilateral trade relationship, particularly one built around energy, where sanctions risk has made dollar clearing feel less like a convenience and more like a liability.

Slow Trust, Fast Corridors

The behavioral economics here matters as much as the plumbing. Holding reserves is a decision made under radical uncertainty about events decades away, so central bankers default to whatever currency feels safest and most liquid, which is a textbook description of an availability heuristic reinforced by decades of habit. Settling a specific trade invoice is a much narrower, much more immediate decision, closer to a repeated game between two counterparties who already trust each other and want to cut transaction costs and currency risk on a deal they do business on every month. It should not surprise us that the second kind of decision is easier to change than the first.

The mBridge project, a multi-currency digital settlement platform backed initially by the Bank for International Settlements and now carried forward mainly by China and the UAE, has processed only about 55 billion dollars in transactions since launch, with the digital yuan accounting for roughly 95 percent of that flow. It is a rounding error next to global trade volumes, but it is also a live demonstration that a wholesale settlement corridor can exist entirely outside the dollar system, purpose built for a narrow use case, without needing to dethrone anything.

An Archipelago, Not an Empire

What emerges from all this is less a race with a single finish line and more a landscape of overlapping, purpose-specific standards, not unlike how the world settled on different voltages, different rail gauges, or different mobile payment standards depending on which region built which infrastructure first and who its trading partners happened to be. The dollar keeps its grip on the deepest, most liquid capital markets and on the reserve function, because information asymmetry and crisis-driven flight to safety both favor the most trusted, most established asset, and trust of that kind is brutally slow to rebuild once broken.

Subscribe now

The yuan carves out dominance in the specific trade corridors that run through China’s own supply chains and its energy relationships with sanctioned or semi-sanctioned partners. The rupee, the dirham, and a growing list of local currency arrangements chip away at dollar intermediation in bilateral trade between developing economies that increasingly trade more with each other than with the West. None of these networks needs to defeat the others. They just need to be good enough for the transaction at hand.

So perhaps the right question is not which currency replaces the dollar, but what kind of world we get once money itself stops behaving like a single market with one winner and starts behaving like an archipelago of parallel systems, each optimized for a different kind of trust. What happens to a country’s leverage when the currency it can weaponize covers less and less of the map every year, not because it lost a war for dominance, but because the game itself quietly stopped being winner take all?

That small, almost comic relocation is a preview of something much larger that has been happening ever since. Every time the United States and China raise a tariff wall between them, the wall does not stop trade. It reroutes it. And the countries that end up catching the overflow are not always the ones you would expect.

The Logic of the Workaround

The basic economic idea here is trade diversion, a concept economists have understood since the 1950s, when Jacob Viner first showed that tariffs and trade preferences do not simply reduce trade, they redirect it toward whichever supplier can now clear the new price bar most cheaply. Diversion is different from destruction. A tariff on Chinese solar panels does not mean Americans stop buying solar panels. It means they buy panels assembled in Malaysia using Chinese components, because Malaysia is not (yet) tariffed and the finished product from there still clears customs at a price American installers can live with.

Subscribe now

What makes this dynamic so durable is that global manufacturing was never really organized by country. It was organized by task. A shirt might be designed in New York, its fabric woven in Guangdong, and its final stitching done in Bangladesh. Economists call this the fragmentation of production, and it means that when a tariff hits one node in the chain, firms do not need to rebuild the whole chain, only swap out that one node. This is why trade diversion happens so fast. It is not a strategic pivot. It is water finding a new crack in a wall.

The numbers make the scale of this hard to dismiss as a rounding error. U.S. imports from China declined sharply after the 2018 and 2019 tariffs, while U.S. imports from Vietnam roughly tripled by 2025. Vietnam’s exports to the United States climbed from about $83 billion in 2020 to $153.2 billion in 2025, a jump of 28 percent in a single year. Mexico’s story is even more striking, because it did not just grow, it took the crown. In 2023, Mexico dethroned China as the top exporter to the United States for the first time in over a decade, and it has held that position every year since.

When the Winner Is Also the Loophole

Here is where the story gets economically interesting rather than just statistically impressive. A striking share of what looks like Vietnamese or Mexican manufacturing success is not really a new industry being born. It is Chinese production wearing a different passport.

Federal Reserve researchers looking at Vietnamese firm-level data this year found something that trade economists had suspected but struggled to prove with hard numbers: a meaningful chunk of Vietnam’s export boom to the United States is driven by Chinese-owned firms that relocated or expanded operations inside Vietnam specifically to sidestep the tariff wall, not by homegrown Vietnamese manufacturers building new capability from scratch. The correlation is almost too clean to be coincidental. Between 2018 and 2024,

Subscribe now

China’s exports of intermediate goods to Vietnam grew at a compound annual rate of 18.7 percent, almost exactly tracking the 15.7 percent annual growth in Vietnam’s finished-goods exports to the United States over the same stretch. Vietnam is not replacing China in the supply chain. In many cases it is standing in front of China, absorbing the components and re-exporting the finished product under a new country-of-origin label.

This is a textbook case of regulatory arbitrage, the practice of restructuring an activity not to change its substance but to change which rulebook it falls under. The tariff is written to apply to goods “made in China.” It says nothing about goods substantially finished in a third country using Chinese inputs, so firms relocate the final stage of production across a border and change the paperwork. Washington has noticed. Vietnam’s own trade surplus with the U.S. grew so large, from $39.5 billion in 2018 to roughly $123.5 billion in 2024, that it triggered a fresh 20 percent U.S. tariff on Vietnamese goods in August 2025, with an even steeper 40 percent levy targeting suspected transshipment. The loophole and the patch for the loophole are now locked in an escalating game of regulatory whack-a-mole.

Why Some Countries Win the Reshuffle and Others Don’t

Not every developing economy gets to be the next Vietnam, and the reason comes down to something economists call agglomeration economics, the tendency for industries to cluster in places that already have the right ports, the right suppliers, the right skilled labor pool, and the right regulatory muscle memory. Trade diversion does not flow evenly across the developing world like rain. It pools in specific places that had already built the infrastructure to catch it.

Mexico is the clearest illustration. Its win was not purely about dodging tariffs, since Mexican goods have enjoyed duty-free access to the U.S. market since NAFTA took effect in 1994. What changed was that geographic proximity and just-in-time supply chains suddenly became a competitive advantage rather than a nice-to-have, once firms burned by pandemic-era shipping delays started pricing in the risk of a twelve-thousand-mile supply chain, not just its sticker cost. Bilateral manufacturing trade between Mexico and the U.S. now represents about 16.5 percent of all U.S. manufacturing trade, ahead of both Canada and China. Tesla’s decision to build a plant in Monterrey rather than expand elsewhere was a bet that globalization’s logic was shifting, from cheapest possible labor to closest reliable neighbor, a shift economists now call friend-shoring.

