Alibaba's AI has undergone a major critical test, surging forward at a breakneck pace while receiving continuous massive resource support.

Artificial intelligence continues to evolve at an unprecedented pace, even exceeding the predictions of many participants in the AI track.

At the Yunqi Conference last September, Wu Yongming, CEO of Alibaba (9988.HK), outlined the blueprint for future development in his keynote speech: to embrace the arrival of the ASI (Artificial Superintelligence) era, compared with 2022, the first year of GenAI, the energy consumption scale of Alibaba Cloud's global data centers will increase 10 times by 2032.

According to UBS's estimates at that time, the total capacity (IT load) of Alibaba Cloud's data centers in 2022 was about 1.5GW. Based on the 10-fold growth target, the total capacity will reach 15GW ten years later, corresponding to an additional computing power of about 13GW.

Just one year later, Alibaba has further ramped up its computing power layout.

At the just-concluded 2026 Yunqi Conference, Wu Yongming made further statements on the construction of AI cloud infrastructure: current customer AI demand is very strong, and medium and long-term demand far exceeds supply capacity. Alibaba will work with all partners to fully invest in AI infrastructure construction. The goal is that by 2032, the scale of global data centers operated by Alibaba Cloud will exceed 20GW.

This also means that Alibaba will continue to increase its investment in AI business.

In February 2025, Alibaba announced an investment plan that its capital expenditure in the next three years will reach 380 billion yuan. By the end of June this year, the company's cumulative investment has reached 190 billion yuan, and the aforementioned budget has been more than half consumed. According to Goldman Sachs' estimates, building a 20GW data center by 2032 means adding 1-2GW of new capacity each year in the next few years, corresponding to an annual capital expenditure of about 200 billion to 300 billion yuan.

As of the end of June this year, Alibaba's total cash and other liquid investments amounted to about 4745 billion yuan. Are these funds sufficient to support Alibaba's AI ambitions?

Where Will the Capital Come From?

With the moat of e-commerce, Alibaba's cash flow has always been stable. In the six years from FY2019 to FY2024, except for FY2022 when free cash flow fell below 100 billion yuan, the free cash flow of the company in the other five fiscal years all stood above the 100 billion yuan mark, and even exceeded 1700 billion yuan in FY2021 and FY2023.

The turning point occurred in FY2025. Since this year, Alibaba has significantly increased its investment in AI. The company's free cash flow dropped to 78.37 billion yuan that year. In FY2026, AI investment superimposed on the "takeaway war" even turned the company's free cash flow from positive to negative, with a net outflow of 46.61 billion yuan for the whole year. In the first quarter of FY2027 (the second quarter of this year), the company recorded another net outflow of 44.67 billion yuan in free cash flow.

Alibaba's large-scale investment in AI and computing power infrastructure is inseparable from the continuous "blood transfusion" of the e-commerce segment.

For a long time, e-commerce business has been Alibaba's core business and the main source of the company's profits. Even under the background of the "takeaway war" in FY2026, Alibaba China E-Commerce Group still contributed 107.51 billion yuan in adjusted EBITA to the company.

In the first quarter of FY2027, Alibaba merged some businesses under the former China E-Commerce Group, International Digital Commerce Group, Freshippo and Cainiao into the Alibaba E-Commerce Group segment. The adjusted EBITA of this segment in a single quarter also reached 39.75 billion yuan.

However, it cannot be ignored that with the change of consumption environment and the improvement of competitiveness of e-commerce platforms such as Douyin and Pinduoduo, the profit growth of Alibaba's e-commerce business has faced phased bottlenecks, and there is limited room for upward growth in the short term.

In the first quarter of FY2027, Alibaba significantly reduced its investment in the flash delivery business, and the loss of this business narrowed significantly. However, compared with the same period of the previous year, the overall adjusted EBITA of Alibaba's e-commerce segment still declined by 1%. Jiang Fan also said on the conference call that the instant retail business is expected to achieve overall profitability by 2029.

Under the background of a substantial increase in AI-related capital expenditure, the e-commerce business cannot provide incremental cash flow support, and the company also needs to reserve funds as a safety cushion for future development. Therefore, seeking external capital injection has become an inevitable choice for Alibaba.

In 2025, Alibaba issued bonds twice. Among them, in July, it issued about 12 billion yuan of convertible bonds linked to Alibaba Health's stocks, and the raised funds were mainly invested in cloud infrastructure and international business. In September, the company issued another US$3.2 billion (about HK$25.1 billion) zero-coupon convertible senior notes, 80% of the raised funds were invested in cloud infrastructure, and 20% were invested in international business operations.

