Alibaba Cloud can’t deploy servers fast enough to satisfy demand for AI

Steven

Alibaba Cloud can’t deploy servers fast enough to satisfy demand for AI

Summary

Alibaba Group says demand for AI is outstripping its ability to deploy new servers, forcing the cloud arm to ration GPU access and prioritise customers that use its broader suite of services. CEO Yongming Wu told investors the company is running GPUs — both current models and kit three to five years old — at full utilisation, citing accelerating and deepening enterprise AI adoption across R&D, manufacturing and customer-facing functions. Alibaba reported RMB120 billion (~US$16bn) in AI-adjacent capital expenditure over the past 12 months and flagged a likely revision of its RMB380 billion three-year AI spend plan.

Key Points

  • Customer demand for AI is growing faster than Alibaba Cloud can deploy servers, creating capacity constraints.
  • Alibaba is rationing GPUs and prioritising customers that use its full cloud stack (storage, big data and other services) over simple GPU renters.
  • The company says GPUs of all vintages are running at full utilisation, which Alibaba cites as evidence that AI demand is genuine and sustained.
  • Alibaba spent RMB120bn (~US$16bn) on AI-related capex in the last 12 months and may increase its RMB380bn three-year budget.
  • Cloud Intelligence revenue rose 34% year-on-year to US$5.6bn for the quarter; the broader Alibaba Group reported US$34.8bn revenue and US$2.95bn net income, with overseas ops turning a profit.
  • Executives did not explicitly link the infrastructure shortfall to US export controls on advanced accelerators, leaving supply-chain effects ambiguous.

Content summary

On the Q2 earnings call CEO Yongming Wu emphasised that AI demand is both accelerating and broadening across enterprise functions. Because deployment of new servers cannot keep pace, Alibaba Cloud is rationing GPU resources — giving higher priority to customers who consume multiple Alibaba Cloud services. The company maintains that high utilisation across its GPU fleet demonstrates real demand rather than a speculative bubble. Alibaba disclosed sizeable recent capex and signalled it may need to increase its planned multi-year AI investment. Financially, the cloud unit grew strongly and the group delivered solid quarterly revenue and profit.

Context and relevance

This is a clear indicator of two industry trends: (1) hyperscalers and cloud providers are wrestling with physical GPU capacity limits as AI workloads scale, and (2) vendors are moving to favour integrated, higher-value customers when supply is tight. For enterprises, the move means potential access constraints, higher spot pricing or longer lead times for GPU-based services. For the GPU market and suppliers, sustained utilisation and rising capex commitments suggest continued demand pressure that will shape procurement, regional supply chains and pricing. Competitors such as Tencent and other cloud providers are facing similar procurement and deployment challenges.

Why should I read this?

Short version: if you buy cloud AI capacity or run inference/training workloads, this matters — big time. Alibaba is basically telling customers: there simply aren’t enough GPUs to go round, so expect prioritisation, delays or price pressure. We skimmed the call and pulled the bits that affect planning, procurement and risk for cloud users, so you don’t have to.

Source

Source: https://go.theregister.com/feed/www.theregister.com/2025/11/26/alibaba_q2_2025/