India offers a quieter, less examined version of the same story, one that rarely gets the magazine-cover treatment Vietnam gets. Apple’s decision to ramp up iPhone assembly in Tamil Nadu and Karnataka was driven only partly by tariffs. It was also a hedge against what economists call concentration risk, the danger of having your entire global supply chain dependent on a single country’s political stability. A tariff war between two superpowers is, from a factory owner’s chair, indistinguishable from any other kind of geopolitical risk. You diversify against it the same way you would diversify against a currency crash or a civil war, by not keeping all your production in one place.

The Part Nobody Put in the Treaty

There is a deeper irony sitting underneath all of this, and it is one that trade negotiators rarely say out loud. Tariffs are usually framed domestically as tools to bring jobs home, to punish a rival, to rebalance a bilateral relationship that feels unfair. But because of trade diversion, that rarely happens. The jobs mostly do not come home. They go to Hanoi, or Monterrey, or Bengaluru. The bilateral deficit narrows, sometimes even reverses, and Washington can claim a symbolic win. Meanwhile a nearly identical volume of goods, often built from nearly identical Chinese components, keeps arriving on American shelves, just with a different stamp on the box.

This is not a flaw in how tariffs work. It is closer to a law of how they work, given how modular and mobile modern manufacturing has become. Which raises an uncomfortable question for the next round of trade brinkmanship, whenever and wherever it happens next. If every act of economic nationalism just relocates globalization rather than reversing it, is anyone actually getting what they wanted? Or has trade policy quietly become a subsidy program for whichever country happens to have the spare industrial capacity, the open port, and the good timing to catch the overflow?

For the past three years, the popular narrative around artificial intelligence has centered on models, on which lab has the smartest one, which country has the most talented researchers, whose transformer architecture will win. That framing made sense when AI progress was bottlenecked by ideas. It no longer does. The frontier has moved from cleverness to capacity, and capacity is a story economists have told before, just never quite at this scale or speed.

Compute as a Heckscher-Ohlin good

Start with factor endowment theory, the old workhorse of international trade economics. The basic insight, going back to Eli Heckscher and Bertil Ohlin in the early twentieth century, is that countries export goods that intensively use the factors of production they have in abundance. A country rich in land grows wheat. A country rich in cheap labor stitches garments. It sounds almost too simple to be useful, until you realize AI has quietly become a Heckscher-Ohlin good. The scarce factors now are electricity, land for data centers, water for cooling, and advanced semiconductor fabrication capacity. Algorithms, by contrast, diffuse fast. Weights leak, papers get published, open source models close the gap within months. What does not diffuse quickly is a gigawatt of reliable power or a functioning fab.

Saudi Arabia and the UAE understood this before most Western commentators did. Neither country has a deep bench of machine learning PhDs. What they have is a sovereign wealth fund with roughly a trillion dollars in the case of Saudi Arabia’s Public Investment Fund, some of the cheapest solar electricity contracts ever signed anywhere in the world, and land.

Subscribe now

Saudi Arabia’s National Renewable Energy Programme produced a solar tariff of just over one cent per kilowatt hour in its Al Shuaibah auction, among the lowest ever recorded globally. That is not a coincidence of geography turned into a coincidence of strategy. It is comparative advantage, David Ricardo’s two-century-old idea, applied to compute the way it used to be applied to wine and cloth.

Taiwan and the economics of agglomeration

This is where a second concept becomes essential, agglomeration economics, the idea that once an industry clusters somewhere, the cluster becomes self-reinforcing through shared suppliers, specialized labor, and accumulated know-how. Taiwan is the textbook case, and the numbers are almost uncomfortable to sit with. TSMC alone now accounts for something like ninety percent of the world’s most advanced chip fabrication, the kind that goes into the GPUs training frontier models.

Subscribe now

The company did not get there by having the best engineers on day one in 1987. It got there because Morris Chang built a pure-play foundry model inside the Hsinchu Science Park, which then pulled in Taiwanese engineers trained abroad, then pulled in suppliers, then pulled in more capital, in a feedback loop that has now run for nearly four decades.

The result is a level of concentration that would be considered dangerously monopolistic in almost any other strategic industry, and yet no government, including the United States, has managed to meaningfully replicate it despite billions in subsidy. That is what agglomeration does. It does not just create an advantage. It creates a moat that ordinary market entry cannot cross, because the thing being sold is not silicon, it is decades of accumulated tacit knowledge distributed across a dense local supply chain.

Chips as chokepoints

The third piece of the puzzle is what trade economists call a chokepoint, and what policy people are now calling compute diplomacy. When a critical resource is concentrated in a small number of hands, control over its distribution becomes a foreign policy lever, not just a commercial one. The United States Commerce Department’s decision in November 2025 to authorize the export of the equivalent of 35,000 Nvidia Blackwell chips each to Saudi Arabia’s Humain and the UAE’s G42 was not a routine trade approval.

It was a calculated exercise of chokepoint leverage, granted in exchange for hundreds of billions of dollars in reciprocal investment commitments and explicit security conditions on how those chips could be used and by whom.

Subscribe now

This mirrors, almost exactly, the logic OPEC nations understood about oil in the 1970s, except now the resource being rationed is not extracted from the ground but manufactured on a single island within missile range of a country that has never renounced the option of taking it by force.

The grid as a natural monopoly

Underneath all of this sits a fourth and perhaps less glamorous force, the economics of the electrical grid as a natural monopoly with genuinely brutal physical constraints. Unlike software, a grid cannot be scaled with a pull request. The IEA reported that global data center electricity demand grew seventeen percent in 2025, with AI-specific facilities growing fifty percent in that single year. Goldman Sachs now estimates the United States faces a structural power shortfall of roughly nine gigawatts in 2026, a gap it forecasts will widen to forty five gigawatts by 2028.

That is a shortfall equivalent to the annual electricity needs of tens of millions of American households, and it cannot be closed by writing better training code. It requires turbines, transmission lines, permitting reform, and in some cases nuclear plants that take a decade to build. Suddenly the country with the friendliest grid interconnection queue has a real edge over the country with the smarter research lab, which is a fairly strange sentence to have to write about the most advanced technology humanity has built.

