Alibaba Raises $10.2B for AI, Launches Wan3.0 Video Model

Alibaba is spending like a company that believes the AI infrastructure race is the only race that matters. Within a single week, the Chinese technology giant priced a $10.2 billion equity placement — described as the largest follow-on share sale in Hong Kong’s history — and launched Wan3.0, a new multimodal AI video model, doubling down on both the money and the models behind its artificial-intelligence push.

The two moves are deliberately linked. Alibaba said it intends to use 100% of the net proceeds from the placement to invest in its full-stack AI capabilities, including expanding and upgrading the data-center infrastructure that trains and serves models like Wan3.0. Taken together, they frame the clearest picture yet of how China’s largest cloud provider plans to keep pace with domestic rivals and Western hyperscalers alike: raise capital aggressively, spend it on compute, and give away or cheaply serve the models that run on top.

A record capital raise

The share placement is the headline financial event. At roughly $10.2 billion, it ranks as the biggest follow-on stock sale ever completed in Hong Kong, a scale that underscores both Alibaba’s ambition and its appetite for external funding to underwrite the AI buildout. The company was explicit that the proceeds are earmarked for AI: infrastructure, model development, and the broader stack of cloud services that sits between the two.

Raising equity to fund AI is a notable choice. It dilutes existing shareholders, and Alibaba is doing it at a moment when its spending is already compressing profits. But it also signals that management views the current window — the land-grab phase of enterprise AI adoption in China — as too important to underinvest in, even at the cost of near-term earnings and per-share value.

Capex is eating the profits

The raise arrives against a backdrop of surging capital expenditure. In the June quarter, Alibaba’s capital spending reached RMB 67.7 billion (about $10 billion), up roughly 75% from a year earlier, with AI-and-cloud capex hitting a record high for the company. That spending has come straight out of the bottom line: net profit fell sharply — reported at around 75% lower year-on-year — as the cost of building AI capacity outran the revenue it currently generates.

The offsetting story is on the top line. Alibaba’s Cloud Intelligence Group has posted accelerating growth, and the company says its AI-related product revenue has grown at triple-digit rates for eight consecutive quarters — the metric management points to when defending the spending. The bet is a familiar one across the industry, laid out in detail in the economics of hyperscaler AI capex: pour capital into infrastructure now, on the conviction that demand for AI compute will grow into it. Alibaba has committed to that thesis at scale, pledging roughly RMB 380 billion — on the order of $53 billion — to AI and cloud infrastructure over three years, and its latest results and share sale show it is spending against that pledge in real time. The full quarterly picture is broken down in our coverage of Alibaba’s June-quarter earnings.

Wan3.0: the model the money is for

If the placement is the capital, Wan3.0 is a piece of what it buys. The new model extends Alibaba’s Wan family of generative video systems, one of the most widely used open lineages in the field, into a broader multimodal flagship. Alibaba made Wan3.0 available in public beta through its Alibaba Cloud Model Studio and Qwen platforms, positioning it as both a product for enterprise customers and a showcase for the infrastructure underneath.

What distinguishes Wan3.0 is the breadth of what it accepts as input. Beyond the text and image prompts typical of video generators, the model can ingest video, audio, web pages, PDFs, and slide decks, turning a wide range of source material into generated video. That “any input to video” framing pushes the Wan family from a creative tool toward a more general content-production engine — the kind of capability Alibaba can sell as a cloud service and wire into its commerce, marketing, and media businesses.

The launch continues a pattern. Earlier in August, Alibaba’s Tongyi Lab open-sourced Wan-Animate-2, a real-time character-animation model released under a permissive license, and the company has steadily expanded its Qwen family of language and multimodal models. Wan3.0 slots into a strategy of competing on both openness and cloud distribution rather than on a single closed frontier model behind an API.

The competitive logic

Alibaba is not spending in a vacuum. Domestic rivals are racing on the same axes — models, cloud, and the chips and data centers that power them — while U.S. hyperscalers continue to expand their own AI infrastructure at a pace measured in tens of billions of dollars per quarter. In that environment, a cloud provider that underinvests risks watching customers migrate to whoever has the most capacity and the best models available on demand.

By coupling a record capital raise to a concrete model launch, Alibaba is telling investors that its AI spending is not open-ended for its own sake but tied to shippable products with a distribution channel already attached. Whether that reassures a market wary of ballooning capex and shrinking profits is the open question. The company is asking shareholders to fund a multi-year infrastructure cycle on faith that AI revenue keeps compounding — the same faith the entire sector is running on.

What it means

Alibaba’s one-two of a $10.2 billion raise and a flagship model launch crystallizes the central tension in AI investing right now: enormous, profit-eating spending justified by fast-growing but still-immature revenue.

For Alibaba, the placement buys runway and firepower, but it raises the stakes. Diluting shareholders to fund capex only pays off if AI and cloud revenue keep scaling fast enough to eventually cover the spend and then some. The eight straight quarters of triple-digit AI revenue growth are the bull case; the 75% profit drop is the bear case. Which number the market fixates on will drive the stock.

For the China AI race, the move raises the capital bar for everyone. When the largest domestic cloud player commits tens of billions to infrastructure and can raise ten billion more in a single placement, smaller rivals face a stark choice between matching the spend and ceding the high-capacity, high-model-quality end of the market. Consolidation pressure follows.

For the broader market, Alibaba is another data point in a global pattern: hyperscalers on both sides of the Pacific are financing AI buildouts through every available channel — equity, debt, and partnerships — and letting near-term margins take the hit. The economics of AI data centers will ultimately decide whether that discipline pays off. For now, watch two things: whether Wan3.0 and its siblings convert into durable cloud revenue, and whether investors keep rewarding capacity over profitability. The moment that trade reverses, the math behind raises like this one changes fast.