Alibaba Raised $10 Billion for AI in One Session, and Its Own Stock Punished It

Alibaba pulled off the biggest share sale in Hong Kong’s history to bankroll its AI ambitions, and the market’s answer was to sell the stock anyway. Meta, which is spending more than $100 billion a year building its own AI infrastructure, is also quietly one of Microsoft’s biggest AI customers. Nvidia is in early talks with a Korean chip startup built specifically to compete with it. And in St Louis, an Indian grandmaster spent Monday taking apart one of the best young players in the world on his way toward a career final.

Alibaba raised $10 billion for AI in one session, and its stock fell anyway

Alibaba sold 710 million new shares at HK$112.70 each on Monday, a 3.6 percent discount to Friday’s close, to raise HK$80 billion, or about $10.2 billion. It’s the largest primary follow on share sale ever completed by a company listed in Hong Kong, and the third largest anywhere in the world this year, behind only Alphabet’s $80 billion raise in June and Intel’s $15 billion sale earlier this month.

Alibaba says every dollar of the proceeds will go toward what it calls full stack AI: chips, data center infrastructure, and the models and products built on top of them. Demand for the offering was strong enough, including interest from sovereign wealth funds, that Alibaba increased the size of the deal before pricing it.

None of that stopped the stock from dropping as much as 10 percent in Hong Kong trading on Monday, its steepest single day fall since April 2025. The timing didn’t help. The raise landed just a week after Alibaba reported that quarterly net profit had fallen 75 percent from a year earlier, a decline the company has pinned largely on AI spending. Investors, in other words, are being asked to fund more of the exact spending that just gutted this quarter’s profit, and plenty of them are not thrilled about it even as they agree AI is where the company has to go.

Meta is paying Microsoft hundreds of millions a year to rent the AI it isn’t building fast enough

Meta has become one of the largest AI customers on Microsoft’s Azure cloud platform, spending hundreds of millions of dollars a year and burning through trillions of tokens, the basic unit of AI computing, every week, according to Bloomberg. That puts Meta in roughly the same tier as ByteDance, which has generally held the top spot among Azure’s AI customers.

What makes this notable is that Meta is simultaneously spending more than $100 billion a year of its own money on data centers and chips to build AI infrastructure it fully owns. Renting serious capacity from a company it’s supposed to be racing against, rather than waiting for its own buildout to catch up, says something about how tight AI compute still is even for a company with Meta’s budget. Both Meta and Microsoft declined to comment on the arrangement when asked, which has left outside observers reading the numbers as a sign that so called circular deals, AI companies paying each other for compute, chips and cloud capacity in overlapping directions, are becoming more common across the industry rather than less.

Nvidia is circling a Korean chip startup built to compete with it

Nvidia chief executive Jensen Huang met this week with Sunghyun Park, the co-founder and chief executive of Rebellions, a six year old South Korean company that designs neural processing chips meant to rival Nvidia’s own hardware for AI inference work. The two sides are in early, preliminary talks that could lead to a technical partnership, a direct investment, or even an acquisition, though people close to the discussions caution nothing has been decided and the talks could still fall apart.

Rebellions has raised roughly $850 million so far from investors including SK Hynix, Samsung Ventures and Arm, plus a direct investment from the South Korean government, and was last valued around $2.3 billion. The company’s chief financial officer has also said Rebellions is preparing for an initial public offering on South Korea’s main stock exchange, which makes the timing of Nvidia’s interest pointed. A company built to chip away at Nvidia’s dominance in AI chips is, at minimum, worth a serious look from Nvidia itself, whether that ends in a partnership, a stake, or a full buyout.

An Indian grandmaster is one strong day from the biggest final of his career

At the Saint Louis Chess Club, R Praggnanandhaa took a commanding 9-3 lead over Germany’s Vincent Keymer in their Grand Chess Tour Finals semifinal on Monday. The format rewards longer time controls more heavily: each semifinal is worth a maximum of 28 points across two classical games worth 6 points for a win, two rapid games worth 4, and four blitz games worth 2, and a player needs 15 points to advance.

Monday’s classical game turned on a tactical mistake. Keymer had built a promising position with the black pieces, but fell into a sequence that let Praggnanandhaa win a rook for a knight. With an advanced passed pawn and Keymer’s rook effectively trapped, the Indian grandmaster closed the game out decisively. Combined with an earlier result, that 9-3 score puts him six points from a place in the final, in a four player field that also includes Wesley So and Fabiano Caruana, competing for a total prize pool of $450,000. The semifinals wrap up Monday, with the final and third place playoff set for Wednesday through Friday.

It’s the kind of day that shows the AI industry paying itself with one hand while it worries about running out of money with the other. Alibaba is betting a record share sale on AI even as the same spending wrecked its own profit line. Meta is building a trillion dollar AI operation of its own while quietly renting from the competition. Nvidia is deciding whether to buy out a company built to beat it. None of that resolves today, and none of it is supposed to. Meanwhile in St Louis, a 21 year old from Chennai spent Monday doing something much simpler: playing better chess than the person across the board.