Behind the HK$80 billion placement: Alibaba's AI gamble and the "heavy asset" destiny

In 2026, the global AI competition has entered an unprecedented capital-burning phase.

Amazon's quarterly capital expenditure exceeded 530 billion U.S. dollars, Google posted its first negative free cash flow in 22 years since its listing, and Meta's cash flow plummeted 91% year on year. The four U.S. tech giants are hoarding GPUs and expanding data centers at a near-costless pace. According to estimates from Goldman Sachs, total global AI-related investment alone will exceed 1 trillion U.S. dollars in 2026.

On the other side of the battlefield, Chinese tech giants have no way out either. The combined capital expenditure of Alibaba, Tencent and Baidu in the second quarter of 2026 reached as high as 131.8 billion yuan, about 2.14 times that of the same period last year, with a combined free cash flow outflow of 66.424 billion yuan. Each of them is verifying strategic value with the growth of AI cloud and computing power revenue, and is paying real money for the entry ticket to the AI era.

On August 23, Alibaba made another bet, announcing the placement of 710 million new shares at HK$112.70 per share, raising a total of about HK$800 billion. The proceeds from the placement will be fully invested in the construction of full-stack AI capabilities. This is Alibaba's first new share placement since its return to the Hong Kong stock market for listing in 2019, and it is also the largest primary follow-on offering in the history of the Hong Kong stock market.

After the news was released, the capital market gave two completely opposite "votes": the primary market was oversubscribed in less than an hour, with sovereign and long-term funds accounting for more than 40% of subscriptions; in the secondary market, Alibaba's Hong Kong shares plummeted more than 8% on the same day, and its market value evaporated by more than HK$200 billion in a single day.

What business considerations are hidden behind Alibaba's "abandoning debt for equity" move? After computing power rewrites the cash flow structure of Internet companies, has Alibaba reached a make-or-break moment?

Counter-intuitive Placement

"With 470 billion yuan lying in its accounts, why do you still ask shareholders for money?" This is the most concentrated question raised by the market after the placement announcement was issued.

As of June 30, 2026, Alibaba's cash and other liquid investments stood at 474.505 billion yuan. This figure is enough to make any company considered "not short of money", but Alibaba still chose to issue additional new shares to raise another HK$800 billion from the market, at the cost of diluting the rights and interests of original shareholders by about 3.8%.

Alibaba's explanation is that the company hopes to retain a core cash safety cushion while continuing to increase AI investment. Under the current financing environment, it finally chooses to supplement funds through equity placement. However, secondary market investors obviously do not buy it, and directly voted with their feet. The well-known short seller Michael Burry publicly stated that he "cannot agree" with this additional issuance, and said that he has sold his Alibaba shares, and will only be interested after Alibaba's share price drops by half.

Some market institutions also analyzed the business considerations behind Alibaba's placement. First of all, in the current global AI competition, it is impossible for giants to pour a large amount of cash into the AI track, which is still a bottomless pit of capital burning. Looking at the super giants of Amazon, Alphabet, Microsoft, Meta and Oracle in the United States, they are all raising debts to expand production, and have issued about 244 billion U.S. dollars of bonds globally in 2026.

Secondly, there is still uncertainty about the commercial realization cycle of AI business. Once the short-term output falls short of expectations, high debt will become a heavy burden. Equity financing does not require principal and interest repayment and does not increase financial leverage, which is more suitable for this kind of high-uncertainty, long-cycle heavy investment.

In its latest comment on August 24, Nomura Securities believes that what Alibaba really exceeded expectations is the financing method. Between paying interest to issue bonds and diluting old shareholders for share placement, Alibaba chose the latter. Nomura believes that Alibaba is not incapable of issuing bonds, but the relative attractiveness of debt capital has declined, and this dilution of old shareholders is still within its controllable range.

But it is undeniable that placing new shares to institutions at a discount and diluting the equity of old shareholders is an objective short-term harm. In the eyes of many investors, Alibaba's move is to use shareholders' rights and interests to bet on the future realization of AI business. To reassure the market, Jack Ma, Joe Tsai, Chairman of Alibaba Group, and Wu Yongming, CEO, have increased their holdings of Alibaba's shares by a total of more than HK$800 million in recent days, of which Jack Ma personally increased his holdings by more than HK$600 million.

This move of "paying out of one's own pocket" undoubtedly sends a signal to the market: Alibaba's management has full confidence in the AI strategy.

AI "Money-devouring"

The flip side of "not short of money" is that Alibaba is still continuing to burn money. Compared with the 474.5 billion yuan of cash-like assets lying in the accounts, what is more noteworthy is the shift of its free cash flow from positive to negative.

In the 2026 fiscal year, Alibaba's free cash flow turned completely from a net inflow of 73.87 billion yuan in the previous fiscal year to a net outflow of 46.609 billion yuan. Between the inflow and outflow, there is a cash flow gap of more than 120 billion yuan. Entering the first quarter of fiscal year 2027, the capital burning speed is accelerating instead, and the quarterly capital expenditure jumped to 67.678 billion yuan, a year-on-year increase of 75%. In the same period, operating cash flow was only 22.945 billion yuan, and the net outflow of free cash flow was as high as 44.67 billion yuan.

Capital expenditure is nearly 3 times that of operating cash flow, which means that Alibaba not only invested all the cash earned in the current quarter, but also consumed a large amount of previous cash reserves. Most of this money goes to Alibaba's GPU servers, AI data centers and computing power infrastructure, and part of the profit sacrifice has brought 45% growth to its instant retail business segment.

