Alibaba Cloud Growth Hits 34% as AI Demand Drives Spending and Profits Face Pressure

AI demand lifts cloud revenue above the group pace

Alibaba's cloud revenue rose 34 percent from a year earlier to 39.8 billion yuan, about $5.6 billion, in its fiscal second quarter, as customers spent more on training and running artificial intelligence models. The performance surpassed expectations and put the cloud division well ahead of the company's revenue growth of 5 percent.

Alibaba reported group revenue of 247.8 billion yuan, about $34.8 billion. Cloud therefore accounted for roughly 16 percent of the total, calculated from those two revenue figures. Its much faster growth shows how AI is changing the mix of a business whose core commerce operations also continued to expand.

AI related product revenue recorded growth of at least 100 percent for the ninth consecutive quarter. That means revenue in this category was at least twice its level a year earlier in each of those quarters. Cloud profitability improved too: the division's EBITA rose 35 percent to 3.6 billion yuan.

EBITA is an earnings measure before interest, tax and amortization, an accounting expense associated with certain assets. It gives readers another way to look at operating performance, alongside revenue. Alibaba's cloud division increased both measures, while heavier spending elsewhere put pressure on group earnings.

Investors initially focused on the cloud figures as evidence that spending on AI infrastructure and software was turning into customer demand. Alibaba shares rose in early U.S. trading before fluctuating as the market weighed the growth against the cost of the company's investment plans.

As of November 27, 2025, Alibaba's central trade off was visible across its results: cloud customers were buying more AI capacity, while investment in instant commerce, data centers and AI infrastructure was weighing on profits and cash flow.

Why AI brings more business to the cloud

AI models need computing power, storage and networking on a scale that can exceed the demands of traditional business software. Training a large language model, the technology behind many chat assistants, requires clusters of advanced graphics processing units, or GPUs, working for days or weeks.

Training is the stage in which a model learns patterns from data. Once that model is ready, it still needs computing resources whenever it answers a question or performs a task. This second stage is called inference. Search, recommendations, chat assistants and productivity tools can create substantial demand for inference capacity when many people use them.

Alibaba Cloud earns revenue by supplying those resources. Customers using more computing power can also need more storage and networking, increasing their consumption of cloud services and, in many cases, the revenue Alibaba receives from each customer.

Businesses in finance, logistics, manufacturing and retail are testing and deploying AI assistants, search tools and systems that generate computer code. Those applications bring demand for other cloud products, including databases, security, content moderation and elastic computing, which lets customers adjust computing capacity as their needs change.

Vector search is another part of this wider demand. It helps an application find information by similarity or meaning, making it useful when an AI system needs to retrieve relevant material from a company's records. These supporting services extend the business opportunity beyond simply supplying the chips that run a model.

Alibaba Chief Executive Officer Eddie Wu linked the acceleration to strong customer adoption and said the company was ready to increase its commitment if demand continued.

Customer demand for AI remains very strong. If demand continues, we are prepared to increase AI investment beyond our previously announced plan.

A 380 billion yuan investment plan could grow

Alibaba previously outlined a plan to invest 380 billion yuan over three years in cloud and AI infrastructure, making AI a central part of its strategy for the longer term. Over the past four quarters alone, it deployed about 120 billion yuan into the capacity and technology needed to support that business.

The spending covered data centers, networking, storage, AI accelerators and the software that makes these systems efficient and reliable at scale. AI accelerators are processors designed to handle the calculations involved in training and running models. Their usefulness depends on the surrounding systems being able to move and store data quickly enough.

Management said capital needs could rise further if customers continued adopting AI at the current pace. Wu also rejected the view that AI was a passing fad or an overvalued investment theme. He argued that limited supply would support pricing and the use of available capacity as businesses moved from pilot projects to production deployments.

Wu, Alibaba's chief executive, described the pressure on infrastructure supply in these terms:

I do not see an AI bubble. Demand will stay high and supply, especially for data centers and semiconductors, will be tight for the next few years.