Who actually owns the grid

Put these four forces together and the emerging map of AI power starts to look less like a leaderboard of research papers and more like a leaderboard of industrial policy from a century ago, rebuilt for electrons and wafers instead of coal and steel. The nations treating this as infrastructure economics, securing energy contracts, courting foundries, and trading chip access for capital, are the ones actually shaping where the next decade of AI capability gets built.

The ones still treating it as a talent competition may find themselves owning the smartest models running on someone else’s grid. The real question worth sitting with is not which lab releases the next great model. It is which government understood, early enough, that AI leadership was never really a technology race at all.

Six million units sold in the console’s first month on shelves, a number that would make most electronics manufacturers giddy, and yet the dominant consumer experience of the Switch 2 was not delight but frustration. Empty shelves. Refresh, refresh, refresh. One American retailer reportedly filled its bare Switch 2 display with old Wii U consoles just to have something to show customers.

Subscribe now

Here is the strange part. Nintendo has been through this before, more than once, and has never seemed particularly desperate to fix it.

A signal disguised as a shortage

The obvious explanation is supply chain failure, and there is truth in that. Component shortages, especially in the kind of high performance memory now being gobbled up by AI data centers, are real constraints. But treat that as the whole story and you miss something Nintendo itself seems to understand better than almost any other consumer hardware company on earth: scarcity is not always a problem to be solved. Sometimes it is a feature that makes the product more valuable than abundance ever could.

Subscribe now

Start with signaling theory. In a world where you cannot directly verify the quality or desirability of a product before buying it, people look for indirect signals. Difficulty of acquisition is one of the strongest signals available. If something is hard to get, buyers reason, other people must want it badly, which means it is probably worth wanting. This is the same logic economist Thorstein Veblen described more than a century ago when he wrote about conspicuous consumption, though Veblen was mostly talking about price as the signal.

Nintendo has found a subtler version. The Switch 2’s price tag alone did not create hype. What amplified the hype was watching it vanish from shelves within minutes, watching restock trackers crash under traffic, watching strangers on forums compare notes about which retailer’s app had briefly shown a green “add to cart” button. Every sold out sign became free advertising, proof by scarcity that the thing behind the counter mattered.

Losing something you never had

Behavioral economics offers a second, complementary explanation, one rooted in loss aversion. Daniel Kahneman and Amos Tversky’s foundational insight was that people feel the pain of losing something roughly twice as intensely as they feel the pleasure of gaining an equivalent thing. Scarcity does something clever to this asymmetry. It reframes a purchase you have not yet made as a loss you are actively suffering.

You are not simply failing to acquire a Switch 2, you are losing the Switch 2, over and over, every time a restock disappears before you can click through. That reframing is psychologically potent in a way that ordinary marketing rarely achieves. Compare this to Sony’s approach with the PlayStation 5, which suffered brutal shortages from 2020 through early 2023 largely against its own wishes.

Sony wanted supply to catch demand as fast as possible and treated every empty shelf as pure lost revenue. Nintendo, by contrast, has a long institutional memory of scarcity working in its favor, and that memory shapes how urgently it chases full supply.

A company that remembers

You can trace this pattern back decades. When Nintendo launched the NES in the United States in 1985, the video game market was still recovering from the catastrophic crash of 1983, and retailers were deeply skeptical of stocking any game console at all. Nintendo deliberately limited initial shipments even as demand grew, partly out of genuine caution but also because a slow, tight rollout built retailer confidence that this was a product worth fighting for shelf space to carry.

The original Wii in 2006 sold out for the better part of a year, with resale prices on eBay routinely doubling retail, and Nintendo’s supply never quite seemed to catch up until well after the frenzy had done its work of turning the Wii into a cultural phenomenon rather than merely a bestselling console. The original Switch in 2017 followed the same script.

Who gets to wait

There is a market design dimension here too, and it is where the Japan lottery example becomes genuinely instructive rather than just colorful. When demand vastly exceeds supply at a fixed price, you get rationing by some mechanism whether you plan for it or not. The default mechanism, in the West at least, tends to be rationing by speed and luck, whoever refreshes fastest or happens to be online at the right second, which functionally advantages resellers running bots over ordinary buyers.

Nintendo’s lottery system in Japan is a more deliberate piece of market design, an attempt to ration a scarce good through a mechanism that at least nominally treats every entrant with equal probability rather than rewarding whoever has the fastest script. It does not eliminate scarcity. It just changes who scarcity burdens, converting a speed contest into something closer to a raffle. Economists who study market design, the same field that gave us kidney exchange algorithms and school choice systems, would recognize this instantly as an allocation problem dressed up in retail clothing.

The next sold out sign

None of this means Nintendo enjoys watching its own president apologize to frustrated customers, or that executives are cackling behind the scenes every time a restock sells out in ninety seconds. Real supply constraints are real, and a company forgoing billions in sales it could easily make is not costless, whatever the branding upside. But it is worth asking why Nintendo, alone among the major console makers, seems so consistently unbothered by the appearance of struggling to keep up.

Xbox and PlayStation have spent years chasing supply as aggressively as engineering allows, treating every unsold console as pure waste. Nintendo has quietly let a few extra months of scarcity run their course, watched resale prices climb, watched forums fill with tales of near misses, and let all of that anxiety do marketing work no advertising budget could buy.

The next time you see a sold out sign, on a console, a sneaker, a limited drop of anything, it is worth asking a different question than the obvious one. Not just why is this scarce, but who benefits from you believing it has to stay that way.

What happened next had almost nothing to do with Neymar’s talent and everything to do with a quirk of human cognition that behavioral economists have been documenting since the 1970s. Once a number that large enters a market, it doesn’t just describe one transaction. It becomes a reference point that every subsequent negotiation gets measured against, whether or not anyone involved thinks the comparison makes sense.

The anchor drops

Anchoring is one of the most robust findings in behavioral economics. Daniel Kahneman and Amos Tversky demonstrated it with a rigged wheel of fortune, showing that people asked to estimate the percentage of African countries in the United Nations gave wildly different answers depending on a completely arbitrary number spun on a wheel moments earlier. The number had no informational content whatsoever, yet it dragged people’s estimates toward it like gravity. Once you’ve been shown a big number, every subsequent number gets judged relative to that one, not against some independent standard of what’s actually reasonable.

Football’s transfer market runs on exactly this mechanism, except the stakes are measured in hundreds of millions rather than trivia answers. A transfer fee is not a price discovered through repeated, liquid trading the way a stock price is. Each deal is a one-off negotiation between a buyer with a specific need, a seller with specific leverage, and a player with specific value that nobody can measure with real precision. In the absence of a clean pricing mechanism, negotiators reach for the most recent comparable headline figure and adjust from there.

