Alibaba's cloud and Qwen momentum is real, but the bill is now showing up clearly in profit and cash flow.
Alibaba did not just report a weak quarter. It showed what China's AI race is starting to cost, even for one of the country's strongest internet companies.
For the March quarter, Alibaba posted revenue of RMB243.38 billion, up 3% from a year earlier, but swung to an operating loss of RMB848 million from operating income of RMB28.47 billion in the same quarter of 2025. That is the kind of turn investors notice because it is not about one soft product line or one accounting adjustment. It is about the price of building the next business while defending the old one.
According to Alibaba's May 13 results, adjusted EBITA fell 84% to RMB5.10 billion, while non-GAAP net income collapsed to RMB86 million from RMB29.85 billion a year earlier. Free cash flow moved in the wrong direction as well, turning into an RMB17.3 billion outflow compared with an RMB3.74 billion inflow last year. Alibaba pointed to quick commerce, Qwen app user acquisition and higher cloud infrastructure spending as the main reasons.
That combination matters because it tells a more useful story than a generic China tech earnings miss. Alibaba is still growing in the places it wants to grow. The problem is that those places are expensive.
Cloud Intelligence Group revenue rose 38% to RMB41.63 billion in the quarter, with revenue from external customers up 40%. AI-related product revenue reached RMB8.97 billion and delivered triple-digit growth for the eleventh consecutive quarter. For a company trying to prove that AI can become more than a market narrative, those are not small numbers.
Alibaba has also been pushing Qwen deeper into its consumer and enterprise ecosystem. The company said it integrated Taobao and Tmall services into the Qwen app, launched a Qwen Shopping Assistant inside Taobao, and expanded Model Studio offerings for developers and companies. Model Studio's customer base grew eight-fold year over year as of March 2026, helped by demand for model access, token plans and AI agents.
This is the payoff side of the investment. Alibaba is not simply buying GPUs and hoping for attention. It is trying to connect cloud infrastructure, foundation models, commerce data, developer tools and consumer applications into one stack. If that works, Alibaba Cloud becomes more than rented computing capacity. It becomes the place where Chinese companies build and run AI products.
The company is also leaning on its own chip design work through T-Head. Alibaba said more than 100,000 Zhenwu PPUs have been deployed on Alibaba Cloud's public cloud platform, with automakers and autonomous driving companies using them for research and development. That matters in China, where access to advanced chips remains strategically sensitive and where local cloud providers are under pressure to prove they can offer scale without depending completely on foreign hardware.
The Margin Pressure Is Not Temporary Noise
The harder part is that Alibaba's AI push is landing at the same time as a costly fight in quick commerce. Quick commerce revenue rose 57% to RMB19.99 billion, driven by the rollout of Taobao Instant Commerce, but the broader China e-commerce group's adjusted EBITA fell 40% to RMB24.01 billion. Sales and marketing expenses climbed to 21.9% of revenue from 15.3% a year earlier, mainly because of quick commerce and Qwen user acquisition.
That is an uncomfortable mix. Alibaba is spending to acquire AI users before monetization is fully mature, and it is spending to defend consumer habits in food delivery and instant retail against Meituan and JD.com. One set of costs is about the future. The other is about not losing relevance in the present.
Net income still rose to RMB23.50 billion, but that headline was helped by mark-to-market investment gains and last year's disposal losses from Sun Art and Intime. Investors usually look through that kind of boost, which is why the RMB86 million non-GAAP net income figure carries more weight. It shows how little core profit was left after the spending wave moved through the income statement.
The cloud segment itself is not the weak link. Its adjusted EBITA increased 57% to RMB3.80 billion, helped by revenue growth and operating efficiency, even as Alibaba continued to invest in customer growth and technology innovation. The issue is whether that improvement can scale fast enough to absorb spending elsewhere, especially as Qwen becomes a consumer product with real acquisition costs.
China's large technology companies are now facing a familiar AI problem. Everyone wants to own the platform layer, but the early economics favor the companies willing to spend first and explain margins later. Tencent, Baidu, ByteDance, Zhipu and DeepSeek have all pushed the market toward faster models, cheaper access and more agentic tools. That makes the user experience better. It also makes pricing power harder to prove.
Alibaba has enough cash to keep going. It ended March with RMB520.82 billion in cash and other liquid investments, and its board approved about US$2.5 billion in annual dividends. So this is not a liquidity story. It is a discipline story.
The next test is whether Alibaba can turn Qwen usage, Model Studio adoption and AI cloud demand into durable, higher-margin revenue before investors lose patience with the spending curve. Cloud growth says the strategy has traction. The operating loss says traction is not the same thing as profit. For Alibaba, the market will now watch the gap between those two numbers more closely than ever.
Also read: Americans are telling AI startups to slow down and earn trust • AI data centers are testing rural America's bargain with Big Tech • Tencent's revenue miss puts its AI payoff under pressure
