China’s AI chip push accelerates under US export curbs—will Nvidia’s dominance crack?

Kuaishou is reportedly funding a chip “spin-off” strategy to expand in-house silicon design, a move framed as a direct response to US export controls that restrict access to advanced foreign technology. The SCMP report links the shift to a broader pattern among Chinese internet firms, where proprietary chips are increasingly used to secure computing infrastructure as AI workloads surge. The funding effort is described alongside players such as TranStreams and QF Capital, suggesting a coordinated push from both corporate and investment channels. In parallel, Huawei and Cambricon are highlighted as leading China’s AI server chip surge, with TrendForce projecting they could capture nearly 80% of the domestic AI server market this year. Strategically, these developments deepen the technology decoupling dynamic between China and the United States, even after a recent leaders’ summit that temporarily “floored” the relationship. A China-based expert from a top state-affiliated think tank warns that structural rivalry and regulatory barriers will likely prevent a durable thaw, using AI and rare earths as examples of persistent constraints. The implication is that export controls are not merely limiting near-term supply, but accelerating indigenous substitution and vertical integration across China’s AI stack. This benefits Chinese chip ecosystems and server makers by reducing dependence on US-linked supply chains, while it pressures global incumbents—especially Nvidia—on both market share and pricing power. Market and economic implications are immediate for semiconductors, AI infrastructure, and the rare-earth-linked supply chain that underpins advanced electronics. If Huawei and Cambricon approach the projected 80% share of China’s domestic AI server chip market, it implies a significant reallocation of demand away from global GPU and accelerator suppliers, potentially intensifying competitive discounting and margin compression for non-local alternatives. The export-control-driven push for proprietary chips also increases capex intensity across Chinese data-center supply chains, from server components to networking and power management. For investors, the signal is a faster-than-expected shift in AI compute procurement toward locally designed silicon, which can influence semiconductor indices, AI hardware sentiment, and risk premia tied to US-China trade and technology restrictions. What to watch next is whether Kuaishou’s funding and design ramp translates into measurable performance and volume shipments, and whether US export controls tighten further in response to rapid substitution. On the diplomacy side, the key trigger is whether regulatory barriers ease in a way that meaningfully changes licensing or compliance pathways for advanced compute components. For the supply chain, monitor rare earth pricing and procurement behavior as firms hedge against bottlenecks that can amplify geopolitical risk. A practical escalation/de-escalation timeline is: near-term (weeks) for announcements of chip tape-outs and server deployments, medium-term (quarters) for market-share confirmation versus TrendForce’s projections, and longer-term (next policy cycle) for any US-China export-control adjustments tied to AI and strategic materials.

Geopolitical Implications

• 01

  Export controls are accelerating vertical integration in China’s AI stack, reducing leverage of foreign suppliers over China’s compute buildout.

• 02

  The AI server chip race is becoming a strategic contest over industrial policy, supply-chain sovereignty, and bargaining power in future negotiations.

• 03

  Rare-earth constraints reinforce that technology decoupling is multi-layered—chips, materials, and manufacturing ecosystems are all implicated.

• 04

  Competitive displacement of global incumbents can reshape corporate alliances and influence future US-China regulatory and compliance frameworks.

Key Signals

• —Chip milestone announcements from Kuaishou (tape-out, qualification) and early deployment data.
• —Any US updates to export-control scope for AI accelerators, networking, or manufacturing equipment.
• —Domestic procurement data confirming whether Huawei/Cambricon approach the projected ~80% share.
• —Rare-earth price moves and procurement hedging behavior by AI hardware and data-center operators.