Nvidia's China market share crashes from 66% to 8% as export curbs fuel domestic AI chip surge

Reviewed byNidhi Govil

19 Sources

Share

Nvidia faces a dramatic decline in China as its market share plummets from 66% in 2024 to just 8% this year. The collapse stems from tightening US export restrictions on high-end AI chips and rapid advances by homegrown Chinese AI hardware makers like Huawei and Moore Threads. While the Trump administration has opened limited sales of Nvidia H200 chips with a 25% surcharge, Beijing is drafting its own purchase rules as domestic suppliers now meet 80% of local demand.

Nvidia Loses Ground as China Market Share Collapses

Nvidia is experiencing a precipitous decline in China as analysts from Bernstein predict the company's China market share will drop to approximately 8% this year, down from 66% in 2024

1

. This dramatic shift represents one of the most significant market transformations in the AI chips sector, driven by a combination of US export restrictions and the rapid maturation of homegrown Chinese AI hardware. Domestic suppliers including Huawei, Cambricon, Moore Threads, and MetaX are now positioned to satisfy approximately 80% of local demand

1

. The collapse marks a turning point for Nvidia, which previously viewed China as a potentially $50 billion annual market

5

.

Source: Analytics Insight

Source: Analytics Insight

Trump Administration Opens Limited H200 Sales with 25% Surcharge

The Trump administration announced it will approve sales of Nvidia H200 chips to China, but with a 25% government surcharge on the dollar value of transactions

5

.

Source: ET

Source: ET

Unlike Nvidia's previous China-targeted H20 chip, the Nvidia H200 represents a full-performance version of the company's Hopper generation sold in other markets. The U.S. Department of Commerce published new AI accelerator export rules that introduce a narrowly defined, compliance-heavy method allowing limited exports under strict case-by-case licensing

4

. Eligible devices must have a total processing performance (TPP) of less than 21,000 and DRAM bandwidth below 6,500 GB/s

4

. The H200 meets these thresholds with a TPP score of 15,832 and memory bandwidth of 4.8 TB/s

4

.

Source: Bloomberg

Source: Bloomberg

Beijing Responds with Purchase Rules and Domestic Push

China is working to establish rules regulating the total volume of high-end AI chips local companies can purchase from foreign makers, effectively controlling rather than banning some sales

3

. Chinese customs authorities told agents that Nvidia's H200 chips are not permitted to enter China, while the government summoned domestic technology companies and explicitly instructed them not to purchase the chips unless necessary

3

. Chinese state media outlet Global Times dismissed the licensing requirements as "discriminatory," noting the stringent export regime targets China specifically while not applying to other US trading partners

2

. The licensing requirements include certifying no shortage of processors exists in the US and limiting chip shipments to China to no more than 50% of total products made for the US market

2

.

Chinese AI Hardware Makers Close Performance Gap

Moore Threads announced its Huashan product, the company's first GPU dedicated solely for AI training and inference, which can compete against Nvidia's Hopper H100 and H200 products

1

. "The new products meet the needs of domestic developers," said Zhang Jianzhong, chief executive of Moore Threads. "There will be no more need to wait for advanced products from overseas"

1

. Huawei's AI CloudMatrix 384 can beat both GB200 NVL72 and GB300 NVL72 systems in BF16 FLOPS, though with four times more power consumption

1

. The company's next-generation Atlas 950 SuperCluster, based on 524,288 Ascend 950DT AI accelerators, is projected to offer up to 524 FP8 ExaFLOPS for AI training sometime in 2026-2027

1

.

Semiconductor Self-Reliance Becomes National Priority

China's transition to domestic AI hardware reflects a long-term national goal for semiconductor self-reliance. A draft five-year plan reportedly circulated by the Communist Party in October calls for a 'new national system' directing state bodies, private companies, and financial institutions toward this objective

1

. At the heart of this effort are the 'four little dragons' of Chinese GPUs: Moore Threads, MetaX, Biren Technology, and Suiyuan Technology

1

. Large hyperscalers are intensifying custom silicon programs, with Baidu's Kunlunxin unit planning to introduce five AI processors by 2030, while Alibaba continues its own silicon efforts. However, China's AI industry remains constrained by SMIC's ability to produce chips on its 7nm-class process technologies in sizable quantities

1

.

Compliance Burden Limits Practical Export Potential

The new AI accelerator export rules impose substantial compliance requirements that may limit practical exports. Exporters must demonstrate US demand is fully met, no domestic orders are delayed, no US advanced-node foundry capacity is diverted, and China-bound shipments don't exceed 50% of the same product shipped to the United States

4

. Every shipment requires review by an independent testing laboratory headquartered in the US with no financial ties to the exporter, and all testing must be conducted domestically

4

. The licensing regime also includes cloud usage controls, requiring exporters to confirm accelerators aren't destined for military or weapons-related end uses and to provide detailed "Know Your Customer" procedures

4

. AMD's Instinct MI325X also qualifies under these rules with a TPP score of 20,800 and memory bandwidth of 6 TB/s

4

. Bloomberg Intelligence analyst Michael Deng noted the combined effect of these requirements may be overly burdensome, posing a significant hurdle for exports

2

. The remaining challenge involves transitioning from an ecosystem centered on Nvidia and CUDA to a fully domestic hardware and software stack, which proves difficult and expensive for existing AI deployments

1

. If SMIC cannot substantially increase output in coming years, China's AI sector faces the prospect of falling dramatically behind America's High-Performance Computing capabilities, unless alternative supply channels emerge

1

.

Today's Top Stories

TheOutpost.ai

Your Daily Dose of Curated AI News

Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Spend less time searching for the latest in AI and get straight to action.

© 2026 Triveous Technologies Private Limited
Instagram logo
LinkedIn logo