Huawei and China's Chipmakers Seize 41% of the AI GPU Market as Nvidia's Grip Loosens

Huawei and China's Chipmakers Seize 41% of the AI GPU Market as Nvidia's Grip Loosens

Two years ago, Nvidia owned virtually the entire Chinese AI accelerator market. That era is over. New data from IDC shows that domestic Chinese chip vendors shipped 1.65 million AI accelerator cards in 2025, capturing 41% of the market and closing fast on Nvidia's shrinking 55% share. The shift has implications that reach far beyond chip sales, reshaping how Asia's largest economy builds and deploys artificial intelligence.

The Numbers Behind the Surge

China's total AI accelerator card shipments reached approximately four million units in 2025. Nvidia still leads with 2.2 million cards shipped, but its dominance has eroded dramatically from a near-total market share before US export restrictions took hold. AMD trails distantly with 160,000 cards and a 4% share.

Among Chinese vendors, Huawei is the clear frontrunner. The company shipped roughly 812,000 chips in 2025, accounting for nearly half of all domestic shipments. T-Head, the chip design arm of Alibaba, placed second with about 265,000 cards, followed by Baidu's Kunlunxin and Cambricon, each shipping approximately 116,000 units.

Chinese GPU^✦ and AI chip makers captured 41% of China's AI accelerator server market in 2025, significantly cutting into Nvidia's historical dominance.

Who Is Building What

The domestic chip ecosystem^✦ is broader than most outside observers realise. Beyond Huawei's Ascend line and Alibaba's T-Head, a second tier of vendors is gaining ground. Hygon captured 5% of Chinese shipments, MetaX took 4%, and Iluvatar CoreX claimed 3%.

This diversification matters. China's AI infrastructure is no longer dependent on a single domestic champion. Multiple vendors competing across different price points and performance tiers creates resilience, something Beijing has been actively engineering through policy.

The Policy Engine

US export controls have been the single largest catalyst for China's domestic chip surge. Washington progressively closed off access to Nvidia's most advanced GPUs, forcing Chinese data centres, cloud providers, and AI labs to find alternatives. Beijing responded with its own accelerants: government procurement mandates favouring domestic chips, subsidies for intelligent computing centres, and local government directives to prioritise Chinese hardware.

The Chinese factory floors where robots are learning new skills increasingly run on domestic silicon. The same is true of the AI-powered kitchens reshaping how China eats. Every new AI application that runs on Huawei or Cambricon hardware deepens the domestic ecosystem's roots.

The momentum behind domestic AI chip adoption in China reflects both policy direction and genuine technological progress across multiple vendors.

What Nvidia Loses

Nvidia retains the technology lead. Its latest chips remain more powerful per unit than anything Huawei or its competitors currently offer. But performance advantages matter less when the market is shifting beneath you. At 55% and falling, Nvidia's China business faces a structural decline that no product cycle can easily reverse.

Analysts at Bernstein have outlined a three-year scenario in which Huawei overtakes Nvidia entirely in China. Under sustained export restrictions, Nvidia's share could fall to as low as 8%, while Huawei's could climb to approximately 70%. That timeline may prove optimistic for Huawei, but the direction is clear.

The Asia-Wide Ripple Effect

China's chip self-sufficiency drive does not exist in isolation. As domestic AI hardware matures, it becomes exportable. Southeast Asian markets, where major tech investments are flowing and AI conferences are drawing global attention, may soon face a choice between American and Chinese AI silicon. For countries navigating the SoftBank-OpenAI axis on one side and Huawei's expanding reach on the other, the chip war is becoming a defining feature of Asia's AI landscape.

• Huawei's Ascend line powers government and state-owned enterprise AI workloads across China
• Alibaba's T-Head serves the country's largest cloud computing platform
• Baidu's Kunlunxin is optimised for search and autonomous driving applications
• Cambricon focuses on edge AI^✦ and inference^✦ for industrial applications
• Hygon targets high-performance computing clusters

Frequently Asked Questions

How did Chinese chipmakers capture 41% of the market so quickly?

US export controls progressively restricted China's access to Nvidia's most advanced GPUs. Beijing responded with procurement mandates, subsidies, and directives pushing domestic adoption. Huawei's Ascend line matured rapidly, and other vendors filled specialised niches across the market.

Is Huawei's chip performance comparable to Nvidia's?

Not yet on a per-chip basis. Nvidia's latest accelerators remain more powerful. However, Huawei compensates with scale, competitive pricing, and the advantage of being the preferred vendor for government and state-backed projects in China.

Could Nvidia recover its China market share?

It is unlikely under current export controls. Bernstein analysts project Nvidia's share could fall to 8% within three years if restrictions continue, while Huawei could reach 70%. Any relaxation of US export policy could change this trajectory.

What does this mean for AI development in China?

China can now sustain large-scale AI training and deployment on domestic hardware. The ecosystem is diversified across multiple vendors, reducing dependence on any single company and creating competition that drives rapid improvement across the sector.

The AI chip map is being redrawn in real time. Will China's domestic ecosystem mature fast enough to compete globally, or will the technology gap keep it confined to the home market? Drop your take in the comments below.