The China Exit

The terms: Didi Chuxing absorbs Uber China. Combined entity valued at $35 billion — Didi at about $28 billion, Uber China at about $7 billion. Uber Global takes 5.89 percent of the combined company in preferred equity carrying a 17.7 percent economic interest; Baidu and Uber China's other shareholders pick up another 2.3 percent, bringing the total to roughly 20 percent. Didi turns around and puts $1 billion into Uber's global business at a $68 billion valuation. Travis Kalanick's blog post called it a new beginning, which is what you call it when the alternative is calling it what it was.

Price of the lesson: about $2 billion over two years, most of it subsidies paid to keep rides cheaper than they cost. Kalanick had conceded in February that Uber was losing over a billion a year in China.

Four days before the announcement, something happened that got a fraction of the coverage. On July 28, 2016, China's Ministry of Transport and six other departments jointly issued the Interim Measures for the Administration of Online Ride-Hailing Services, legalizing ride-hailing nationally, effective November 1. Buried in the text: operators may not price below cost to squeeze out competitors. Fares are to be set by market mechanism.

The regulation legalized the business and outlawed the strategy, in the same document.

Be careful with the sequencing, though, because the obvious reading is the wrong one. A $35 billion combination is not negotiated in four days. Uber and Didi had been in talks for months, under investor pressure to stop the bleeding and with an IPO calendar to protect. The Interim Measures did not kill the deal. They priced it. A state that legalizes your industry and forbids your only strategy four days before you sign has not ended your negotiation — it has removed the last argument for staying in the room. That is a colder instrument than a veto, and a more precise one.

There is a second problem with reading the rule as the cause, and it is the more interesting one. The below-cost pricing ban applied to everybody. Didi was subsidizing at least as hard as Uber was, and Didi is still here. A symmetric rule cannot by itself produce an asymmetric outcome.

What closes the gap is enforcement. A rule that binds everyone and is applied to someone is a far more sophisticated instrument than a rule that names its target: unimpeachable on its face, discretionary in practice, and never obliged to explain itself. Which leaves only the question of why the discretion ran the way it did.

Didi already held the payment rails through WeChat Pay and Alipay, the municipal licensing relationships, the mapping data, the state's unspoken preference — and, the part usually skipped, roughly equivalent capital. Tencent, Alibaba, and a $1 billion check from Apple. You cannot out-subsidize an opponent whose subsidies are underwritten by the national champions.

This is where the network-effect thesis goes to die. Metcalfe's law doesn't clear customs. The pitch every deck made in that era was that liquidity compounds until the market tips, and geography is a rounding error on the way there. Uber arrived with $2 billion and the sincere conviction that a two-sided marketplace is a two-sided marketplace anywhere on the planet.

The correction is unglamorous and load-bearing, and it is not really about networks. Uber's error was a category error about which layer it owned. It believed the app and the marketplace were the sovereign layer. They weren't. The sovereign layers were payment, identity, mapping, municipal permission, and legal standing — every one of them rented, and rented from a landlord who lives there and votes. Uber owned the interface and rented the ground. eBay, Google, Amazon, Groupon, Uber: each showed up with the same theory, and each discovered that what looks like a platform is actually a stack, and that most of the stack belongs to somebody else.

The clearest proof arrived three years later and reads like a footnote. Bloomberg reported in 2019 that the transaction still had not received Chinese antitrust approval. Three years after the sale, the deal that ended the war sat in regulatory superposition — completed, unblessed, reversible in principle. That is not administrative sloth. The state never needed to block the merger; it needed the merger to remain blockable. Indefinite reviewability is not a delay in the exercise of power. It is the exercise of power — a standing conditional hold on a private company, costing nothing to maintain and available the moment it becomes useful.

One thing this story is usually told without. The $2 billion was not vaporized; it was transferred. It went to Chinese riders and drivers, as artificially cheap fares and inflated payouts, for two years — and when the price war ended, fares rose and the payouts fell. Told from the altitude of capital, this is a story about who won and who spent. The people who actually lived through the subsidy war lived through it as a brief, strange period when the city was cheap, followed by the bill.

The shares turned out fine as a trade. But what the stake actually bought was a seat at the table of the company that had beaten Uber, in a market Uber could no longer enter. Ten years on, every platform CEO can recite the lesson and none has absorbed it. A network effect is a legal artifact wearing a math costume.