Alibaba Buys Its Stock and Sells Its Stock

Alibaba Buys Its Stock and Sells Its Stock

Alibaba has spent the last couple of years being one of the most voracious buyers of its own stock in the world, and it still has about $19 billion of buyback authorization left. On Sunday it announced that it is also selling about $10 billion of its own stock to the rest of the world. Both of those statements are true at the same time. That is sort of the whole story.

Here is the deal. Alibaba proposes to place HK$80 billion — about $10.2 billion — of newly issued ordinary shares in Hong Kong. A "placing" is just an accelerated sale of new shares to big investors at a discount to the market price. Because they are new shares, the money goes to the company, not to a founder or a private-equity firm cashing out; this is not an insider exit that you can wave away as smart selling. And all of the net proceeds are earmarked for AI — "full-stack AI capabilities," in the company's phrasing, expanding and enhancing AI infrastructure, to extend its global AI leadership. That is the respectable label. The plainer reading is that Alibaba is paying for the biggest AI build of its corporate life by selling its own equity, at a discount, in a stock that is down.

Now, what sort of machine is that? A buyback is a company spending cash to take its shares out of circulation. A placement is a company printing new shares and selling them, at a discount, to people who want in. Most people treat those as opposites: returning capital versus raising it. Alibaba is running both at once, on the same equity, with tens of billions on each side. The respectable way to describe that is "flexible capital allocation." The more honest way is: the company that keeps demonstrating by deed that it thinks its own stock is cheap is now selling that stock wholesale, at a schedule and price chosen by its banks.

The timing is the sharp part. Three days before, Alibaba reported June-quarter results in which net profit fell about 75%, not because anything went wrong with the business but because the company is shoveling money into AI infrastructure; cloud growth strengthened sharply, and the stock fell anyway. Market data show the US-listed ADS (American depositary share) closed around $119, down about 8.6% in its latest session and roughly 19% year to date — about 38% below the 52-week high near $193. The placement is arriving right after the worst profit print in a long time, into a falling stock. The options market is already pricing around 44% implied volatility into the name, which is the market's way of saying: we expect a move.

The bill underneath it is enormous. In early 2025, Alibaba pledged at least RMB 380 billion — roughly $52 billion — to cloud and AI infrastructure over three years, and by early 2026 it was reportedly considering pushing that toward $69 billion. A build that size eats cash faster than even Alibaba generates it, and the incremental funding has to come from somewhere; in this case, somewhere is its own equity. This is very 1999 of it. American telecoms sold stock to lay fiber, then the economics followed. Alibaba is selling stock to buy GPUs and build data centers, and the economics, presumably, will follow. Alibaba has even run this exact pump before: its 2019 Hong Kong listing was built to raise up to $13.8 billion of fresh equity.

There is a market-structure detail worth pausing on. This is the largest primary follow-on ever done by a Hong Kong-listed company — a "primary" deal, meaning new shares, not insiders selling — and the biggest Regulation S offering on record, and the third-largest primary follow-on of the year anywhere, behind Alphabet and Intel. In other words, the incumbent version of the AI buildout is being financed, in part, by equity issuance at the very companies doing the building. The pattern is old; only the wrapper is new.

And then there is the securities-law plumbing, which is the funniest part. The new shares are offered only to certain non-U.S. persons in offshore transactions, under Regulation S — the rule that lets issuers sell unregistered securities to people who are not in America — and they are not registered under US securities law. So a company that is one of the biggest names on the New York Stock Exchange is printing shares and selling them exclusively to people who are not in the United States. The Americans who own the stock through ADSs get diluted by a transaction they were structurally never offered.

US holder: "Wait, you just diluted me with shares I wasn't allowed to buy?" Alibaba: "Yes. But contractually, those shares were never for you."

The dilution is real but, on the numbers, modest. HK$80 billion at roughly HK$125 per share is on the order of 640 million new shares, a bit over 3% of the roughly 18.5 billion ordinary shares outstanding. Alibaba has not announced the price, the share count, or the discount; the deal is sized at HK$80 billion, and the number of shares floats with whatever price clears the book on Monday. The discount — the gap between where the shares trade and where the banks' order book clears — is the number that will tell you how eager the market really is for China's AI story, and how much of the tab Alibaba's own shareholders are being asked to pick up. A small discount means the market is genuinely paying for the AI future. A big discount means the market is being paid to take it.

The defense has a real case. A bit over 3% dilution is a cheap price for an unobstructed shot at leading China's AI build, and a company can rationally sell equity at a discount to fund a project it expects to be worth far more than the discount. The reason it still reads as selling the dip is the sequencing: right after that profit print, into a stock already down a fifth on the year. And the buyback makes the picture stranger in a clarifying way. The company is still a huge net buyer of its stock in the US and now a huge issuer of its stock in Hong Kong. One machine supports the price; the other converts the price into cash. The two are not contradictory; they are two doors on the same share register. But a $10 billion placement is a bigger machine than a buyback tranche.

Strip the labels off, and Alibaba is doing the oldest trick in the corporate-finance book: capital spending is outrunning operating cash flow, and rather than slow the build or lean harder on debt, it is paying with its own equity, priced near the low end of its range. The new shareholders get the discount. The existing shareholders get the dilution. The company gets $10 billion and an option on whether China's AI build clears its cost of capital. A buyback is a company's opinion that its stock is cheap. A placement is a company's statement that it needs the money. The two sat side by side in the same announcement, and only one of them is actually an opinion.