Subscribe now

Before August 2017, that comparable was Paul Pogba’s £89 million move to Manchester United. After that day, it was Neymar. Every agent representing a promising attacker suddenly had a new opening bid to invoke, and every buying club’s board had a new number against which spending felt psychologically permissible.

You can watch this play out with almost comic precision in what happened to Barcelona that same summer. Having lost their best player and pocketed a quarter of a billion euros, the club went shopping in a market it had itself just repriced. It paid Borussia Dortmund €105 million for Ousmane Dembélé, a promising but largely unproven twenty-year-old, and then handed Liverpool €160 million for Philippe Coutinho the following January. Barcelona spent north of €260 million replacing one departed superstar with two unproven ones, and both signings became case studies in overpayment, with Dembélé missing well over a hundred matches through injury during his years at the club and Coutinho never rediscovering his Anfield form. Barcelona wasn’t valuing these players against their actual output. It was valuing them against the shape of the hole Neymar’s departure had left, and that hole had just been priced at €222 million.

Signaling, status, and the sovereign wealth problem

Anchoring explains why the number spread, but it doesn’t fully explain why PSG paid it in the first place. That requires a second concept: the transfer fee as a signal rather than a straightforward market price.

Subscribe now

PSG in 2017 was not a football club optimizing for wins per euro spent. It was, and remains, an instrument of Qatari soft power, owned via Qatar Sports Investments with a mandate that extends well beyond the pitch into national branding and geopolitical visibility. In that context, the Neymar fee wasn’t a mispriced asset purchase. It was a extraordinarily expensive advertisement, deliberately loud enough to be heard everywhere, that signaled Qatar’s capacity and willingness to bend the world’s most popular sport to its ambitions. The economist Thorstein Veblen described conspicuous consumption over a century ago as spending whose entire value lies in being seen to spend it. A €222 million transfer fee is Veblen goods economics played out on a global broadcast, and once a market has a buyer for whom price is not a constraint but the point, every rational actor around that buyer has to recalibrate.

This is where the story stops being purely psychological and becomes structural. State-backed ownership introduced a category of buyer immune to the usual budget discipline that keeps transfer markets tethered to something like reality. When Saudi Arabia’s Public Investment Fund took controlling stakes in four Saudi Pro League clubs in 2023, the same dynamic repeated at a different scale. Saudi clubs spent roughly 970 million dollars on transfers that year, a leap from just 50 million dollars the year before, and the fee that symbolically opened that door was PSG selling Neymar himself to Al Hilal for a reported package worth close to 98 million dollars in wages and fees combined. The anchor that reshaped Europe in 2017 was exported to reshape Asia’s football economy six years later.

Wages follow fees, and fees follow wages

There’s a third mechanism at work, one closer to game theory than to psychology. Transfer fees and wages aren’t separate markets, they’re coupled through a kind of coordination problem. A player’s agent negotiating a new contract doesn’t argue from the player’s output alone.

Subscribe now

They argue from what comparable players have recently been valued at, because that comparison is legible to a club’s board and defensible to its fans in a way that a bespoke performance metric never quite is. Once Neymar’s fee reset the ceiling on valuation, his wage packet at PSG, reportedly exceeding 30 million euros a year net, became the new reference point every superstar’s representative could cite.

This produces a self-reinforcing loop economists would recognize as a coordination equilibrium, sustained entirely by shared expectations rather than fundamentals. No single club wants to be the one that breaks from the escalating pattern, because doing so risks losing top targets to rivals willing to pay the going anchored rate. FIFA’s own transfer data shows international spending hit a record 9.63 billion dollars in 2023, up nearly 50 percent on the year before, and landed at 8.59 billion dollars in 2024, still comfortably the second highest total ever recorded. Financial Fair Play was supposed to impose the discipline that would break this loop, yet clubs have shown remarkable creativity in structuring fees and add-ons to stay technically compliant while remaining anchored to the same inflated expectations.

The information problem underneath it all

Strip away the psychology and the geopolitics and you’re left with a genuine information asymmetry. Nobody, including the buying club, actually knows with confidence what a 25-year-old attacker will produce over five years of a volatile, injury-prone sport. In markets with this much uncertainty, price becomes less a measure of expected value and more a coordination device, a shared fiction that lets negotiations conclude. Kylian Mbappé’s move to Real Madrid in 2024 shows the strange endpoint of that logic.

He arrived on a free transfer under the Bosman ruling, technically costing PSG nothing in fees, yet the signing-on payments and salary commitments Real Madrid absorbed made the true economic cost comparable to a nine-figure transfer anyway. The fee had simply moved underground, because the market’s expectations around what a player of his caliber costs had never actually reset.

Which raises the more interesting question for anyone watching football’s economics rather than its scorelines. Anchors eventually break when a market participant with genuinely different incentives refuses to play along, the way a disciplined value investor ignores a hyped stock price. Bundesliga clubs, largely insulated by the 50+1 ownership rule that keeps majority control with member associations rather than wealthy individuals, have mostly resisted the inflationary spiral that gripped England, Spain, and France. Whether that discipline holds, or whether it eventually gets priced out of competing for the same talent everyone else has decided is worth these sums, might be the clearest live experiment in whether an anchored market can ever really unanchor itself.

What Pritchard had stumbled into, without using the phrase, was a market where the price of the asset had become disconnected from the value of the asset, and almost nobody involved in selling it had an incentive to say so.If that sentence sounds familiar, it should. It is roughly the plot of 2008.

Same structure, different asset

The parallel between digital advertising and the mortgage crisis is not a loose metaphor. It is a structurally similar failure, and understanding why makes both crises easier to see clearly. In the mid-2000s, mortgage originators made money by originating mortgages, not by whether those mortgages ever got repaid. The loan got bundled, sliced into tranches, rated by agencies who were paid by the very banks whose products they were rating, and sold on to investors who had no way of independently verifying what was actually inside the package.

This is a textbook information asymmetry problem, the kind George Akerlof described in his famous essay on used cars, where sellers know more about the quality of what they’re selling than buyers do. When buyers can’t distinguish good assets from bad ones, they pay an average price for both, good sellers exit the market because they’re underpaid, and what’s left skews increasingly toward lemons.

Subscribe now

Programmatic advertising runs on the same architecture. An advertiser buys engagement, but engagement is bundled, tranched, and resold through a chain of ad exchanges, supply-side platforms, demand-side platforms, and header bidding intermediaries so long that almost nobody in it, including the platforms themselves, can fully verify what’s underneath.

