Alibaba's Flywheel Still Spins. The Question Is How Hard the Wind Now Blows Against It.

In February 1999, in a second-floor apartment at Lakeside Gardens in Hangzhou, Jack Ma gathered a room of friends and pitched a marketplace.⁷ Twenty-five years later that marketplace was, by gross merchandise volume, the largest digital retail business on earth - Taobao and Tmall together moving more goods than any commerce platform anywhere.⁴ The machine that did it has a name everyone uses and almost nobody examines: the flywheel. More buyers pull more sellers; more sellers mean more selection; more selection pulls more buyers; payments and logistics grease the whole loop until it spins on its own momentum.

The official story is that this flywheel is a self-reinforcing monopoly - a perpetual-motion machine immune to competition. That story is half right and dangerously incomplete. The flywheel is real and still turning. But it is turning against headwinds the celebration leaves out, and the place to see those headwinds is not the GMV trophy. It is the thin, quiet number the flywheel actually converts its motion into.

The flywheel never charged a toll. It ran an auction.

Here is the part the word 'flywheel' hides. Most marketplaces take a fixed cut of every sale - a clean commission stamped on each transaction. Alibaba's China retail business barely works that way. The majority of its money is 'customer management revenue': merchants paying to be seen, through pay-per-click, pay-per-impression, time-based slots, and pay-per-sale ads.⁴ In other words, Alibaba doesn't mostly tax the cart. It runs an auction for the cart's attention, and the sellers bid against each other for it. In FY2024 that auction generated RMB414,414 million - about US$57.4 billion - up 5% year over year for Taobao and Tmall.² The flywheel's genius was never the swipe fee. It was making millions of sellers compete to pay for visibility on a stage only Alibaba owned.

This is why scale mattered so violently. An auction is only worth bidding in if the buyers are there. When Alibaba's three China marketplaces moved US$296 billion of GMV across 279 million active buyers in the year leading into its 2014 IPO¹, no merchant could afford to ignore the bidding - the buyers were nowhere else. The flywheel didn't just attract sellers. It cornered them. Every additional buyer raised the price of attention, and Alibaba collected the difference.

The trophy that hides the slippage

Watch the GMV crown and you'll conclude Alibaba is fine. Watch the share of the market and the price of attention, and you see the flywheel laboring. PDD surpassed Alibaba in annual active consumers in the China retail marketplace as early as the fiscal year ended March 2021 - the buyer side of the loop, the part everyone assumed was unassailable, already had a peer.⁵ Worse, Alibaba's share of China's online retail GMV fell from roughly 72% to roughly 62% in the single year ended March 2023.⁵ Ten points of share, gone in twelve months. That is not a flywheel tightening. That is a flywheel leaking.

And the auction is softening underneath. Spending is mix-shifting toward Taobao, which monetizes at a lower take rate than premium Tmall, while PDD and Douyin offer merchants cheaper rooms to buy attention in. Morningstar expects marketplace monetization rates to keep declining long-term for exactly these reasons.⁵ When sellers have somewhere else to bid, the bids fall - and a flywheel that runs on the price of attention slows the moment the price has competition. The B2B arm tells the same story in miniature: Alibaba.com still projected about US$60 billion in GMV for 2024, up roughly 20%, but its president conceded growth had 'slowed after a sevenfold increase' over the prior five years and named business-model transformation as the core problem.⁸

But isn't a falling take rate just the price of staying number one?

The fair objection is that this reads too bearishly. The flywheel is structurally intact: Taobao and Tmall were still the world's largest digital retail business by GMV for the twelve months ended March 2024⁴, 88VIP loyalty membership crossed 35 million², and as recently as the December 2024 quarter, customer management revenue rose 9% year over year on GMV growth and an improved take rate.³ If Alibaba is deliberately trading take rate for volume - lowering the price of attention to keep buyers and sellers from defecting - that's not a leak; it's a rational defense, and the December quarter shows the take rate can recover when it chooses. There is truth in this. A high take rate you can't keep is worth less than a lower one you can defend, and Alibaba is clearly choosing to defend the loop.

But the steelman quietly concedes the thesis. A flywheel that has to lower its own toll to keep buyers and sellers from leaving is no longer the self-reinforcing monopoly of the legend - it is a strong network being actively held together against gravity. The 2021 buyer-side overtake and the ten-point share fall happened anyway, take-rate management notwithstanding. The machine still spins, beautifully, at the largest scale in commerce. It just no longer spins for free. Every turn now costs something it didn't have to spend when there was nowhere else for a Chinese shopper or seller to go.

The flywheel was never the trick. The trick was building a marketplace so total that sellers had no choice but to bid against each other to be seen on it, and then collecting the difference. For two decades nobody could route around it, so the price of attention only rose. Now there are other rooms - PDD's, Douyin's - and the rooms compete on the one number the flywheel quietly lives or dies on. Alibaba's wheel still spins at a scale no rival has matched. The question the celebration skips is the only one that matters: not how big the wheel is, but how hard the wind now blows against it, and how much of each turn is spent just holding it in place.