By Grace Kim

Huawei is closing the performance and market gap with Nvidia inside China, according to a report by Chosunbiz, as Chinese AI chip development shifts beyond traditional GPU architectures.

The report frames the shift as structural rather than incremental. Chinese designers are no longer chasing Nvidia only in general-purpose GPUs for AI training. They are broadening the portfolio into alternative accelerator categories, a move that changes the competitive equation for the U.S. chipmaker in its largest historical market.

Why does the GPU focus no longer tell the whole story?

For years, the benchmark for Chinese semiconductor progress was simple. Could a domestic GPU match an Nvidia data center product? The answer, mostly, was no. Export controls tightened that gap further by restricting Nvidia's most advanced accelerators from Chinese customers.

Chosunbiz reports that the picture is now broader. Huawei and other Chinese suppliers are winning designs across a wider set of AI compute products, not just GPUs built to Nvidia's template. This diversification reduces the single point of comparison that long favored the U.S. vendor and gives Chinese buyers more routes to domestic supply.

Huawei sits at the center of this transition. The company has built a domestic stack that pairs its own accelerators with software and systems integration, an approach aimed at enterprise and government customers that must source locally under U.S. export rules.

What does this mean for Nvidia's China business?

Nvidia remains the global leader in AI accelerators. No Chinese vendor yet matches its top-end training products or the software ecosystem built around its CUDA platform.

The competitive question is different inside China. Washington's export controls bar Nvidia from selling its most advanced chips to Chinese customers. That regulation hands protected market space to Huawei and other domestic suppliers, who face no such restriction on their home turf.

Chosunbiz characterizes the result as Huawei closing on Nvidia. The gap narrows not only through engineering progress but through policy. Each tightening of export rules widens the addressable market for domestic alternatives.

Is the shift beyond GPUs a lasting change?

The move toward non-GPU AI silicon reflects how Chinese AI workloads are evolving. Inference, edge deployment and domain-specific compute reward different design tradeoffs than large-scale training clusters. Chinese suppliers are positioning products for those segments rather than fighting Nvidia head-on in its strongest category.

This strategy carries risk. Nvidia's ecosystem lock-in remains deep, and Chinese software alternatives must prove they can support large model development at scale. Huawei's stack has advanced, but matching developer mindshare takes years.

It also carries leverage. If Chinese accelerators prove adequate for a growing share of domestic workloads, Nvidia's addressable China revenue shrinks further, and Beijing gains a credible supplier base for strategic programs.

What comes next?

The trajectory Chosunbiz describes suggests the China AI chip market splits into two tracks. Nvidia keeps a durable lead in products it is permitted to sell. Huawei and domestic rivals expand across the rest, on favorable regulatory ground and with state-backed demand behind them.