The rating agencies of this system are the platforms’ own dashboards, reporting impressions and click-through rates and viewability scores, graded by the seller, on the seller’s own product. In 2025, global digital ad spend crossed seven hundred fifty billion dollars for the first time, and advertisers lost a conservative one hundred sixty five billion dollars to invalid traffic that same year. That is not a rounding error sitting at the edges of the market. That is a fifth of the entire asset class being, in some meaningful sense, fake. eMarketer

When the measure becomes the target

The mechanism that makes this worse than ordinary fraud is Goodhart’s Law, the observation that once a measure becomes a target, it stops being a good measure. Engagement was originally a proxy for genuine interest. Advertisers wanted attention, and clicks and time-on-page were reasonable stand-ins for attention that was hard to measure directly. But the moment platforms began optimizing their algorithms and publishers began optimizing their content specifically to maximize that proxy, the proxy detached from the thing it was supposed to represent.

Rage-bait headlines, autoplaying videos, infinite scroll, and notification traps all exist because they move the needle on engagement metrics regardless of whether they produce anything an advertiser would actually recognize as commercial attention. The target got hit. The underlying goal got abandoned somewhere along the way.

Whose incentives you’re actually buying

Layered on top of that is a genuinely useful case of the principal-agent problem. The advertiser is the principal. The publisher, the ad network, and increasingly the algorithm itself function as agents, and their incentives point somewhere else entirely, toward maximizing the metrics that determine their own revenue, not toward the advertiser’s actual goal of selling more product. Uber discovered this the hard way around 2017, when it began forensically auditing its mobile marketing spend and found that a substantial share of its install attributions, run through partners including the Singapore-based ad network Fetch Media, were essentially fictional

Bots were clicking fake ads for an app already installed on the phone, claiming credit for installs that would have happened anyway. Uber slashed roughly a hundred million dollars from its marketing budget. App installs did not fall. The signal advertisers had been paying to receive had never really been connected to the outcome they cared about.

The correlated risk nobody priced

This is where the mortgage analogy gets uncomfortably precise. Bots do to attention what synthetic CDOs did to mortgage risk: they take a market with some real underlying value, wrap it in a level of complexity that resists inspection, and let the wrapping itself become the thing that’s priced. Bad bots surpassed a third of all web traffic by 2024, according to Imperva’s annual bot report, meaning that more than a third of the raw traffic flowing through the pipes advertisers pay for was never human to begin with.

Just as with MBS, the risk here is correlated rather than idiosyncratic. If one publisher’s traffic is fraudulent, that’s a bad site. If the entire measurement infrastructure across an industry systematically overstates the value of engagement, that’s a mispriced asset class, and everyone holding it is exposed to the same underlying flaw at once.

Who’s left holding the bag

The people most exposed when this corrects are not the platforms. Google, Meta, and the major exchanges have enough scale and enough proprietary walled-garden data to weather a re-pricing, the same way the biggest banks in 2008 got bailed out while smaller institutions vanished.

Subscribe now

The exposed parties are the small and mid-sized publishers who depend entirely on programmatic revenue with no other business model, the direct-to-consumer brands that built their entire growth strategy on performance marketing metrics they never independently verified, and the marketing agencies whose value proposition is built on being able to explain a dashboard nobody has stress-tested. When advertisers start demanding the ad-tech equivalent of loan-level data instead of trusting the rating, a lot of businesses built on the spread between reported and real attention will discover they were never as profitable as they looked.

What nobody has built yet is the equivalent of a credit default swap for attention, some instrument that lets an advertiser hedge against the possibility that the engagement they bought was never real. Until something like that exists, or until independent verification becomes as standard as an audited balance sheet, the market will keep pricing bot traffic and human attention as though they were the same asset. The 2008 crisis needed millions of individually rational decisions, made by people who each believed someone else was checking the underlying quality, to compound into something catastrophic. It’s worth asking who, exactly, is checking now.

That scene is worth sitting with, because it destroys the comforting story we tell about the 2008 crisis, which is that nobody saw it coming. Plenty of people saw it coming. Rajan saw it. Nouriel Roubini saw it. A hedge fund manager named Michael Burry saw it clearly enough to bet nearly a billion dollars of his investors’ money on it, years before Lehman collapsed. The real puzzle isn’t why the crash was invisible. It’s why visibility didn’t matter. Why an entire profession of extremely smart, extremely well paid analysts kept stamping AAA on securities that were, in retrospect, barely investment grade.

Subscribe now

Efficient market hypothesis says prices reflect available information. What 2008 taught us is that the hypothesis quietly assumes the people processing that information have a reason to process it honestly. Take that assumption away and the theory doesn’t fail randomly. It fails in a specific, predictable direction, and that direction has a name in economics: the principal-agent problem.

The issuer-pays trap

Here is the mechanism, stripped of jargon. Credit rating agencies like Moody’s and Standard & Poor’s are supposed to be neutral referees, grading the safety of bonds for the investors who buy them. But the agencies are not paid by investors. They are paid by the banks issuing the bonds. Moody’s rates Goldman Sachs’s mortgage bond, and Goldman Sachs writes the check. This is the issuer-pays model, and once you see it you cannot unsee how strange it is. It is roughly equivalent to a restaurant paying its own health inspector, and the inspector knowing that a bad grade means the restaurant simply hires a friendlier inspector next time.

Subscribe now

The numbers show exactly how this incentive reshaped behavior. Structured finance, the category that includes mortgage-backed securities and collateralized debt obligations, made up about 28 percent of Moody’s revenue in 1998. By 2007 it was close to half, and internal accounts describe it as having driven nearly all of the company’s growth over that decade. When your fastest growing, most lucrative business line depends on a steady stream of AAA stamps for mortgage bonds, you do not need an explicit order from management to go easy on the ratings.

The incentive is baked into the org chart. An analyst who got too skeptical about subprime pools wasn’t committing fraud in any legal sense. He was just quietly less useful to the firm than the analyst next to him who kept the deals flowing. Economists call this an agency problem, a gap between what the agent is supposed to optimize and what he is actually rewarded for optimizing, and it explains far more of the crisis than any story about pure incompetence.

Why nobody broke ranks

But incentive structures alone don’t explain why so many individually smart people converged on the same wrong answer at the same time. For that you need a second concept, one that goes back to Keynes’s observation that professional investing resembles a newspaper beauty contest, where the winning strategy isn’t picking the prettiest face but guessing who everyone else will vote for. Economists Douglas Scharfstein and Jeremy Stein formalized this decades later into what’s called a model of herding driven by career concerns. If you are a fund manager or a ratings analyst and you make an unconventional call that turns out to be right, you get a modest bonus.

