Alibaba's $10.2bn placement prices AI compute off the equity market, not the profit line

The ratio fixes the position better than the headline does. Capital expenditure of 67.68bn yuan in the April to June quarter ran against 48.44bn yuan of cloud and AI revenue in the same three months ^[3][6], roughly 1.40 yuan of spend for every yuan the segment billed ^[1]. A 75% fall in net profit is the arithmetic consequence of that, and a Hong Kong placement is what lets the spending continue without the profit line getting a vote on it.

It is worth being precise about what customers are currently paying for. AI model services above 16bn yuan in annual recurring revenue ^[7] work out to about 8% of cloud and AI revenue at the current run rate ^[2]. The other 92% is conventional cloud demand underwriting machines bought against a forecast rather than a backlog.

Chief executive Eddie Wu has not dressed up the sequence: "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" ^[8]. The repair he points to on the way back is silicon, with Wu saying Alibaba's own chips could lift margins substantially as they scale ^[10]. For a buyer, that means the margin recovery route runs through a chip programme rather than through list price increases, at least while the chips are still ramping.

The plan being topped up is also being rewritten mid-flight. The 380bn yuan committed over three years in early 2025 is about half spent, and reports suggest a rise to 480bn yuan ^[9], an increase of 100bn yuan to a programme not yet two years old ^[3]. The placement is not a one-off gap filler on a fixed budget; it is the funding method for a budget that keeps moving.

Scale gives two useful reference points. At $10.2bn the raise is around 12% of the $85bn Alphabet has taken in equity this year ^[4], so the largest primary follow-on in the history of the Hong Kong market ^[4] is roughly an eighth of what the largest issuer raised. Set against the European Union's roughly 20bn euro AI gigafactory programme, one afternoon's placement is about half the continent's flagship compute budget ^[11].

The consequence for procurement is not scarcity. Capacity funded from equity does not have to clear a return hurdle in the quarter it is installed, which is why capex can run at 1.4 times segment revenue ^[1] while cloud and AI revenue grows 45% ^[6]. Buyers should expect capacity ahead of demand and prices that the segment's current economics do not support on their own. What that borrows against is the renewal, once the equity story requires the machines to earn their keep.