Alibaba Dilutes Shares To Feed Its AI Spending Spree

Our analysts just identified a stock with the potential to be the next Nvidia. Tell us how you invest and we'll show you why it's our #1 pick. Tap here.

You know the old saying, "you've gotta spend money to make money?"

Historically, that plays out in one of two ways: you spend money and make more, or you go broke. Alibaba, like U.S. hyperscalers Meta and Amazon, is figuring it out the hard way.

Alibaba, China's Amazon, AWS, and PayPal all rolled into one, decided the best way to raise money for AI infrastructure is to dilute its shares: The company announced a $10.2 billion capital raise Monday morning.

Alibaba will sell 710 million shares at HK$112.70 (about $15). The stock immediately fell 10% on the news. Settlement and delivery should be completed in the next couple of days.

This comes after another correction last Wednesday, when the company reported a 75% drop in profit for the June quarter due to heavy AI spending.

Early last year, Alibaba pledged RMB 380 billion (~$53 billion to $56 billion) over the next three years dedicated to AI and cloud infrastructure. To close out 2025, it spent about ~$18 billion. In 2026, projections range from ~$25 billion to $28 billion, leaving ~$10 billion to $12 billion for 2027 and 2028.

Something tells us that it's going to be more, hence the placement of newly issued shares to non-U.S. investors. The question is: will it be enough to get them to the finish line of 2028, or are they going to have to do this again?

Given Alibaba's AI models' scores across several leaderboards, we're assuming the latter. It's a hint that this may not actually be about intelligence, but a cloud computing takeover.

One stock. Nvidia-level potential. 30M+ investors trust Moby to find it first. Get the pick. Tap here.

Alibaba's AI model Qwen is good, but not great.

Alibaba's Qwen 3.8 Max and its open-weight version (Qwen 3.8 2.4T A95B) both landed at 58 on Artificial Analysis's Intelligence Index, a benchmark that ranks 600+ models on overall capability. That puts them a few points behind Claude's Opus 5 and Fable 5 (63 and 62), but ahead of GPT 5.6 Sol and Meta's Muse Spark 1.2. Not bad for Alibaba, especially since one of those Qwen models is fully open-weight and downloadable.

Cost is where Qwen takes a hit, taking 10th and 11th place out of 616 models on Artificial Analysis's "Cost per Intelligence Index Task." OpenAI and Anthropic's models lead in this category. Qwen's high cost per task perhaps points to why Alibaba's profit dropped 75% in June and why they've tapped capital markets for the third time. It's all about compute.