From October 20 to 22, along the banks of Shanghai's Huangpu River, the 2026 Sustainable Global Leaders Conference returns once again. Nobel laureates and Turing Award winners will take the stage to discuss the future and trends of technology, and the internet-famous professors who "rescue" people from anxiety and involution 鈥?McFarland and Xiang Biao 鈥?will hold a live dialogue!

Limited seats available, first come first served. Let us journey together toward the sustainable covenant! [Click to register] The UK's Financial Times reported on October 13, citing industry estimates, that China's domestic AI chip output is expected to reach 3.5 million units in 2026, while market demand may exceed 5 million units. The figures are industry estimates cited by media, not officially disclosed procurement or capacity data from companies, but they still reveal an industry issue worth continuously tracking: domestic computing power construction is accelerating, yet a gap remains between supply and demand, and cloud vendors need to find more effective combinations among in-house chips, external procurement and heterogeneous computing power scheduling.

This also gives Alibaba (BABA) an opportunity. According to Tianyancha, the latest quarterly financial report published in August 2026 shows that Alibaba's cloud and computing services revenue reached 48.437 billion yuan, up 45% year-on-year; AI-related product revenue reached 12.376 billion yuan, achieving triple-digit year-on-year growth for the twelfth consecutive quarter. At the same time, Alibaba continues to advance the commercialization of T-Head's in-house chips and incorporates chips, servers, networks, models and inference systems into its full-stack AI strategy. Growth in AI demand has already been reflected in Alibaba's cloud revenue side, and what needs to be observed next is whether computing power investment can be further converted into customer retention, resource utilization and improvement in operating profit margins. For Alibaba, the value of in-house chips lies not only in improving supply chain autonomy, but also in optimizing the unit cost of AI services; for investors, the growth quality of the cloud business, the return cycle of capital expenditure and the commercialization progress of in-house chips will jointly influence the market's long-term judgment of Alibaba's AI business.

Domestic computing power supply still has a gap, and cloud vendors must simultaneously solve "having computing power" and "using computing power well"

AI infrastructure is entering a stage of two-way expansion of supply and demand. On one end, large model training, enterprise-level inference and AI agent applications are continuously increasing computing power demand; on the other end, advanced chip supply, cluster construction and software adaptation still face constraints. Growth in domestic chip shipments does not mean that all types of computing power have achieved full substitution. Training, inference, general-purpose computing and high-speed interconnection have different requirements for chip architecture and system configuration, and cloud vendors need to allocate resources according to actual workloads. The supply-demand estimate cited by the Financial Times provides an observation window: if an obvious gap still exists between domestic chip output and market demand, cloud vendors will need to combine chips from different sources and different architectures for a considerable period. External procurement can ease short-term supply pressure, while in-house chips may improve long-term resource allocation capability, but there is no simple substitute relationship between the two. Whether chips can be delivered stably, whether software can be migrated smoothly and whether clusters can maintain high utilization will all affect actually available computing power.

This is also why Alibaba continues to emphasize full-stack AI capability. In its June 2026 quarter earnings disclosure, the company said that the latest generation AI processor of the Zhenwu series, the M890, has already been commercialized through Alibaba Cloud services among more than 650 external customers across more than 20 industries including autonomous driving, the internet and financial services, covering workloads such as training, fine-tuning and inference. Once in-house chips have been validated by external customer applications, they are no longer merely an internal R&D capability. However, the number of external customers cannot directly indicate how much revenue the chips contributed, nor can it independently prove their cost advantage. What still needs to be observed going forward is the actual deployment scale, depth of customer use, speed of product iteration and the adaptation effects of the chips in different business scenarios. For Alibaba Cloud, the focus of competition is extending to the system level. The performance of a single chip is certainly important, but if servers, networks, storage and software scheduling are not optimized in coordination, local performance advantages may not necessarily translate into efficiency advantages for the entire system. Especially against the backdrop of rapidly growing inference demand, whether computing power resources can be allocated on demand and whether different chips can jointly handle tasks will affect cloud vendors' costs and delivery capability in serving customers. The domestic chip supply gap also has two sides for Alibaba: in the short term, supply constraints may increase the difficulty of procurement and resource scheduling; in the medium to long term, vendors able to combine in-house chips with cloud platforms, models and applications have an opportunity to form more complete infrastructure capabilities. What is truly worth tracking is whether such capability can continuously translate into commercialization revenue and operating efficiency.