If you make an unconventional call that turns out to be wrong, you get fired, because your boss cannot easily tell whether you were unlucky or incompetent. Making the same wrong call as everyone else, however, is much safer. Nobody gets fired for losing money the way the whole industry lost money. This is precisely why Rajan’s warning was so easy to dismiss. Following him meant taking on career risk that dwarfed the reputational cost of simply agreeing with the consensus.

Subscribe now

Michael Burry’s story is the sharpest illustration of the flip side of this dynamic, something financial economists Andrei Shleifer and Robert Vishny called the limits of arbitrage. In a textbook efficient market, mispricing gets corrected immediately because smart money rushes in to exploit it. Burry identified the subprime rot as early as 2005 and structured credit default swaps to bet against it. He was right, spectacularly right, and it nearly destroyed his fund before it paid off, because his investors watched their money bleed for two straight years and tried to pull out before the thesis played out.

Being early and being wrong look identical in the short run, and that resemblance is enough to keep most rational capital on the sidelines. The market can stay irrational, as the old line goes, longer than you can stay solvent, and that single fact is the load-bearing wall of the entire efficient market hypothesis’s failure mode.

A signal manufactured, not discovered

It’s worth adding one more layer, because the AAA rating itself is a signaling problem, not just an incentive problem. In markets with information asymmetry, ratings exist precisely because ordinary investors cannot independently evaluate the risk of a thousand-page mortgage pool prospectus. The rating is supposed to compress that complexity into a trustworthy signal. But a signal only works if it is costly or difficult to fake. When the agencies started competing with each other for issuer business, the signal became cheap to manufacture, and cheap signals stop signaling anything real.

This pattern didn’t retire with the 2008 crisis. Watch how the same three ingredients showed up in China’s property sector. Chinese developers like Evergrande carried debt loads north of 300 billion dollars while ratings agencies maintained investment grade assessments well into the sector’s decline, under a system where local and international agencies alike depended on continued issuance business from the very developers they were grading. The names change, the geography changes, but the architecture of the incentive is identical.

So here is the model worth carrying forward. The next time an entire industry of experts agrees confidently on something, don’t ask whether they are smart enough to be right. Ask who pays their salary, what happens to their career if they dissent, and whether anyone with real capital is structurally able to bet against the consensus without being destroyed before they’re proven correct. Consensus built on aligned incentives is evidence. Consensus built on everyone facing the same asymmetric risk of being the lone dissenter is not evidence at all. It is just fear, wearing a suit.

Twelve years later, that same country won its first Best Picture Oscar with a film about class resentment shot almost entirely in two apartments, and the year after that, a Korean show about people playing children’s games to the death became the most watched series in Netflix’s history. None of this was luck. It was industrial policy that happened to produce longing instead of steel.

A crisis reroutes an economy

The story has to start in 1997, because that is where the deliberate part begins. The Asian Financial Crisis hit Korea harder than almost anyone expected, forcing the government to accept a $57 billion IMF bailout, of which it drew down roughly $19.5 billion. Chaebols that had spent decades expanding on cheap, government-directed credit suddenly looked fragile, and the entire model of growth through heavy manufacturing and debt-fueled conglomerates was exposed as a single point of failure.

Korean policymakers walked away from that experience with a lesson straight out of trade theory. A nation’s comparative advantage should not sit in one basket. Concentration risk, the thing a portfolio manager loses sleep over, had nearly sunk the country.

Subscribe now

What followed was a genuinely strange bet for a government to make. Instead of doubling down on steel and cars, the Kim Dae-jung administration began treating culture as an export category with the same seriousness previously reserved for semiconductors. The Ministry of Culture, Sports and Tourism eventually built out a dedicated content office, and by the early 2020s its cultural content budget alone had swelled into the billions of dollars, with a slice ring-fenced specifically for nurturing and exporting popular culture.

This is the part people miss about Hallyu, the Korean Wave. It gets described as an organic cultural phenomenon, something that simply happened because Korean art got good. It did get good. But it got good inside a structure engineered to make that possible.

Manufacturing a comparative advantage instead of discovering one

Classical trade theory tells you that comparative advantage comes from what you already have, cheap labor, abundant land, a skilled workforce shaped by history. Korea’s cultural strategy is interesting precisely because it inverts that logic. The government did not look around and notice Korea was naturally gifted at pop music.

It looked at what other small, resource-poor economies with strong education systems and no natural advantage in commodities or land had going for them, and it decided that intellectual property, unlike oil or farmland, could be built from scratch through sustained investment in training, technology, and distribution.

The clearest evidence of this is the trainee system itself, which functions less like an artistic tradition and more like a supply chain. Companies such as SM Entertainment, JYP, and YG began scouting children as young as eleven or twelve, then put them through years of subsidized vocal, dance, and language training before a single one of them ever appeared on stage.

This looks bizarre through a Western lens, where fame is assumed to be discovered. In Korea it is manufactured, with attrition built in, the way a chip fab expects a certain yield loss. Out of every hundred trainees, a handful debut. The company treats the ones who never make it as sunk cost, the same logic a venture fund applies across a portfolio of startups it knows will mostly fail.

Why coordination beats individual genius

Here is where market design becomes the more useful lens than simple state subsidy. A single talented director or a single gifted singer does not create an export industry. What creates one is coordination across an entire value chain, training academies, production studios, distribution platforms, and government trade offices, all pointed in the same direction at once. This solves a coordination problem that decentralized markets are often bad at solving on their own.

Subscribe now

No individual studio has the incentive to build, say, subtitling infrastructure or global marketing capacity that benefits the entire industry rather than just itself. The state stepped in and absorbed exactly that kind of externality, building shared infrastructure like international content promotion offices and cultural centers abroad, essentially subsidizing the public good of “Korea as a recognizable content brand” that every individual studio then free-rides on profitably.

CJ ENM’s trajectory captures this well. The conglomerate did not simply produce Parasite and hope. It had spent two decades building multiplex chains, financing structures, and international festival relationships specifically designed to get Korean film in front of global gatekeepers, so that when Bong Joon-ho made a genuinely brilliant film, the distribution muscle to carry it to a Best Picture win already existed.

