Alibaba CapEx jumps 75% to $10B as AI infrastructure spending accelerates

Alibaba's capital expenditure jumped 75% year-over-year to RMB 67.7 billion ($10 billion) in the June quarter, as the Chinese cloud giant pours record sums into AI infrastructure to meet surging demand for compute and agentic AI workloads.

"With our full-stack AI strategy, we have put Alibaba in a superior position to capture the substantial growth of demand for artificial intelligence and AI compute," CEO Eddie Wu said in a statement Thursday.

The spending surge — up from RMB 38.7 billion a year earlier — was driven by procurement cycle fluctuations, rising CPU compute capacity tied to customer adoption of AI agents, and higher prices across chip components, the company said. Revenue rose 9% to RMB 269 billion, slightly above the LSEG consensus of RMB 268.9 billion. Cloud division revenue reached RMB 48.4 billion, up 45% year-over-year, with AI-related product revenue posting triple-digit growth for a twelfth consecutive quarter.

The spending trajectory places Alibaba alongside global hyperscalers in the AI capex race, with direct implications for the semiconductor supply chain. As memory bandwidth becomes the binding constraint for AI inference — South Korea's memory chip exports surged 277% year-over-year in July while logic chip exports fell 0.7% — Alibaba's procurement decisions will shape demand for HBM, DRAM, and compute infrastructure through 2028 and beyond.

Qwen Models Anchor Alibaba's Cloud Monetization Push

Alibaba has been aggressive in expanding its AI model portfolio to drive cloud adoption. Earlier this month, the company unveiled Qwen3.8-Max, which it described as its "most powerful" AI model, with benchmark results comparable to or better than Anthropic's Claude 5 in certain tests. The company also released Qwen3.8-27B, a model designed to run on consumer hardware like laptops — a new battleground for AI model developers as inference moves to edge devices.

The model strategy is central to Alibaba's cloud monetization play, mirroring how Microsoft and Google use their AI models to drive Azure and Google Cloud consumption. Alibaba's cloud unit is seen as the key to the company monetizing artificial intelligence, much like its US counterparts.

Memory Supply Constraints Keep AI Infrastructure Costs Elevated

The capex surge comes at a time when the AI infrastructure supply chain is under unprecedented strain. High Bandwidth Memory (HBM) — the stacked DRAM technology that delivers up to 2 TB/s of bandwidth per stack, roughly 33 times that of conventional DDR5 — is the binding constraint for AI inference workloads. Every HBM stack requires approximately three times the fabrication capacity of an equivalent DDR5 unit, and all three major memory suppliers (Samsung, SK Hynix, and Micron) have already committed their full 2027 output.

SK Hynix's board approved ₩54.3 trillion ($38.3 billion) in new fabrication spending on August 8, with new capacity not expected to reach market until mid-2029 at the earliest. Counterpoint Research's Neil Shah told CNBC that memory prices are "unlikely to soften before the end of 2028." For Alibaba and other hyperscalers, this means the cost of AI infrastructure will remain elevated for the foreseeable future.

Alibaba's US-listed shares fell 3.1% in premarket trading Thursday as investors weighed the profit impact of the spending surge. The company reported a 75% fall in June-quarter profits, with the capex increase weighing on margins. However, the market's reaction was muted compared to the initial 4% drop, suggesting investors are beginning to price in the long-term growth potential of AI infrastructure investment.

For investors, the key question is whether Alibaba's aggressive capex will translate into sustained cloud revenue growth. The company's cloud division grew 45% year-over-year, and AI-related product revenue has now posted triple-digit growth for twelve consecutive quarters. But with memory prices expected to remain elevated through 2028 and HBM supply constrained until mid-2029, the margin pressure from AI infrastructure spending is unlikely to ease soon. Alibaba's bet is that the demand for AI compute — particularly from agentic AI workloads that require persistent memory states — will outpace the cost curve.

This article is for informational purposes only and does not constitute investment advice.