Alibaba's AI Is a Hit, but Hard to Turn Into a Moneymaker
The New York Times
Alibaba's open-source AI models are winning fans worldwide because they're cheap and modifiable. Problem is, being popular and being profitable are turning out to be two very different things.
Based on reporting by The New York Times — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Alibaba has a strange problem: everybody loves its AI, and almost nobody pays for it. The company's open-source models have spread fast across the developer world, prized for being cheaper to run than the proprietary systems coming out of OpenAI or Anthropic, and free for anyone to download, tweak, and deploy however they like. That's exactly the kind of adoption story most AI labs would kill for. Yet Alibaba is discovering what open-source hardware and software companies have known for decades: giving something away is a great way to build a user base and a terrible way to build a balance sheet.
The economics here are blunt. When you release model weights openly, you can't easily charge a licensing fee the way a closed-model company can. Anyone with enough GPUs can take Alibaba's work, fine-tune it, and run it as a competing service without sending a cent back to Hangzhou. So the popularity numbers look great in a press release, but they don't show up cleanly in quarterly revenue. Alibaba's cloud division, which is presumably where the monetization is supposed to happen eventually, has to find other ways to convert goodwill into dollars — selling compute, selling enterprise support, selling the surrounding infrastructure rather than the models themselves.
This isn't a new dilemma, it's just a new arena for it. Red Hat figured out how to make billions off free Linux by selling support and services, not the code. Alibaba is effectively betting on a similar playbook: give away the model to win mindshare and lock developers into its ecosystem, then monetize the cloud services, APIs, and enterprise deals built on top. But AI infrastructure is capital-intensive in a way Linux never was — training and serving frontier models costs real money, and that bill doesn't shrink just because the model itself is free.
There's also a geopolitical wrinkle. Open models are Alibaba's most effective tool for winning developers outside China, especially in markets wary of relying on American AI companies. That soft-power angle matters to Beijing as much as it does to shareholders. So even if the direct revenue math looks ugly right now, Alibaba may keep subsidizing this strategy anyway, treating global adoption as a strategic asset rather than a line item that needs to turn a profit next quarter.
My take — AI-written commentary, not fact-checked reporting
Open-weight models were always going to hit this wall — adoption is not the same as revenue, and anyone who thought otherwise hasn't been paying attention to how open-source economics actually work. I like that Alibaba is pushing open models globally because it keeps the closed labs honest on pricing, but let's not pretend this is a business model yet. It's a subsidized land grab, and someone in Hangzhou is going to have to answer for the cash burn eventually.
Read more about this at: The New York Times