Nvidia Loses China AI Market as Beijing Blocks H200 Sales
Key insights
- Beijing blocked H200 purchases despite U.S. approval, giving Nvidia zero percent of China's AI accelerator market.
- Huawei's AI chip revenue is projected to reach $12 billion in 2026, up 60 percent from $7.5 billion in 2025.
- Chinese hyperscalers Alibaba, ByteDance, and Tencent are deploying Huawei Ascend 950 while hedging on future Nvidia access.
Why this matters
For AI infrastructure teams and hardware procurement leads, China's enforced domestic-chip strategy means Nvidia's addressable market has structurally contracted by one of the world's largest AI compute markets, reshaping Nvidia's long-term revenue ceiling and competitive positioning. Huawei's rapid revenue trajectory signals that Chinese domestic chips are scaling fast enough to credibly serve hyperscaler workloads, which changes the calculus for any global AI vendor banking on China re-opening. Founders and investors building on assumptions of a unified global AI compute supply chain now face a world where two parallel hardware ecosystems are hardening, with policy, not performance, as the deciding variable.
Summary
Beijing has blocked Chinese tech firms from purchasing Nvidia H200 chips despite explicit U.S. approval, leaving Nvidia CEO Jensen Huang to acknowledge the company holds effectively zero percent of China's AI accelerator market.
The mechanics are deliberate: major Chinese hyperscalers including Alibaba, ByteDance, and Tencent have been directed toward domestic alternatives, primarily Huawei's Ascend 950, as part of a state-backed push to eliminate dependency on foreign chip infrastructure. These companies are running a dual-stack approach, deploying Huawei domestically while keeping optionality on Nvidia if access conditions ever change.
Essentially: (Nvidia, Huawei) are now on opposite sides of a structurally enforced market split, and Beijing is the one drawing the line.
- Huawei projects $12 billion in AI chip revenue for 2026, up from $7.5 billion, signaling it can absorb demand at scale.
- Moore Threads and MetaX are also scaling to fill remaining gaps, broadening China's domestic supply base beyond a single vendor.
- The U.S. approved the H200 sales to 10 named Chinese firms, meaning the block originates entirely from the Chinese side.
This isn't a trade-war side effect anymore; it's a deliberate industrial policy outcome that has now hardened into infrastructure.
Potential risks and opportunities
Risks
- Nvidia faces accelerating revenue compression in the China segment through 2026-2027 if Huawei closes the performance gap on next-generation Ascend chips before any policy reversal occurs
- U.S. semiconductor firms lobbying for export-license expansions lose political leverage as Chinese hyperscalers publicly commit capex to domestic alternatives, making re-entry arguments harder to sustain in Washington
- Moore Threads and MetaX, scaling rapidly on state-directed demand, could produce cost-competitive chips within 18 months that Chinese firms then export to third-party markets, undercutting Nvidia in Southeast Asia and the Middle East
Opportunities
- Huawei's ecosystem partners and Ascend software stack developers gain immediate enterprise demand from Alibaba, ByteDance, and Tencent as these firms standardize on domestic hardware
- Nvidia can redirect China-bound H200 inventory and manufacturing capacity toward accelerating delivery timelines in the U.S., Europe, and Gulf markets where demand backlogs remain significant
- Western AI cloud providers (AWS, Azure, Google Cloud) offering Nvidia-accelerated compute gain a structural advantage over Chinese cloud rivals in attracting non-Chinese AI companies that require H100/H200-class performance without supply uncertainty
What we don't know yet
- Whether the 10 named Chinese firms received any formal government directive blocking purchases, or if the block operates through informal guidance with no public paper trail
- How Ascend 950 benchmark performance compares to H200 on large-scale LLM training workloads, since public head-to-head data remains sparse as of May 2026
- Whether Nvidia's reported 'zero percent' share reflects lost bookings, cancelled contracts, or simply no new pipeline, which would change the revenue impact timeline
Originally reported by cnbc.com
Read the original article →Original headline: Jensen Huang Says Nvidia Has 'Zero Percent' AI Market Share in China After Beijing Blocks Approved H200 Purchases to Protect Domestic Chip Industry