How Nvidia lost China's AI chip market: From 95% dominance to zero in two years

Nvidia's dominance in China's AI chip market has vanished. In roughly two years, the company's market share collapsed from an estimated 95% to effectively zero — a shift driven not by competitive failure but by a three-way collision of U.S. export controls, Chinese industrial policy, and a domestic competitor's rapid ramp-up .

The market share collapse in numbers

Before U.S. export controls began tightening in 2022, Nvidia commanded about 95% of China's market for advanced AI accelerators . By 2025, that share had fallen to roughly 40%, according to Bernstein Research, with Huawei reaching a comparable market share in the same period . In April 2026, IDC reported that Chinese GPU and AI-chip manufacturers had already secured nearly 41% of China's AI accelerator server market .

Bernstein projects Nvidia's share will drop to around 8% in 2026, while Huawei's will likely grow to about 50% . Jensen Huang, Nvidia's CEO, has publicly acknowledged the company has "largely conceded" the Chinese AI chip market to Huawei . In a CNBC interview, Huang said, "We have effectively conceded that market to them" . By mid-2026, Nvidia stopped including China in its revenue and profit forecasts .

U.S. export controls: the primary trigger

The cascade began in October 2022, when the Biden administration banned the export to China of any AI chips equal to or more capable than Nvidia's A100 chip . Nvidia responded by designing slower variants — the A800, H800, and later the H20 — specifically for the Chinese market. But each was eventually restricted.

In April 2025, the Trump administration required licenses for sales of Nvidia's H20 chips to China, marking a major new restriction . Nvidia disclosed charges tied to the new restrictions, underscoring the financial impact . Just months later, in May 2026, the U.S. Commerce Department moved to close a loophole that had allowed AI chip exports to Chinese companies operating outside China .

Huang publicly called the export controls "a failure" in May 2025, saying they cost U.S. firms billions in lost revenue and accelerated Chinese self-sufficiency . The Brookings Institution summarized the outcome starkly: "Trump Approved a Nvidia Chip for Sale in China. Beijing Doesn't Want It," noting that as of mid-2026 not a single H200 chip had been sold to Chinese companies despite administration approval .

China's $295 billion AI data-center plan with an 80% domestic-sourcing mandate

In June 2026, Bloomberg reported that China's National Development and Reform Commission was drafting a five-year plan to spend roughly 2 trillion yuan (about $295 billion) on a nationwide network of AI data centers . The plan explicitly requires that at least 80% of the underlying technology, including AI chips, come from domestic suppliers such as Huawei . State-owned carriers China Mobile and China Telecom will operate most of the hubs, with a target of linking them into a unified computing grid by 2028 .

This mandate alone would sharply limit Nvidia's role in China's largest planned AI infrastructure push, even if U.S. export restrictions were later relaxed. The total investment, when integrated with power grid upgrades, could reach 5 trillion yuan (~$735 billion) .

Huawei's Ascend chip ramp-up: the supply-side enabler

Huawei has become the primary beneficiary of Nvidia's exit. The company's Ascend AI chip division, HiSilicon, has ramped production aggressively . By 2025, Huawei's AI chip revenue reached $7.5 billion; the company projects $12 billion in 2026 — a 60% jump .

Chinese hyperscalers alone represent $12–15 billion in projected AI chip demand for 2026, and ByteDance has committed $5.6 billion specifically in orders for Huawei's Ascend 950PR chips . Analysts at Counterpoint note that Huawei's advantage lies not just in a single chip but in its full technology stack, including software and integration .

The shift has been rapid: Chinese developers and data-center operators are actively transitioning to domestic hardware, with Huawei, Cambricon, Moore Threads, and MetaX advancing both chips and software ecosystems .

How the three forces worked together

Factor	Mechanism	Impact
U.S. export controls	Multiple rounds of restrictions starting in 2022 progressively limited Nvidia's ability to sell advanced and then even China-specific chips .	Reduced Nvidia's legal market access, triggered $4.5B+ in charges, and pushed Chinese firms to seek domestic alternatives .
China's 80% domestic-sourcing mandate	The $295B AI data-center plan requires at least 80% domestic technology, explicitly including AI chips .	Shifts all state-backed AI infrastructure demand toward domestic suppliers, functionally excluding Nvidia even if export controls are relaxed .
Huawei's Ascend chip ramp-up	Huawei scaled AI chip production from near-zero in 2023 to projected $12B revenue in 2026 .	Domestic supply became available to absorb demand, with Huawei capturing roughly 50% of the market by 2026 .

What it means for the global AI chip landscape

The collapse of Nvidia's China market share signals the formal bifurcation of the global AI computing stack into two separate ecosystems . China is no longer just a market Nvidia lost — it is a market where domestic competitors are rapidly building a parallel supply chain.

The bottom line, across multiple high-authority sources, is unambiguous: this was not a market-driven loss of competitiveness. It was a policy-driven decoupling in which U.S. export controls constrained Nvidia's supply, China's industrial policy redirected demand toward domestic chips, and Huawei and other Chinese suppliers moved decisively into the resulting gap .