Alibaba Cloud revenue is accelerating, and the next step is to see how capital expenditure translates into operating returns

Alibaba's latest quarterly financial report has already provided clear demand-side data. For the quarter ended June 30, 2026, Alibaba's cloud and computing services revenue reached 48.437 billion yuan, up 45% year-on-year; among this, AI-related product revenue was 12.376 billion yuan, achieving triple-digit year-on-year growth for the twelfth consecutive quarter. The company disclosed that increased adoption of AI-related products is an important driver of the acceleration in cloud business growth. This means AI has become an important part of Alibaba Cloud's growth, and customer demand for AI models, computing infrastructure and related cloud services is gradually being reflected in revenue. For enterprise customers, purchasing cloud services can reduce the complexity of building infrastructure themselves and adjust resource allocation according to actual business needs; for Alibaba, growth in AI demand provides an opportunity to expand the scale of its cloud business, broaden customer relationships and drive product upgrades. But revenue growth is only the first half of the operating chain. AI infrastructure is highly capital-intensive, and chips, servers and data center construction require upfront investment, while revenue is realized gradually along with customer deployment, actual usage and continuous renewals. The faster capital expenditure grows, the more necessary it is to observe whether newly added assets can generate sufficient revenue and cash flow; otherwise, high growth may also come with a longer investment payback period.

Alibaba's capital expenditure in the June 2026 quarter reached 67.678 billion yuan, up 75% from 38.676 billion yuan in the same period a year earlier. The company explained that the increase in related spending was affected by continued investment in AI infrastructure, increased CPU computing power demand driven by customer adoption of AI agents, and rising prices of chip components. From the perspective of operating performance, Alibaba Cloud has already shown signs of growth proceeding in parallel with profit improvement. The company disclosed that adjusted EBITA for the cloud segment increased 133% year-on-year, with the margin rising to about 12%. This provides a positive signal for Alibaba's AI infrastructure investment: expansion of scale and improvement in profitability in the cloud business are occurring at the same time. However, the profit performance of a single quarter still needs to be judged alongside subsequent capital expenditure, depreciation and amortization, product prices and customer structure. Especially as the company continues to expand computing infrastructure, the impact of new investment on profit margins may involve a time lag, and short-term profit improvement should not be directly equated with full realization of long-term returns. Alibaba's advantage lies in possessing multiple links including the cloud platform, models and in-house chips, allowing it to try to reduce service costs through system coordination; its operating challenge is that it needs to continuously prove that these investments can bring repeatable commercial returns. In the coming quarters, whether cloud business revenue growth can be maintained, whether profit margins can be sustained under continued investment and whether cash flow can gradually improve will be important bases for judging the quality of the AI business.

The value of in-house chips must be verified by external customers and the economics of each unit of computing power

Another advantage of Alibaba is that its in-house chips do not exist in isolation. T-Head can provide underlying hardware, Alibaba Cloud handles computing resource scheduling, and models such as Qwen and enterprise-level AI applications can form upper-layer demand. Coordination among different links has the potential to reduce software-hardware adaptation costs, improve computing power utilization and allow customers to complete model deployment and application development within the same technology system. The commercial value of this system ultimately depends on why customers are willing to continue paying. If customers choose Alibaba Cloud only to obtain available computing power, the business remains susceptible to price competition and hardware supply; if Alibaba Cloud can simultaneously provide model services, development tools, deployment environments and enterprise-level solutions, customer migration costs and service stickiness may increase, and the cloud vendor will also have a better chance of obtaining recurring service revenue. In its latest financial report, Alibaba disclosed that the Zhenwu M890 has been commercialized for external customers through cloud services, and heterogeneous chip cluster management capability has also become part of its full-stack AI services. This provides a realistic foundation for the company to continue expanding the application of in-house chips, but more sufficient operating data is still needed to judge the scale of commercialization and economic benefits.

For in-house chips, the indicators most worth observing going forward include revenue contribution from external customers, actual deployment scale, utilization of chips and complete machine systems, and comprehensive service costs under similar workloads. For the AI cloud business, attention also needs to be paid to customer renewals, inference call volume, the share of AI product revenue and cloud segment profit margin. Only if these indicators continue to improve can in-house chips potentially evolve from technical reserves into an operating advantage. Industry competition is also developing in this direction. Cloud vendors develop specialized chips mainly to optimize performance, power consumption and cost according to the characteristics of their own businesses, but specialized architectures usually require supporting software ecosystems, development tools and customer migration capability. External chips have mature software ecosystems and a broad developer base, while in-house chips have the opportunity to achieve better system adaptation in specific tasks. For a considerable period in the future, the two routes may coexist in parallel, and enterprises need to allocate computing power according to different workloads rather than simply pursuing the usage ratio of a particular type of chip. Alibaba's long-term opportunity comes from the synergy between full-stack technology capability and cloud service commercialization; whether its operating performance can improve further needs to be continuously verified by subsequent revenue, profit margin, capital expenditure efficiency and cash flow data. For investors watching China's AI industry, this main line is worth long-term tracking: computing power construction determines how fast the business can expand, while commercialization efficiency determines how much value growth can ultimately retain.