A pilot project tests whether an AI tool works for a particular job. A production deployment makes it part of everyday operations, such as answering customer requests or helping employees write software. Moving between those stages can turn occasional computing demand into continuing use.

Chips and data centers limit the pace

Alibaba faces practical limits on how quickly it can add computing capacity. Access to advanced chips, tighter export rules between the United States and China, and the time needed to build power and cooling systems all make new capacity harder to bring online quickly.

Data centers need more than processors. The chips must receive a dependable electricity supply and stay cool while operating. Those requirements help explain why strong customer demand can coexist with shortages of available computing resources.

Alibaba is working to stretch its existing capacity through model pruning, quantization, better networking and improved scheduling. Pruning reduces parts of a model that are less useful, while quantization uses simpler numerical representations to reduce the resources needed for calculations. Both approaches aim to make models less demanding to run.

Networking improvements help computing systems exchange data more efficiently. Scheduling determines when and where workloads run, helping available equipment serve more tasks. Alibaba is also using a wider mix of accelerators, including domestic alternatives, to manage costs and changes in supply while maintaining reliable performance.

Cloud leadership described the revenue growth outlook as high, rather than promising further acceleration in the very near term. Component supply, the availability of GPUs and data center slots, quarterly changes in Alibaba's own AI usage, and a demanding comparison with the previous year could all affect the growth rate.

Management's view for the medium term depends on customers moving more AI workloads into production on Alibaba Cloud as models and tools mature. The immediate picture is therefore one of strong demand, with growth also shaped by how much usable capacity the company can supply.

Commerce investment weighs on group earnings

Alibaba's cloud gains came alongside pressure on group profitability. Adjusted EBITA fell sharply as the company spent heavily on instant commerce and AI infrastructure. Adjusted earnings measures modify reported earnings for selected items; here, the direction of the figure showed the strain from the investment phase.

Instant commerce, also described as quick commerce, centers on delivering purchases rapidly. Alibaba increased subsidies to build logistics density while accelerating data center projects to meet AI demand. Logistics density means having enough orders and delivery activity in an area to use the network more efficiently.

Management's message was that Alibaba was prioritizing market share in strategic businesses and computing capacity for AI over immediate improvement in profit margins. That choice affected both earnings and cash flow, even while cloud revenue and cloud EBITA were rising.

China e commerce revenue increased 16 percent, supported by strong order growth, while quick commerce revenue rose 60 percent. Alibaba set a target of reaching 1 trillion yuan in quick commerce gross merchandise value within three years.

Gross merchandise value, or GMV, measures the value of purchases passing through a platform. Revenue measures what the company earns from its business. The quick commerce target is therefore a target for transaction activity, while the reported 60 percent increase refers to revenue.

Management said losses per order were narrowing as greater scale and operational improvements took effect. This is an improvement in unit economics, meaning the revenue and costs associated with an individual order. Alibaba still faced intense competition in e commerce and food delivery, with price promotions across the sector squeezing margins.

Qwen connects consumer interest with developer demand

Alibaba's Qwen app, its consumer rival to ChatGPT, exceeded 10 million downloads in its first week. The rapid uptake showed strong interest in generative AI tools among users in China and gave Alibaba a consumer outlet for its flagship family of large language models.

Generative AI produces material such as text or computer code in response to instructions. Qwen is also part of Alibaba's developer strategy, with models that businesses and software builders can adapt to their own tasks.

The latest releases, including the Qwen3 series, added stronger reasoning and coding capabilities. Alibaba offered a mix of model sizes for different tasks and cost requirements. A company can choose a model that suits the complexity of its work and the computing resources it is willing to use.

Developers can access Qwen through services hosted on Alibaba Cloud or use open source model checkpoints on their own infrastructure. A checkpoint is a saved version of a trained model that developers can use or customize. These choices allow Alibaba to serve customers seeking a managed cloud service while also reaching developers who prefer to operate their own systems.

