Nvidia now holds 34 percent of the Chinese artificial intelligence chip market, ceding share to Huawei and a clutch of domestic rivals, the latest market tracking data shows. The figure marks a reversal for a company that once treated mainland China as one of its most lucrative data-center customer pools.

The headline number comes from a fresh market tracking exercise reported by The Economic Times. Nvidia does not break out China-specific revenue in its segment reporting, so the 34 percent figure reflects third-party estimates rather than audited disclosures. The direction of travel, however, has been visible for some quarters.

What does the 34 percent figure capture?

The number measures Nvidia's share of accelerator silicon sold into mainland China for AI workloads, set against the total addressable market including domestic alternatives. It does not measure revenue in dollars, and it does not capture systems routed through Hong Kong or Taiwan channels. Treat it as a relative gauge rather than an absolute scorecard.

Huawei's Ascend series of accelerators has been the principal beneficiary of Nvidia's retreat. The Shenzhen-based group has leaned on domestic foundries to scale production of training-oriented parts aimed at hyperscale cloud operators and state-owned telecom carriers. Volume remains modest by global standards; the trajectory is the story.

Why is Huawei gaining ground?

Three forces explain the shift.

• US export controls now bar Nvidia's top-performing data-center parts from sale into China.
• The H20 chip, a cut-down Ada Lovelace derivative, has lost ground as domestic alternatives improved.
• Procurement preferences inside large Chinese cloud operators increasingly favor domestic suppliers for supply-security reasons.

Nvidia's revenue from China has not collapsed outright. The company continues to book meaningful income from H20 shipments, and its data-center segment still posts record quarterly figures globally. The damage is relative rather than absolute. Nvidia sells more chips in absolute terms than it did two years ago; it is simply losing share inside a market that is growing around it.

The competitive picture carries a geopolitical tail. Washington tightened licensing rules for advanced accelerators in late 2023, then expanded restrictions in successive rounds. Each tightening gave Huawei another quarter to close the performance gap on training parts and to position its products as the default choice for sovereign AI infrastructure inside China. The Department of Commerce has signaled that further restrictions may follow, which would compress Nvidia's addressable China market further still.

How should investors read the print?

The 34 percent figure functions as a leading indicator rather than a terminal number. Huawei is not yet displacing Nvidia on raw silicon merit; it is winning on availability, procurement policy, and the willingness of Chinese hyperscalers to accept a domestic default. Should Huawei's roadmap deliver comparable software maturity over the coming cycle, the figure could fall further.

The numbers also speak to a broader reshaping of the global AI accelerator supply chain. TSMC's CoWoS packaging capacity, the bottleneck for Nvidia's top-end GPUs, has expanded through 2024 and 2025, easing supply but doing nothing to restore access to Chinese hyperscalers. SMIC, by contrast, has prioritized domestic AI parts and now treats Huawei as an anchor customer. The bifurcation is becoming structural rather than cyclical. Put simply, the bottleneck moved from tape-out to trade policy.

Nvidia's response, so far, has been to develop further cut-down variants and to lobby for licensing clarity. Whether Washington grants the necessary export permits remains the single most important variable for Nvidia's China revenue in the near term. The company's longer-term bet is that its software stack, the CUDA programming environment, keeps customers tethered even when the silicon underneath is not Nvidia's own.

For now, the 34 percent figure is a snapshot of a market in transition. The next quarterly reading will move the dial again. The direction of that move depends less on Chinese demand than on decisions made in Washington and Shenzhen. Somewhere between those two capitals sits the future of Nvidia's second-largest regional market.