Compare this to Nigeria’s Nollywood, which produces an extraordinary volume of films with real cultural reach across Africa but has never built the equivalent export scaffolding, and so it remains a continental phenomenon rather than a global one despite arguably comparable raw creative output per dollar spent. Talent alone was never the binding constraint. Infrastructure and coordination were.

Signaling, information asymmetry, and why global audiences trust the brand

There is also a signaling story here that economists will recognize immediately. Foreign buyers like Netflix or a European broadcaster face genuine information asymmetry about which unfamiliar market’s content is worth the risk. A show from an unknown industry is a lemon problem in miniature, nobody wants to be first to invest heavily in overseas content only to find it does not travel.

What Korea built over two decades is a reputational signal so strong it now substitutes for individual due diligence. Netflix does not evaluate each Korean drama from scratch the way it might a debut show from a market with no track record. The “K” prefix itself has become a quality signal, the cultural equivalent of a country-of-origin label that lets buyers economize on information.

The scale this has reached is now genuinely large. Hallyu-related exports hit nearly $19 billion in 2025, up almost 16 percent from the year before, and the games industry alone, still the largest single component of Korea’s cultural IP exports, pulled in over $5 billion, more than music, film, broadcasting, and animation combined.

Subscribe now

Music exports specifically surged 84 percent in a single year, driven partly by the unexpected global reach of an animated film called KPop Demon Hunters, itself a strange late-stage confirmation that the Korean pop machine has become so recognizable a genre that it can be exported as a Hollywood animation concept and still read as authentically Korean.

The uncomfortable question this leaves behind

The part that should unsettle anyone who believes culture is supposed to emerge spontaneously from a people’s soul is that Korea’s version so plainly did not. It was budgeted, targeted, and iterated on with the same policy tools once used to build container ships. That raises a genuinely uncomfortable question for every other government now watching Hallyu’s numbers and wondering if they can replicate it.

Comparative advantage in cultural production might turn out to be buildable in the same way a semiconductor fab is buildable, given enough capital, patience, and willingness to treat feeling itself as a supply chain.

Or it might turn out that Korea’s model worked precisely because it was first, because it built the trust signal before anyone else tried, and that the second country to attempt the same playbook will find the global audience simply is not interested in a copy. Nobody actually knows yet which of those two stories is true, and the answer will probably arrive from Saudi Arabia’s entertainment budget, or Indonesia’s, or Nigeria’s, before it arrives from an economics paper.

A separate analysis of a 2022 University of Cambridge study found that the deforestation threat used to justify these projects had been overstated by roughly four hundred percent on average. Companies including Disney, Shell, Gucci, and easyJet had been buying certificates for trees that were, statistically speaking, never in much danger of being cut down.

This is not really a story about bad actors gaming a good system. It is a story about a market that was, by its own design, almost impossible to police, and that is what makes it one of the most interesting broken markets in the world right now. Carbon credits are a case study in what happens when you build a multibillion dollar asset class on a product whose core attribute, whether a ton of carbon dioxide was actually kept out of the atmosphere, cannot be directly observed by the buyer.

The lemon problem, but for trees

Economists have a name for this, and it comes from a paper about used cars. George Akerlof’s 1970 essay on the market for lemons described what happens when sellers know the quality of what they are selling and buyers do not. Sellers of good used cars, unable to prove their quality, get the same price as sellers of bad ones, so they exit the market.

What remains skews toward lemons, prices fall further, and the cycle repeats until the market shrinks to a shadow of what it could be. This is adverse selection, and it does not require anyone to be dishonest. It only requires that verification is expensive or impossible and that buyers cannot tell good from bad at the point of sale.

Carbon offsets have almost exactly this structure. A credit that represents genuine, additional, permanent carbon reduction looks identical on a spreadsheet to a credit generated by a forest that was never actually threatened, or one that gets logged three years after the credits are sold, or one where the underlying carbon simply migrates to unprotected land nearby, a problem called leakage.

The buyer cannot see the forest. They cannot verify the counterfactual, meaning what would have happened to that land without the money. All they can see is a certificate and a logo from a standards body. And here is where the data gets genuinely revealing about the mechanics of the collapse.

According to MSCI’s 2025 carbon market analysis, credits rated A or higher by independent quality raters sold for an average of about fourteen dollars and eighty cents per ton in 2025, while credits rated CCC to B sold for around three dollars and fifty cents. That is a real price signal, proof that buyers who can assess quality will pay for it.

Subscribe now

But the same MSCI data found that forty three percent of all credits retired in 2024, meaning actually used to claim an offset, came from those low rated projects anyway. The lemons are not just surviving in this market. They are still dominating volume, years after everyone learned they were lemons.

Who is actually grading the exam

The second mechanism compounding this is a straightforward principal-agent problem, and it sits inside the verification system itself. Carbon standards like Verra do not generate credits directly. Project developers hire independent auditors to verify their own projects, and the developer pays the auditor’s fee. The auditor’s future business depends on developers continuing to bring them work, which means the auditor’s incentive is not perfectly aligned with rigor. It is aligned with approval.

Subscribe now

This is the same structural flaw that shows up in credit rating agencies paid by the bond issuers they rate, or, closer to home for anyone who has watched the restaurant world, Michelin inspectors whose stars shape a chef’s entire business model. When the party being evaluated pays the evaluator, skepticism becomes an expensive thing to sell.

The Northern Rangelands Trust project in Kenya is a vivid illustration of how this plays out on the ground. The project generated carbon credits across two million hectares of land used by Indigenous pastoralist communities, credits that were purchased by major corporate buyers seeking offsets. In January 2023, following an investigation by Survival International into land rights abuses and questionable carbon accounting tied to the project, Verra suspended its credits.

By November of that year Verra completed a review and reinstated the project, but critics pointed out that the review relied heavily on a single 2022 audit conducted by a verification firm with an ongoing commercial relationship with the project, and did not meaningfully engage with the substance of the complaints. The reinstatement did not resolve the underlying question. It just illustrated who gets to decide what counts as resolution.

A market with no scarcity to sell

There is a third layer here, and it is more structural than the other two. A functioning market for an environmental good needs the underlying scarcity to be real and enforced, the way a fishing quota or a taxi medallion creates scarcity through a hard legal cap. Voluntary carbon credits have no such cap. Anyone can, in principle, generate a new supply of credits by starting a new project and finding a standards body willing to certify it.

Supply is not fixed by nature or by regulation. It is fixed by how many developers are willing to write an optimistic baseline. That absence of a hard constraint is why prices for low quality credits stayed near three to four dollars a ton for years even as reputational scandal after reputational scandal broke, according to Ecosystem Marketplace’s State of the Voluntary Carbon Market data, which also shows the market’s total transacted value falling to five hundred thirty five million dollars in 2024, down from roughly two billion dollars at its 2021 peak.

