Overseas AI Investment Is Cooling — and the Shockwaves Reach China’s Export Machine
Growth in overseas AI capital expenditure is decelerating, and the ripple effects on China’s exports and its domestic AI supply chain are becoming a central question for macro strategists. Research from CICC argues that the export channel transmits the shock quickly, while the investment channel reacts with a lag — with a more visible marginal shift potentially arriving around 2027.
Two Channels, Two Speeds
The analysis distinguishes between two transmission mechanisms:
- Export channel (fast): Slower overseas AI capex immediately reduces demand for servers, optical modules, PCBs, power equipment and other hardware where Chinese manufacturers are deeply embedded in global supply chains. Order books and shipping data tend to react within quarters.
- Investment channel (slow): Domestic AI-related capital spending — data centers, chip capacity, cloud infrastructure — responds with a lag because projects are planned over multi-year horizons and financing decisions are sticky.
That asymmetry means the near-term pain is likely to show up first in trade and industrial production data, while the domestic investment response — and any offsetting policy support — may only become clear later.
Why 2027 Matters
The 2027 reference point reflects where the cumulative effects of slower external demand, shifting global AI supply chains and the maturation of domestic capacity could converge into a discernible marginal change in the economic trajectory. By then, today’s capex decisions by overseas hyperscalers and AI chip designers will have fully worked through equipment orders, inventory cycles and capacity planning.
Implications for Crypto and Digital-Asset Markets
For crypto markets, the AI capex cycle has become an unexpected but real driver. Several listed miners have pivoted toward hosting AI and high-performance computing workloads, turning their power contracts and data-center shells into AI infrastructure assets. A sustained slowdown in AI capex would compress demand for that conversion trade, pressuring the economics of miner-to-AI pivots and the valuation premium many of these names now carry.
At the same time, decentralized compute and GPU networks settled on-chain are directly exposed to the same demand curve. If enterprise AI spending cools, the pricing power of both centralized cloud providers and decentralized alternatives weakens — a dynamic worth watching for tokens tied to compute marketplaces and inference networks.
Forward-Looking View
The key question is not whether AI capex slows, but how fast and how unevenly. Export-linked hardware names face the earliest test; domestic AI investment and policy offsets will determine whether the drag is a soft patch or a structural shift. For digital-asset investors, the AI-crypto convergence trade — miners turned data-center operators, decentralized compute tokens, on-chain inference marketplaces — now carries a macro beta it did not have two years ago. Positioning should account for the lag: trade data will warn first, investment data will confirm later, and 2027 is the horizon where the full picture emerges.
