Alibaba's net income collapsed more than 75 percent year on year to 10.5 billion yuan — about $1.6 billion — in the June quarter, results released Thursday showed, even as revenue rose a healthy 9 percent. The gap between those two numbers has a name: capital expenditure, poured into chips, data centers, and compute at a pace that turns one of China's most profitable companies briefly unprofitable on purpose. [1][2]
The bill is enormous and half-paid. Chief executive Eddie Wu told analysts the company has already spent half of its planned 380 billion yuan — $56.4 billion — in AI infrastructure investment scheduled for 2026 through 2029. "In order to be able to capture that future growth, we first need to make these capex investments to build out the necessary compute capacity," Wu said, promising the spend breaks even within three years at current gross margins. [3]
The paradox is that the strategy is working exactly as designed, which is precisely why the income statement looks like this. Cloud revenue grew 45 percent in the quarter on enterprise demand for training and serving AI models, the fastest clip in years. [1] Alibaba's in-house T-head chips are already deployed at scale across supernode clusters for model training, which Wu argues will lift gross margins substantially as proprietary silicon replaces purchased chips. He was candid about the frontier-model side of the ledger: monetization remains inadequate today, but the end game is artificial general intelligence, where he expects monetization to be significantly higher than anything visible now. [3]
The reorganization tells you how seriously the company takes the pivot. Alibaba split its AI businesses from its cloud arm earlier this year, created a dedicated Token Hub group under Wu, and now reports in four units: e-commerce, AI cloud and computing, AI model applications, and other. CFO Toby Xu set expectations elsewhere in the empire too — quick commerce reaches overall profitability in fiscal 2029, and China's consumption slowdown is a real headwind the AI story cannot paper over. [3]
Markets took both halves of the message Thursday morning, sending U.S.-listed shares down around 4 to 5 percent premarket. [1] Then the feeds did what they do. Mainstream coverage performed sober arithmetic: reinvestment cycle, visible cost, credible management timeline. Capex Twitter went tribal instead — one camp posting buildout-greatest-ever threads, the other circulating 1999 telecom-fiber comparisons where infrastructure got built, profits did not follow for a decade, and most of the builders died broke. Both camps are arguing about the same income statement; neither can be proven wrong for three years.
What makes Alibaba worth watching rather than dismissing is that it is running the experiment faster and more visibly than its American peers, who spread similar spending across four balance sheets and gentler accounting quarters. Here is one company, one number, one promise: eat the profit now, own the compute layer, collect later. The quarter proves the eating. The collecting remains, for now, a matter of faith backed by $56 billion. [3][1][2]
-- THEO KAPLAN, San Francisco