Subscribe now

What would actually fix this is not a mystery, economically speaking. It looks like mandatory, government backed emissions caps rather than voluntary purchases, third party verifiers paid by a pooled industry fund rather than by the project they are grading, satellite and remote sensing data as a baseline rather than a spot check, and liability that follows a credit if it later turns out to be fraudulent, the way a bad bond can trigger legal exposure for the underwriter years later. The European Union’s shift toward mandatory carbon border adjustments and stricter Article 6 rules under the Paris Agreement are early, clumsy attempts at exactly this kind of hard constraint.

None of that has fully arrived yet, and in the meantime companies keep announcing net zero pledges that lean, at least partly, on a market that a majority of its own biggest supplier’s credits turned out not to represent. The real question is not whether carbon credits are broken. It is whether a market this structurally prone to adverse selection can ever be fixed without simply becoming a tax by another name, and whether anyone with the power to make that trade would actually choose to.

Roughly half of every Galaxy phone sold anywhere in the world was built by workers who, three decades ago, were farming rice paddies that this same land used to be. That is not a footnote to the story of global trade realignment. It is the story.

The boom everyone miscalls China plus one

The easy explanation for Vietnam’s manufacturing surge is “China plus one,” the idea that multinational companies, spooked by tariffs and geopolitical risk, are simply diversifying their supply chains and Vietnam happened to be next door. There is truth in this. But it flattens a much more interesting picture, one where the outcome was not luck but the payoff of a bet Vietnam placed on itself starting in 1986, when the Communist Party launched Doi Moi, a set of market reforms that opened the country to foreign capital while keeping political control tightly centralized.

Vietnam spent the following thirty years building the very things that make a China-plus-one strategy possible: deep water ports, industrial zones with predictable land leases, tax holidays for exporters, and free trade agreements with nearly everyone who matters, including the EU, the CPTPP bloc, and the United States. When American and Korean firms went looking for an alternative to China around 2018, they were not discovering Vietnam. They were cashing in on infrastructure Hanoi had been laying for two generations.

An advantage that was built, not inherited

This is where the economics gets genuinely interesting, because it complicates the tidy Ricardian story of comparative advantage that most people learned in an introductory course. The textbook version says countries specialize in what they are naturally good at, based on fixed endowments of land, labor, and capital.

Vietnam’s rise shows something closer to what economists call dynamic comparative advantage, the idea that a country’s productive strengths can be manufactured through deliberate policy rather than inherited from geography. Vietnam did not have a natural gift for semiconductor packaging or smartphone assembly.

It built the conditions for that advantage to exist, through vocational training pipelines aligned with employer needs, through industrial parks purpose built for electronics supply chains, and through decades of signaling to global capital that contracts here would be honored and currency would stay stable. By the time the trade war with China escalated, Vietnam did not need to compete on cost alone. It had already built the ecosystem.

Why clusters compound

That ecosystem is where a second concept, agglomeration economies, starts doing real work. When Samsung opened its first Vietnamese factory in Bac Ninh in 2009, it was a single company making a bet.

Subscribe now

But manufacturing clusters compound. Local suppliers learn to meet a foreign multinational’s quality standards, then start supplying other foreign multinationals. The number of Vietnamese firms inside Samsung’s own supply chain grew from 25 in 2014 to 257 last year. Engineers trained at one plant move to competitors down the road. Logistics networks built for one company’s needs become the default option for the next one that arrives.

This is the same dynamic that made Shenzhen impossible to replicate quickly, and it explains why Vietnam pulled ahead of countries with cheaper labor, like Bangladesh or Myanmar, which never built the surrounding density of skilled suppliers and specialized logistics that turns a factory into an ecosystem.

The subsidy nobody sees

The externality here cuts both ways, and it is worth sitting with, because it explains why so few countries manage to seize a moment like this even when the opportunity is sitting in plain view. Every additional firm that locates in a cluster makes the cluster more attractive to the next firm, but no individual firm captures the full value of that improvement, which means the initial investment, the one that gets the cluster started, requires either extraordinary confidence or a government willing to subsidize the early movers until the agglomeration effects kick in on their own.

Vietnam’s government did exactly that, offering four year corporate tax holidays and reduced rates for another decade to qualifying exporters, essentially eating the cost of building critical mass so that private capital would eventually arrive on its own. Ethiopia tried a version of this with its Hawassa Industrial Park, built with World Bank backing specifically to lure garment manufacturers away from Asia. It opened in 2016 with real ambition, but political instability, currency controls, and weak worker retention meant the cluster never achieved the same self-reinforcing density.

Occupancy stalled, several anchor tenants pulled back, and the park became a cautionary tale about how the same playbook can fail without the underlying institutional credibility to sustain it.

Safety in numbers

There is a game theoretic layer too, and it helps explain the timing. Once a handful of major manufacturers relocate production away from China, other firms face a coordination problem. Moving alone is risky, since suppliers, logistics, and skilled labor might not follow. Moving alongside others is far safer, since the ecosystem effect described above only exists once enough firms have made the jump together.

Vietnam’s early wins, Samsung’s smartphone plants, Intel’s semiconductor assembly and test facility, Nike and Adidas footwear contractors, functioned as coordinating signals that told the next wave of companies it was safe to follow. Foreign direct investment into manufacturing hit roughly 38 billion dollars in 2025, and processed industrial exports reached 421 billion dollars, nearly 89 percent of everything Vietnam sold abroad.

Subscribe now

The United States alone bought over 153 billion dollars of Vietnamese goods that year, producing a trade surplus in Vietnam’s favor of 134 billion dollars, a gap so large it has started drawing its own scrutiny in Washington over transshipment, where Chinese goods are lightly finished in Vietnam to dodge tariffs meant for China.

The success that might become the problem

That scrutiny points to the tension sitting underneath this entire success story. Vietnam’s rise was built on being the destination where global capital could avoid a trade war it did not start. But the same forces that made Vietnam useful for that purpose, cheap redirection of Chinese supply chains, deep integration with Chinese component makers, and a widening deficit with China itself of over 115 billion dollars in 2025, are the forces that could make Vietnam the next target once policymakers in Washington or Brussels decide the workaround has become the problem.

Every country that wins a supply chain reshuffle inherits the question its predecessor never had to answer, which is what happens when the very originality of the win starts to look, from a distance, exactly like the thing everyone was trying to avoid in the first place.