Alibaba's approach brings together infrastructure, foundation models and application platforms. Foundation models are the underlying models that can support a range of applications. The company aims to let enterprises train, fine tune and deploy AI without having to assemble all the necessary services from different vendors.

Enterprise tools turn models into working applications

Alibaba's Platform for AI combines data integration, training and inference orchestration, vector databases, monitoring and security. Orchestration coordinates the steps and resources involved in running an AI workload, while monitoring helps teams track how their systems are performing.

The platform is intended to shorten development times and reduce the operational work required to run AI in production. Alibaba is seeking to connect these tools to commerce, logistics and office workflows within its own businesses and among external customers.

Lingma, Alibaba's coding assistant, is gaining traction among corporate developers. Customers are also adopting tools for content generation, product search, customer service bots and knowledge management, which involves organizing information so employees or applications can find and use it.

These applications can create continuing demand for cloud services once they become part of a company's routine work. Switching providers can be difficult and expensive when an AI system is connected to business records, software and established workflows.

Open source widens the Qwen ecosystem

Qwen models have been widely downloaded by developers and used as the foundation for many derivative models. A derivative model builds on an existing model and adapts it for a different purpose, rather than starting the entire training process from the beginning.

Alibaba's open source strategy allows a wider developer community to customize Qwen and contribute to its development. That activity helps produce specialized variants for fields including finance, retail and manufacturing, while feedback from developers supports further improvements.

A growing ecosystem of plug ins, retrieval tools and guardrails is helping enterprises put applications based on Qwen into production with greater confidence. Plug ins connect applications to extra functions. Retrieval tools find information that an AI system can use, while guardrails set controls around its behavior or output.

Alibaba's strategy therefore reaches beyond its consumer app. The company is building a network of developers, business applications and cloud services around Qwen, with open source access serving as a way to broaden adoption.

Competition and regulation shape expansion

Alibaba remains China's largest cloud provider, with market share of around one third. Its position is supported by scale, cost efficiency and connections to services across its network of businesses. Those strengths sit alongside a crowded competitive field.

In China, major domestic rivals compete on performance, price and solutions designed for particular industries. Internationally, Alibaba faces hyperscalers, the large cloud operators whose infrastructure serves vast numbers of customers, that have a lead in markets outside China.

Restrictions on advanced U.S. chips add complexity to Alibaba's plans. Its efforts to improve efficiency and diversify accelerator options can offset part of that pressure, while investment in infrastructure remains necessary to serve growing AI workloads.

Outside China, Alibaba is focusing on markets where it sees a match between its products and customer needs, including parts of Southeast Asia. Regulatory oversight is also increasing, especially in Europe, where large platforms face stricter obligations under digital services rules.

Alibaba must manage those compliance demands while expanding its AI products. The company also maintains a sizable cash position and an active authorization to repurchase shares, giving it flexibility as it balances investment, growth and returns to shareholders.

Key Points

• Fiscal second quarter cloud revenue rose 34 percent to 39.8 billion yuan, about $5.6 billion, surpassing expectations.
• Group revenue increased 5 percent to 247.8 billion yuan, about $34.8 billion.
• AI related product revenue recorded a ninth consecutive quarter of growth of at least 100 percent.
• Cloud EBITA increased 35 percent to 3.6 billion yuan, while adjusted group EBITA fell sharply under investment pressure.
• Alibaba spent about 120 billion yuan on cloud and AI infrastructure over four quarters and may increase investment beyond its plan of 380 billion yuan over three years.
• Qwen's consumer app passed 10 million downloads in its first week, alongside expansion of its models and enterprise tools.
• China e commerce revenue grew 16 percent; quick commerce revenue rose 60 percent, with a target of 1 trillion yuan in GMV within three years.
• Chip supply and data center capacity could keep cloud growth high but uneven between quarters.
• Alibaba holds around one third of China's cloud market while pursuing expansion, including in Southeast Asia.