In the AI track, full-stack AI is regarded as a capital-burning game that only giants can afford to play. The cost of computing power infrastructure alone has made small and medium-sized companies daunted. Among them, hardware procurement and data centers are rigid expenditures, and the prices of computing power chips are dominated by a few suppliers, leaving limited room for procurement negotiation; the total investment in each GW computing power park for data center construction is about 50-80 billion U.S. dollars, which further raises the entry threshold for players.

Looking at the global track, it is not only Alibaba that is under the pressure of continuous AI investment. Data shows that the total capital expenditure of the four giants Microsoft, Amazon, Google and Meta this year has approached 750 billion U.S. dollars. In China, Tencent's capital expenditure in the second quarter was as high as 52.78 billion yuan, a year-on-year increase of 176%, and the net outflow of free cash flow in the quarter was 13.8 billion yuan. Baidu's capital expenditure in the second quarter was 11.39 billion yuan, with a free cash flow outflow of 1.173 billion yuan. Free cash flow has collectively turned negative.

Giants are verifying strategic value with the growth figures of AI business revenue. At the earnings call on August 20, Wu Yongming said that AI has become the core engine for Alibaba Cloud to accelerate growth. The financial report shows that Alibaba Cloud's external revenue increased by 45% year on year, and revenue from AI-related products has achieved three-digit growth for the 12th consecutive quarter.

The growth rate is accelerating, and money is continuing to burn. In February 2025, Wu Yongming announced that at least 380 billion yuan will be invested in the construction of cloud and AI infrastructure in the next three years. As of the end of the June quarter, a total of 190 billion yuan has been invested, and the progress is just over half. And he reiterated at the earnings meeting in May this year that Alibaba Cloud has set a "very high revenue target" for the next five years, and investment in AI infrastructure will far exceed 380 billion yuan.

According to Alibaba's capital expenditure in the first quarter of fiscal year 2017, this HK$800 billion placement is only enough to cover about one quarter of expenditure.

Heavy Asset "High-stakes Bet"

The pressure on free cash flow explains why Alibaba needs to raise more money from the market. A deeper proposition is: this company that started with an asset-light platform economy is embarking on a transformation path to become a heavy-asset AI infrastructure provider.

Over the past two decades, Internet platforms have long been regarded as asset-light companies, and Alibaba is a model of the asset-light route. Taobao connects buyers and sellers, Cainiao Network integrates logistics resources, and Alibaba Cloud provides computing power services. The core is to leverage social resources through platforms and technologies, rather than making heavy investments on its own.

But in the AI era, this business logic has been completely rewritten - to train large models, you must buy GPUs; to provide AI computing power, you must build data centers; to reduce costs, you must self-develop chips. The common features of these investments are extremely high upfront costs, extremely long payback periods, and extremely fast asset depreciation.

In Wu Yongming's view, the business model of full-stack AI platform services is essentially a heavy-asset business model. All AI monetization methods are based on the AI computing power center, and the premise of achieving rapid growth requires upfront investment in capital expenditure. Alibaba's rising capital density interprets this point.

This transformation can also be seen from its business architecture. In this quarter, Alibaba re-divided its business into four major segments: Alibaba E-Commerce Group, AI Cloud and Computing Power Services, AI Labs and Applications, and "All Others". The Cloud Intelligence Group and T-Head Semiconductor are included in "AI Cloud and Computing Power Services", while the Qwen model, Qwen App and Qwen Office form "AI Labs and Applications".

Alibaba is accelerating the integration of the entire AI industrial chain from chips to computing power, from models to applications. The results after such investment are also very significant. In this quarter, Alibaba's AI Cloud and Computing Power Services revenue reached 48.437 billion yuan, a year-on-year increase of 45%, hitting a new high in 22 quarters. Among the 49.5 billion yuan of annualized recurring revenue (ARR) from AI-related products, the ARR of model and application services including MaaS has exceeded 16 billion yuan. The adjusted EBITA profit margin of AI Cloud and Computing Power Services rose to 12%, indicating that AI investment has begun to be converted into profits, and it is no longer in the pure capital-burning stage.

Wu Yongming expects that the annualized revenue of AI-related products in the next quarter will approach 10 billion U.S. dollars. And he gave a clear judgment: "According to the current average gross profit margin of Alibaba Cloud's AI products, the investment in AI computing power can be paid back within three years. As the gross profit margin of AI products continues to increase and the proportion of self-developed chips for replacement increases, the payback period is expected to be shortened to 2.5 years or even 2 years in the future."

But the other side of "heavy asset transformation" is equally cruel.

The upstream makes money by selling computing power, while the downstream burns money by developing applications. In Q1 of fiscal year 2027, Alibaba's AI Labs and Applications revenue was 3.338 billion yuan, but the adjusted EBITA loss was as high as 13.861 billion yuan, and the loss scale expanded by more than 10 billion yuan in one year. The financial report directly stated that the expansion of losses is mainly due to the investment in AI capabilities and the increase in reasoning costs related to the Qwen App.

Alibaba Cloud is profitable to support the group, while the Qwen App keeps burning money, which is a realistic portrayal of Alibaba's "heavy asset transformation". The profits of the entire group are being squeezed by both the upfront investment in AI infrastructure and the continuous capital burning of AI applications. Before the returns come, all players in this heavy-asset AI arms race have no way out.

This article is from the WeChat official account "Digital Tech Society" (ID: sktxs0), written by Bei Ye, and published with authorization from 36Kr.