Alibaba Cloud made Wan3.0 generally available on Monday, August 24. The day before, the company priced 710 million newly issued Hong Kong shares at HK$112.70 apiece — HK$80 billion, or about $10.2 billion, its first placement since the Hong Kong listing. Four days before that, it reported that fiscal first-quarter net income had fallen 76% to RMB 10.54 billion, roughly $1.55 billion.
That is the whole story in five days: a video model, a capital raise sized almost exactly to one quarter of capex, and an income statement showing what the spending costs. Alibaba is converting its profit margin into position in generative video and AI compute, and the market now has to decide whether that is discipline or panic.
What Wan3.0 actually does
The headline capability is duration and input breadth. Wan3.0 generates clips of up to 30 seconds at up to 1080p, and it accepts far more than prompts and images. Alongside text, image, audio and video references, it takes DOC, XLS, PPT, PDF and Markdown files, plus live web pages. Feed it a slide deck or a spreadsheet and it produces video.
That input list is the differentiated part. Rivals compete on reference count — ByteDance's Seedance 2.5 accepts up to 50 references across images, video and audio — but document- and webpage-to-video is not something Sora, Veo or Seedance ship today. It also points at where the revenue is: not cinematic auteurship, but the unglamorous middle of marketing production.
The model has been in public beta since August 6. Alibaba says users have applied it to short dramas and film production, advertising, tourism promotion, marketing and music videos — an accurate map of China's short-video content economy rather than a research agenda. Pricing runs $0.05 per second at 480p, $0.10 at 720p and $0.20 at 1080p, so a 30-second 1080p clip costs about six dollars.
One caveat, stated plainly: there is no independent quality benchmark for Wan3.0. It has the longest confirmed generation duration in its class and a live endpoint, and it is also the only model in that class with no third-party quality signal whatsoever.
The money
The placement priced at an 8.4% discount to the previous close and dilutes existing holders by roughly 3.7%. Shares fell as much as 10% in early Hong Kong trade. The book closed within hours: demand hit about $28 billion, nearly three times covered, with close to $6 billion from sovereign wealth funds and long-only investors, who took more than 40% of the allocation.
Alibaba said it intends to use 100% of net proceeds on its "full stack" AI capabilities — a category it defines as chips, infrastructure, and the development and deployment of AI models. The phrasing is not incidental. CEO Eddie Wu used almost the same words on the earnings call: "We delivered a strong quarter, driven by the improving commercialization of our full-stack AI capabilities."
The quarter that phrase describes: capex of RMB 67.68 billion, up 75% year over year, roughly $9.98 billion. Free cash flow swung to an outflow of RMB 44.7 billion, against RMB 18.8 billion a year earlier. Alibaba has spent RMB 190 billion of a RMB 380 billion three-year plan — exactly halfway — and pulled its projected payback on AI investment forward to two and a half years from three.
The offsetting numbers are real. Cloud external revenue grew 45%, a 22-quarter high. AI-related product revenue hit RMB 12.4 billion, an annualized run rate above RMB 49.5 billion, and a twelfth straight quarter of triple-digit growth. Wu's framing is that the sequencing is deliberate: "In order to be able to capture that future growth, we first need to make these capex investments to build out the necessary compute capacity."
Discipline or desperation
The bear reading is not about AI at all. Nicholas Mugalli, founder and CEO of World Trade Securities, argued that resorting to equity dilution "proves that balance sheet burn driven by aggressive AI Capex and price wars in ecommerce—finally caught up," and that "Alibaba's core cash engine can't fund its multifront war against $PDD and Tencent on its own anymore." In that version the raise is not offense; it is the commerce business no longer covering the bill.
Charles Wang, chairman of Shenzhen Dragon Pacific Capital Management, put the same tension more neutrally, saying the placement "dilutes shareholders' interest" and that "investors generally don't like capex … though the investment is beneficial in the long term."
Both can be true. Raising $10.2 billion into three-times-covered demand, with sovereign funds anchoring, is a sign of access, not distress. Needing to raise it in the same month capex matched the raise dollar for dollar is a sign that internal cash flow no longer clears the hurdle.
There is also a strategic wrinkle the market has mostly ignored. Alibaba built enormous global goodwill by open-weighting Qwen and earlier Wan models. That line has quietly stopped for video: no core weights for a Wan video model since Wan 2.2 in July 2025, and Wan3.0 shipped as a closed, paid beta. The company that made "Chinese open weights" a category is now monetizing its best video model exactly the way OpenAI and Google monetize theirs. When you are funding capex with equity, free distribution gets harder to justify.
What to watch
Three things. Whether AI product revenue keeps compounding at triple digits once the base is RMB 50 billion annualized rather than RMB 12 billion. Whether the RMB 190 billion still unspent on the three-year plan comes out of cash flow or out of another placement. And whether independent evaluation confirms Wan3.0's 30 seconds hold up — right now the strongest claim about the model is its runtime, and Alibaba is the only one making it.
One more: OpenAI's Sora 2 and sora-2-pro API endpoints are scheduled to shut down on September 24, with no announced replacement. Alibaba just shipped a longer model with a live endpoint into that gap. The timing may be the most disciplined thing in this entire story.
“In order to be able to capture that future growth, we first need to make these capex investments to build out the necessary compute capacity.”— Eddie Wu, CEO, Alibaba Group