China Cloud Capex Could Hit RMB 800 Billion by 2027, Splitting the AI Chip Market in Two

Chinese cloud capital expenditure could reach RMB 800 billion in 2027, implying RMB 400–480 billion in chip procurement if semiconductors absorb 50–60 percent of that spending. That figure, from China Asset Management Company, frames an AI compute market that is now effectively two markets — one domestic, one overseas — with different drivers, different pressures and, notably, different clocks.

The analysis comes from ChinaAMC, a Beijing-headquartered asset manager founded in April 1998 with RMB 3.15 trillion (US$465.3 billion) in assets under management as of June 30, 2026. Through its Hong Kong subsidiaries and a global partnership network, the firm gives international investors access to China's capital markets, which lends its supply-chain read a certain institutional weight.

Why are the two AI compute markets diverging?

AI infrastructure investment continues to grow strongly in both markets, but the reasons for that growth are drifting apart, according to the analysis. In China, two forces compound: expanding cloud capital expenditure and accelerating domestic substitution. Together they create opportunities across the local AI chip supply chain rather than at any single point in it.

Overseas, the picture is more conditional. Robust AI spending still supports demand, but pricing power, capital efficiency and earnings delivery are becoming the metrics that separate winners from the merely exposed. In plain terms, revenue alone no longer clears the bar.

What underpins China's near-term chip opportunity?

The domestic thesis rests on the capex arithmetic. If chips account for half to three-fifths of cloud spending, the RMB 800 billion 2027 capex estimate translates into an RMB 400–480 billion procurement pool. Foundry capacity, meanwhile, is expected to double by 2027, according to the analysis. Two variables remain critical: validation of domestic supply share and the direction of overseas supply policy.

The distinction matters. Doubling capacity is a roadmap figure, not booked revenue. Whether it converts into share depends on qualification cycles and on whether foreign supply remains available at all.

Which catalysts should investors watch?

ChinaAMC points to a short, concrete list:

• Q4 shipments from Huawei and Cambricon, the two most visible domestic AI compute suppliers.
• Progress in domestic HBM qualification, the memory bottleneck for AI accelerators.
• Cloud vendors' December capex guidance, which will test the 2027 spending trajectory.
• CXMT's HBM, expected to enter small-scale production in Q4.

The CXMT milestone deserves emphasis. High-bandwidth memory has been among the tightest links in the AI supply chain, and small-scale production is the first step from laboratory claim to commercial fact.

Where does overseas pricing power sit?

The overseas story has one firm datapoint: PCB supply constraints. Lead times in some segments have stretched to approximately six months, supporting price increases across both materials and finished boards. For substrate and board suppliers, that is genuine pricing power — the kind that shows up in gross margin rather than in press releases.

The offsetting risk is hyperscaler free-cash-flow pressure. Disappointing AI returns, ChinaAMC notes, could affect supply-chain valuations. The logic is straightforward: if the buyers of all this silicon cannot convert spending into cash flow, the chain behind them gets repriced.

What is the verdict on timing?

Despite near-term volatility, both markets retain a positive medium-term outlook, the analysis concludes. The asymmetry is short-term: domestic AI compute enjoys the better near-term outlook, courtesy of state-backed substitution and committed capex, while overseas suppliers must prove that pricing power survives investor scrutiny of returns. The December capex guidance and the Q4 shipment data will tell us which half of the tale is running ahead of schedule.