Alibaba Is Raising $10.2B to Fund Its Full AI Stack
Matthew Leo · Published August 29, 2026 · AI
Alibaba is asking investors for another $10.2 billion, and it says every dollar of the net proceeds is going into artificial intelligence.
The August 23 Hong Kong share placement is remarkable for its size and for how broadly Alibaba defines the AI bet. The company is not raising money for one model or one data centre. It plans to fund chips, infrastructure, model development and deployment across what it calls its “full stack” AI capabilities.
That makes the deal a useful measure of how expensive serious competition in AI has become.
This is a record-sized Hong Kong follow-on offering
Alibaba launched a HK$80 billion placement, worth roughly US$10.2 billion. Reuters reported that it would be the largest primary follow-on offering ever by a Hong Kong-listed company.
The placement represents about 3.6 percent of Alibaba’s outstanding shares. Investors ordered far more stock than was available, although the discounted issue price still sent the shares lower because existing holders are being diluted.
Alibaba says 100 percent of the net proceeds will support AI-related development.
That is an unusually clean declaration of priorities for a company that built its empire in e-commerce.
“Full stack” means Alibaba does not want to rent the future
There are several ways to participate in the AI market. A company can build applications on someone else’s models, rent compute from someone else’s cloud and buy chips from an established supplier.
Alibaba is trying to own far more of the chain.
Its plan covers semiconductor development, cloud infrastructure, foundation models and the services that put those models into products. That approach is partly strategic and partly forced by geopolitics.
U.S. export restrictions limit Chinese access to Nvidia’s most advanced accelerators. Chinese technology companies therefore have a much stronger incentive to build domestic alternatives instead of assuming they can always purchase the world’s best chips.
The money also shows how capital-intensive AI has become
Software businesses used to be celebrated because successful products could scale with relatively little physical infrastructure.
Frontier AI has reversed some of that logic. Training and serving leading models requires data centres, accelerators, memory, networking equipment and enormous amounts of electricity. The companies trying to compete at the top are beginning to resemble industrial infrastructure businesses as much as traditional software firms.
A $10.2 billion equity raise would be transformational capital for most technology companies. In the current AI race, it is one funding layer inside a much larger spending plan.
Alibaba has already outlined tens of billions of dollars in capital expenditure around AI and cloud infrastructure.
Shareholders are paying for the bet today
The strategic logic does not make dilution disappear.
New shares reduce the ownership percentage represented by each existing share unless the capital produces enough future value to compensate. Alibaba priced the placement at a discount, which is normal for large offerings but still painful for current investors.
The company therefore has to do more than show that AI is important. It has to earn returns on a very large capital base.
That is the challenge facing nearly every hyperscaler. The market has largely accepted huge AI spending because demand remains strong. Eventually, investors will want to see those investments produce sustainable revenue and margin rather than another round of infrastructure spending.
China’s AI economics are different from America’s
U.S. companies can generally buy the newest Nvidia systems, although supply and cost remain constraints. Chinese companies have to design around export controls, domestic chip capabilities and a different cloud market.
That can produce a surprising advantage in efficiency.
When the best hardware is scarce, developers have a stronger incentive to squeeze more from every chip, optimize model architectures and build software around locally available accelerators. Chinese AI labs have already shown they can remain competitive without matching U.S. infrastructure spending dollar for dollar.
Alibaba’s capital raise suggests it does not intend to rely on efficiency alone. It wants both scale and deeper control over the stack.
This matters well beyond China
If Alibaba succeeds in building competitive chips, models and cloud infrastructure as an integrated platform, it gives countries and companies outside the U.S. another major AI ecosystem to choose from.
That has consequences for pricing, standards and geopolitics.
More competition can reduce the power of any one supplier. It can also split the global technology stack along political lines, with different countries favouring U.S.- or Chinese-developed infrastructure based on security rules and trade relationships.
Canada is firmly connected to the U.S. technology ecosystem, but Canadian companies still compete in a global market. Cheaper or more capable Chinese models can influence what customers expect even when Canadian businesses cannot or will not deploy the underlying infrastructure.
The AI race is becoming a financing race
The most striking thing about Alibaba’s $10.2 billion raise is that it does not feel extraordinary in the context of 2026 AI spending.
Nvidia is selling nearly $90 billion of data-centre hardware in a quarter. American hyperscalers are committing hundreds of billions to compute. Governments are funding domestic infrastructure. Now one of China’s largest technology companies is issuing a record amount of stock to make sure it can keep pace.
Better models still matter. So do researchers and software.
But the companies that remain at the frontier increasingly need something else: access to capital on a scale that very few organizations on Earth can match.
Tags: Alibaba, AI, China Tech, semiconductors, Business