Alibaba committed 380 billion yuan to AI and cloud infrastructure in February 2025, called it the largest privately financed compute project in Chinese history, and is roughly a third of the way through it (SCMP). The announcement was framed as ambition. Read against the eighteen months that preceded it, it reads as triage. Pinduoduo set the floor on price, Douyin took the discovery layer, JD opened a logistics flank, and the e-commerce growth that once justified Alibaba’s premium had to be replaced with something the market would still pay for. A capex line in cloud was what was available.

The squeeze on Taobao and Tmall has been measurable since 2024. Pinduoduo gained ground through a value-segment push the incumbent could not price-match without surrendering margin (Kavout). Douyin compounded the other way; its share of livestream commerce overtook Taobao before 2025 began and pulled the discovery layer onto a feed Alibaba does not own (ECDB). JD then opened a third front in food delivery and instant retail, converting new delivery users into JD e-commerce buyers and dragging Alibaba into a subsidy war it had not budgeted (SCMP). Three competitors are pulling at three different parts of the same purse.

Timing is the tell. The 380-billion-yuan commitment landed in February 2025, after Taotian had spent successive quarters below 5% revenue growth (Genuine Impact). When Eddie Wu told analysts the company would “aggressively invest” in cloud, a strategist quoted by SCMP read the number as a forecast rather than a plan: a sum that size required Alibaba to find growth outside its core e-commerce business, or it could not be justified (SCMP). Wu was describing this as ambition. The market read it as obligation.

The institutional moves around the spend tell the same story. In March 2026, Wu reportedly consolidated several AI units into a new Alibaba Token Hub Business Group under his direct control (TechNode). Qwen is being pushed into Taobao, Alipay, and Fliggy as a usage layer; the Quark browser sits as the consumer wrapper; the in-house T-Head chip unit supplies the silicon (DCD). Each move makes the cloud-and-AI line more legible to investors who can no longer underwrite Taotian on growth alone. The narrative has been re-tooled to the extent the business had to be.

The cost of the new strategy is being absorbed by the old one. Q2 FY2026 free cash flow swung from a 13.7 billion yuan inflow a year earlier to a 21.8 billion yuan outflow, with capex up 80% on the year (Nasdaq). Adjusted EBITDA margin compressed in the same window from 17.4% to 3.7%. The marketplace is paying for the cloud build in real time.

The case for reading the spend as strategy rather than defence is not weak. Cloud Intelligence Group reached 34% revenue growth in Q2 FY2026 — an acceleration that nothing else in the portfolio matches (Nasdaq). Qwen crossed 100 million monthly active users by January 2026; Alibaba’s T-Head chip unit has shipped 470,000 AI accelerators, with more than 60% deployed to external customers (SCMP on Qwen, DCD). For the defensive reading to fail, Chinese enterprise AI demand has to absorb the new capacity at margins above the 3.7% the rest of the company is currently delivering, inside a four-to-five-year depreciation window. The condition is not absurd. Microsoft posted 47% operating margins in the same period and Google delivered $34.5 billion in quarterly net income; Alibaba is making the same bet from a thinner balance sheet (Nasdaq).

Defensive capex and strategic capex can be the same line item, in sequence.

The price of the 380-billion-yuan commitment is the admission embedded in it. Three of Alibaba’s four largest competitive threats sit outside cloud. PDD on price, Douyin on discovery, and JD on logistics cannot be answered inside the e-commerce P&L without surrendering more margin than the company has left to spend — so Wu has chosen to grow a different one. That works if the cloud line compounds faster than the marketplace line erodes, and if domestic AI demand absorbs the new capacity at acceptable yield. If either condition slips, the next 380 billion is not a choice. Alibaba reports its full fiscal year on May 13; that print is the first full read on which story Wu has committed Alibaba to.