Alibaba’s AI Bet Is Growing Fast—but Crushing Near-Term Profit

Alibaba’s June-quarter results reveal a clear strategic trade-off: the company is accepting a steep near-term earnings hit to build an integrated AI platform. Revenue growth and cloud demand provide evidence that customers are buying AI capacity, but the quarter also shows that Alibaba has not yet converted that demand into group-level profitability.

The headline numbers

For the quarter ended June 30, 2026, Alibaba reported revenue of RMB268.95 billion, up 9% year over year. Net income fell 75% to RMB10.44 billion, while adjusted EBITA declined 30% to RMB27.33 billion. 67

The earnings miss was concentrated in profitability rather than sales. Alibaba’s adjusted earnings per American depositary share came in below analyst expectations, as higher AI-related costs and capital spending weighed on results. 910

The quarter’s key figures are:

• Group revenue: RMB268.95 billion, up 9% year over year. 211
• Net income: RMB10.44 billion, down 75%. 67
• AI cloud and computing-services revenue: RMB48.44 billion, up 45%. 48
• Capital expenditure: RMB67.68 billion, up 75%. 68
• Model and application services annual recurring revenue: more than RMB16 billion. 8

That combination matters. Alibaba is demonstrating demand for AI services while simultaneously spending heavily on the computing, semiconductor and data-centre capacity needed to supply them. The result is a business with stronger AI revenue momentum but weaker consolidated earnings and cash generation in the short term. 36

AI cloud is the clearest commercial proof point

AI cloud and computing services were the strongest part of the quarter. Revenue rose 45% to RMB48.44 billion, with growth supported by AI-related products and model-as-a-service offerings. 48

The more than RMB16 billion in annual recurring revenue from model and application services is important because it points to recurring customer demand rather than a one-off infrastructure sale. It does not, however, establish that Alibaba’s overall AI investment has reached an attractive return. The company is still carrying the cost of training, inference, chips and capacity expansion across a much larger investment programme. 8

Alibaba’s own reporting also indicates that AI-related products are becoming a meaningful part of external cloud revenue. One earnings-call summary put AI-related product revenue at RMB12.4 billion, or 35% of Alibaba Cloud’s external revenue, although the supplied reporting describes the segment and product definitions differently across sources. 10 The consistent conclusion is that AI is becoming a major cloud growth engine, even as the precise accounting scope requires care.

Why profit fell so sharply

The 75% net-income decline should not be read as evidence that AI demand disappeared. The supplied results instead attribute the pressure to the cost of accelerating the AI build-out, alongside other charges and expenses. Capital expenditure alone rose 75% to RMB67.68 billion. 68

Alibaba said the spending was driven in part by increased CPU-compute capacity, the timing of customer purchases and higher prices across a broad range of chip components. Additional reporting also points to regulatory and impairment-related charges affecting profitability. 717

This is the central financial tension in the results:

1. Demand is arriving: AI cloud and model-service revenue are growing rapidly.
2. Capacity must arrive first: Alibaba is spending heavily on servers, chips and infrastructure to meet that demand.
3. Returns come later: management expects AI-related capital expenditure to break even within roughly three years, based on current average gross margins. 4

That timetable is a management expectation, not a demonstrated outcome. The investment case therefore depends on whether revenue growth, utilisation and margins improve quickly enough to justify the scale of the spending.

The full-stack strategy: chips, cloud and models

Alibaba is trying to control more of the AI stack instead of relying solely on commercially purchased chips and third-party infrastructure. Its planned AI and cloud investment for 2026–29 totals RMB380 billion. 4

The company is also working to deploy proprietary T-Head chips in its data centres and replace some commercially procured chips over time. Alibaba said its in-house chips are already being deployed at scale on large server systems for model training and inference.

If the chips deliver competitive performance and reliability, the strategy could give Alibaba greater control over supply, reduce dependence on external processors and potentially improve infrastructure economics. Those benefits remain conditional. Chip performance, software compatibility, production scale and customer demand will determine whether proprietary hardware becomes a margin advantage or simply another costly investment.

The approach also has a strategic dimension in China, where access to advanced foreign processors is constrained by export controls and domestic technology companies are developing alternatives. Alibaba unveiled a new T-Head AI chip in May 2026 as part of that broader domestic-alternatives push.

Models must become more than a cost centre

Alibaba continues to invest in frontier models and AGI despite limited evidence that the most advanced model work is currently producing commensurate profits. That spending is partly offensive—an attempt to create new products—and partly defensive, because model capabilities and inference prices are moving quickly across China’s AI market.

In August, Alibaba unveiled its largest and most capable AI model to date. The launch came as competitors such as DeepSeek applied intense pricing pressure; one research estimate cited by Reuters said DeepSeek’s pricing was more than 100 times cheaper than an Anthropic model.

Alibaba is also exploring ways to monetise open-source models. Reuters reported that the company planned to seek a share of revenue from major users of a future Qwen model when those users commercialise it. That proposal illustrates the challenge: open models can expand adoption, but the provider must still find a sustainable way to capture value from downstream usage.

The company’s move into robotics models points to another possible route. Rather than limiting AI to chatbots, Alibaba is pursuing models that can support agents and machines performing more complex tasks. These applications could eventually support higher-value enterprise use cases, but the quarter’s financial results do not yet prove that they will.

Organisation is part of the commercialisation effort

Alibaba has been consolidating its AI teams and products under a dedicated Alibaba Token Hub business group. The structure brings together model development, model-as-a-service, Qwen and related AI businesses under one umbrella.

The strategic purpose is to connect research, model services, infrastructure and customer distribution more closely. In principle, that can shorten the path from a model breakthrough to a paid enterprise product. It does not automatically reduce spending: the immediate objective is clearer accountability and faster commercial execution while the company continues investing.

Quick commerce adds another margin test

AI is not Alibaba’s only investment-heavy priority. China’s quick-commerce business also expanded rapidly, with revenue reported up 45% to RMB53.3 billion in the quarter. Management is targeting overall profitability for the business by fiscal 2029. 3

That target implies several more years in which Alibaba may be balancing two large strategic programmes: building AI infrastructure and expanding local commerce. Even if both businesses are growing, their combined investment demands can continue to weigh on group margins before they reach the profitability milestones management has outlined.

What investors should watch next

Alibaba’s results do not settle whether its AI strategy will succeed. They establish a set of measurable tests for the next several quarters:

• AI revenue growth: Can cloud and model-service growth remain strong as the customer base expands?
• AI margins: Does higher utilisation improve the economics of inference and infrastructure?
• Capital intensity: Does capex begin to moderate, or does demand require another major spending cycle?
• T-Head execution: Can proprietary chips perform reliably at scale and reduce Alibaba’s dependence on commercial processors?
• Model monetisation: Can Qwen and related services generate durable revenue despite open-source competition and falling inference prices?
• Quick-commerce discipline: Can the business move toward its fiscal 2029 profitability goal without extending the group’s margin pressure?

The most useful reading of the quarter is therefore neither that Alibaba’s AI strategy is already profitable nor that the profit collapse invalidates it. The evidence shows a company with substantial AI demand and a rapidly expanding infrastructure business, but one that is deliberately bringing forward costs and delaying the payoff. Alibaba’s next phase will be judged by whether that spending produces durable margins—not simply by whether AI revenue continues to grow.