Nvidia Loses China: How Chinese AI Chips Just Took Over Overnight

Nvidia's dominance in China's AI chip market has crumbled, with local competitors capturing over 40% of the market in 2025. Backed by government pressure and rapid innovation, Chinese firms like Huawei and Alibaba's T-Head shipped 1.65 million AI GPUs, signaling a major shift in the global tech landscape.

Key Takeaways

• Chinese AI chip makers captured 41% of the domestic market in 2025, shipping 1.65 million GPUs.
• Nvidia's share fell to 55%, down from 95% pre-sanctions, despite still leading the market.
• Huawei emerged as the top local player, shipping over 800,000 AI accelerators.
• U.S. export bans in 2025 forced China to fast-track domestic chip development.
• Beijing now faces a balancing act: boost local tech while keeping AI firms globally competitive.

In This Article

• Nvidia's China Empire Cracks
• China's Homegrown AI Chip Surge
• U.S. Export Chaos: Ban, Reverse, Repeat
• Beijing's Tough Choice: National Chips vs. Global Competitiveness

Nvidia's China Empire Cracks

For years, Nvidia was the undisputed king of AI chips in China, controlling up to 95% of the market. But 2025 marked a turning point — the U.S. giant's grip has loosened dramatically.

• New data from IDC shows Nvidia’s share in China's AI server market dropped to 55%, a steep fall from its previous dominance.
• Though Nvidia still shipped around 2.2 million chips, local competitors are catching up fast, driven by government mandates and national pride.
• The decline wasn't sudden — it was triggered by a series of U.S. export restrictions that pushed Beijing to accelerate its semiconductor independence.

China's Homegrown AI Chip Surge

With U.S. supplies unreliable, Chinese tech firms didn't just adapt — they innovated at breakneck speed.

• Domestic chipmakers delivered a staggering 1.65 million AI GPUs in 2025, claiming 41% of the 4 million-unit market.
• Huawei led the charge with 812,000 units, thanks to strong government ties and its powerful new Atlas 350 AI accelerator.
• Alibaba’s T-Head followed with 256,000 chips, while AMD lagged behind with just 160,000 units and 4% market share.
• Baidu’s Kunlunxin and Cambricon each shipped 116,000 chips, rounding out the top five in a rapidly diversifying market.

U.S. Export Chaos: Ban, Reverse, Repeat

American restrictions on AI chip exports created a rollercoaster that ultimately backfired — pushing China to go all-in on local tech.

• In April 2025, President Trump banned all AI GPU exports to China, cutting off access to top-tier hardware.
• By July, partial access returned for chips like the Nvidia H20 and AMD MI308, but confusion lingered after Commerce Secretary Howard Lutnick’s vague 'addition' comment halted new orders.
• In December, the U.S. allowed H200 shipments — but only for select institutions and after months of delay.
• This stop-start policy gave Chinese firms the final push they needed to ditch foreign dependency.

Beijing's Tough Choice: National Chips vs. Global Competitiveness

China now faces a tricky dilemma — how to support its homegrown semiconductor industry without holding back its AI ambitions.

• While Chinese AI chips have improved fast, experts say they still trail Nvidia and AMD by 5 to 10 years in performance and efficiency.
• The government wants data centers to use domestic chips, but top AI labs worry about falling behind global rivals.
• Huawei claims its new Atlas 350 offers triple the performance of Nvidia’s H20 — if true, it could reshape the playing field.
• The real test will be whether local chips can power cutting-edge AI research, not just scale up basic infrastructure.

“Chinese semiconductor firms have taken a big chunk of the domestic market, claiming 41% of the local AI server market”

— IDC, reported by Reuters

Final Thoughts

Nvidia may still lead in China, but the momentum has shifted. With government backing, aggressive innovation, and U.S. policy whiplash as a catalyst, Chinese chipmakers are no longer just alternatives — they're serious contenders. The next few years will determine whether China can close the tech gap — or whether reliance on domestic hardware could slow its AI revolution.

Sources & Credits

Originally reported by Latest from Tom's Hardware