In August this year, Alibaba launched a placement to raise funds, with a fundraising amount of HK$800 billion, which is also the first placement since the company's listing on the Hong Kong Stock Exchange. All the funds raised this time will be used to invest in full-stack AI capabilities and strengthen the construction of AI infrastructure.

However, the day after the placement was announced, Alibaba's share price plummeted 8.51% to HK$112.5 per share. Since then, the company's share price has fluctuated downward. As of the close on September 24, the latest share price of the company is HK$110, with a total market value of about HK$2.19 trillion.

In the next few years, Alibaba still needs to continuously inject capital into its AI business, but from the feedback of the capital market, it is difficult for the enterprise to pass all the investment pressure on to the external market. In this context, whether the self-sustainment capability of Alibaba's AI business can be improved will become the core factor affecting the subsequent development direction of the company's AI business.

Self-sustainment Capability

In the second quarter of this year, Alibaba re-adjusted its business segments.

At present, among Alibaba's four major business segments, AI business accounts for two segments, namely the AI Cloud and Computing Power Service segment, and the AI Lab and Application segment, which reflects the core position of AI in the company's business.

At a time when Alibaba's computing power infrastructure is expanding rapidly, Alibaba Cloud's performance is growing rapidly, becoming one of the few segments in Alibaba's AI business that can achieve self-sustainment. In the second quarter of this year, Alibaba's AI Cloud and Computing Power Service segment achieved revenue of 48.44 billion yuan, a year-on-year increase of 45%; adjusted EBITA was 5.63 billion yuan, a year-on-year surge of 133%.

The AI Lab and Application segment, which includes the AI model lab and Qwen-related businesses, continued to lose money. In the second quarter of this year, its revenue increased by 16% year-on-year to 3.34 billion yuan, with an adjusted EBITA of -13.86 billion yuan, and the loss expanded by 330% year-on-year.

In Wu Yongming's view, when a technology is in the early stage, especially in the case of supply shortage, a large amount of value often concentrates on infrastructure and core hardware. This is manifested in the current AI field, that most of the value concentrates on chips and AI cloud infrastructure.

With the growth of Alibaba's computing power infrastructure, Alibaba Cloud's business has great room for growth in the future. Some analysts estimate that if each GW corresponds to about US$120 billion in revenue, Alibaba Cloud's external revenue is expected to reach about 1.2 trillion yuan by 2032.

At the same time, with the development of AI technology, the commercial value of large models will gradually increase.

Some market views hold that the market scale that AI large models can leverage in the future is much higher than that of pure cloud services. Among them, the search market corresponding to AI Chat is about US$1 trillion, the Coding market is US$5 trillion, and the Co-Work market for knowledge workers is as high as US$50 trillion.

Alibaba is also continuously improving the level of its foundational large models to reach a broader market.

Wu Yongming mentioned at the Yunqi Conference that the first cornerstone of machine intelligence is the AI model. At present, Alibaba's Qwen team is exploring RSI (Recursive Self-Improvement), and continuously promoting research on model architecture and data optimization, planning to train new models with a parameter scale of 5-10T, aiming to complete more complex and longer-range tasks and move towards ASI.

However, the problem is that large model companies are currently facing fierce competition, and how to occupy a larger market share in such a market is still a problem that Alibaba needs to think about.

In addition to large models, Alibaba's AI chip business is developing rapidly, which may provide new opportunities for the company's AI business to reduce costs and increase revenue in the future.

Wu Yongming said that the second cornerstone of machine intelligence is the AI chip. At present, T-Head is making a full layout of data center chips, including the "Zhenwu" series of GPU chips, the "Yitian" series of CPU chips, the "Panmai" series of smart network cards, and ICN interconnection chips. The core chips for building ultra-large-scale AI clusters have been fully covered.

At the Yunqi Conference site, Alibaba also released the new-generation AI chip Zhenwu V900. A single cluster built on it can be scaled up to 500,000 cards, supporting the training and inference of cutting-edge models. This chip is expected to be mass-produced in 2027.

If Alibaba's AI chip business develops smoothly in the future, the cost of the company's AI infrastructure construction will also decrease accordingly.

For Alibaba, AI is a strategic choice for the enterprise to go through the industry cycle. However, the capital market's expectations for AI are becoming increasingly rational. Commercial implementation capability and core competitiveness have replaced simple vision and become the core criterion for evaluating AI business. Whether it can seize the AI commercialization window and promote related businesses to achieve self-sustainment will determine how far Alibaba's AI strategy can go.

This article is from the WeChat official account "Damo Finance" (ID: damofinance), written by Damo Finance, and authorized for release by 36Kr.