China AI chip market

The numbers, via IDC data reported by Reuters on October 13, are stark. In the first half of 2026, Nvidia shipped 678,000 AI accelerators in China — down 44% year over year — for a 34% share of the data-center AI chip market. Huawei shipped 572,000 units, nearly double its volume a year earlier, taking 28%. The gap between them has collapsed from 35 percentage points to six.

This isn’t a blip. It’s the structural consequence of a policy choice — and it deserves a closer look than the headlines give it.

The policy backfired in the most predictable way

Washington’s export controls were designed to deny China access to cutting-edge AI compute. They succeeded — at denying Nvidia access to China’s $50 billion AI chip market. Jensen Huang has said exactly this, repeatedly: the curbs forced an American company to hand the world’s second-largest AI market to Huawei.

The mechanism is textbook. Restrictions created scarcity; scarcity created a price umbrella; Beijing layered on procurement preferences and import-approval requirements for compliant Nvidia chips. The result: nine of the eleven Chinese chipmakers tracked by IDC grew shipments year over year. Alibaba’s T-Head took 7%, Cambricon 5%, Baidu’s Kunlunxin 4%, Hygon 3%. The ecosystem didn’t just survive the controls — it diversified under them.

The quieter shift: GPUs are losing the volume game

Buried in the IDC data is a second story. GPU shipments in China fell 26% to 991,000 units, while non-GPU AI accelerators rose roughly 80% to 1.016 million — overtaking GPUs by volume. These are domain-specific chips optimized for inference and particular workloads.

This matters because it suggests the market is maturing past the “throw H100s at everything” phase. As inference dominates AI spending, specialized silicon — exactly the kind Chinese vendors are now shipping at scale — becomes structurally more important. The moat around general-purpose GPU dominance is narrowing from the demand side, not just the supply side.

What this means for investors

Three implications worth pricing in:

  1. Nvidia’s China revenue is structurally impaired, not cyclically delayed. The 44% shipment decline isn’t waiting for a policy reversal — domestic alternatives are now good enough for a growing share of workloads. Even if controls loosened tomorrow, Huawei’s 28% doesn’t evaporate.

  2. The AI trade is fragmenting. Tencent is raising at least $4.5 billion in fresh debt as its AI capex more than doubles. Asian tech stocks sold off hard on the news — Korea’s KOSPI dropped around 6% — as investors repriced the assumption that AI demand automatically flows to the incumbent semiconductor leaders.

  3. Scarcity → competition is the new regime. The first phase of the AI boom was defined by shortage: too few advanced chips for exploding demand. The next phase may be defined by competition — more suppliers, more architectures, thinner margins for incumbents. Good for customers. Harder for shareholders of the old winners.

The bottom line

Export controls bought time; they also bought China a domestic AI-chip industry. The 34% figure isn’t just a market-share data point — it’s evidence that the attempt to freeze China’s AI progress accelerated the one outcome Washington most wanted to avoid: a self-sufficient competitor ecosystem.

The frontier of the AI race is no longer just about who trains the best model. It’s about who owns the silicon underneath — and that map is being redrawn.

Sources: Reuters via Mix 92.9, Aju Press, PaisaKawach