Forecasts cited by three media outlets suggest China’s domestic AI chip suppliers could take most of the country’s market by 2026, but the reports do not agree on how sharply NVIDIA and AMD will be displaced.

A Wccftech headline, citing an unnamed report, says NVIDIA and AMD could together hold only 10% of China’s AI chip market in 2026, leaving domestic chips with 90%. KuCoin separately reported that China’s AI chip market was approaching 90% domestic share as NVIDIA’s influence declined. Shane the Gamer cited TrendForce for a less extreme estimate: NVIDIA and AMD’s combined share could fall to 21% in 2026.

The common signal is a weakening position for U.S. chipmakers in China. The size and timing of that shift remain uncertain because the supplied source material contains headlines and summaries, not the underlying research, methodology, market definition or detailed data.

Conflicting forecasts point to the same pressure

The most aggressive forecast comes from the Wccftech report. Its headline projects that NVIDIA and AMD will collectively account for 10% of China’s AI chip market in 2026, while domestic chips reach 90%. The article summary identifies this as a report-based claim rather than an announced target from either chipmaker.

KuCoin’s coverage uses similar language, saying China’s AI chip market is nearing 90% domestic share and that NVIDIA’s influence is declining. However, the available evidence does not show whether that 90% figure refers to current shipments, installed systems, procurement value, accelerator demand, or a forecast for 2026. It also does not clarify whether AMD is included in the calculation.

The Shane the Gamer item attributes its forecast to TrendForce and gives a materially different result: NVIDIA and AMD together would retain 21% of the market in 2026. That would still represent a major reduction in their position, but it would leave them with more than twice the combined share suggested by the Wccftech headline.

These figures should not be treated as interchangeable. A market-share estimate can change substantially depending on whether researchers count chip revenue, units, computing capacity, cloud deployments or state-backed procurement. The available source extracts do not identify the metric used by either forecast.

What the evidence establishes—and what it does not

The strongest substantiated conclusion is that analysts and market observers expect China’s domestic chip ecosystem to gain ground against NVIDIA and AMD. The three items also connect that expectation to 2026, making the year a focal point for forecasts about China’s AI infrastructure.

The evidence does not establish that domestic suppliers have already captured 90% of the entire Chinese AI chip market. KuCoin’s summary says the market is “nears” that level, while the Wccftech headline presents 90% as a 2026 outcome. Those are different claims. Without the full articles or original TrendForce research, the distinction cannot be resolved.

Nor does the evidence identify the domestic companies, chip families or purchasing programs expected to drive the shift. It provides no verified shipment totals, customer lists, product specifications or independent performance comparisons. NVIDIA and AMD are not quoted in the supplied material, and no official statement from a Chinese chipmaker is included.

The figures are therefore best read as reported market forecasts, not confirmed market results. TrendForce is named as the source of the 21% estimate, but the underlying report is not available in the evidence. The 10% estimate is described only as coming from a report in the Wccftech headline. That attribution gap matters for investors, product teams and researchers trying to use the numbers for planning.

Why the forecast matters to AI builders and buyers

If either forecast is directionally correct, China’s AI infrastructure market could become less accessible to U.S. suppliers even where their hardware remains competitive. For AI builders, that would affect the choice of AI accelerators, software stacks, inference capacity and cloud partners available for deployments inside China.

A larger domestic-chip presence could also increase the importance of local compiler support, model optimization and hardware-specific tooling. Developers may need to maintain more than one deployment path rather than assume that code optimized for NVIDIA’s ecosystem will transfer cleanly to alternative processors. The business cost would include engineering time, testing and operational support, not just the purchase price of chips.

Enterprise buyers would face a different set of trade-offs. Domestic chips may offer advantages in availability, regulatory alignment or local procurement, while established U.S. platforms may retain strengths in software maturity, developer familiarity and compatibility with existing workloads. The supplied reports do not compare those factors, so the market-share forecasts alone cannot determine which option is technically or economically superior.

For NVIDIA and AMD, a decline in China share would affect more than regional revenue. China is also a significant environment for testing demand, building developer relationships and supporting large-scale AI deployments. Export controls and product restrictions may accelerate local substitution, but the available evidence does not quantify how much of the forecast is attributable to policy, supply constraints, pricing, performance or customer preference.

What to watch next

The first signal will be the release of the original TrendForce analysis or another source that explains the 21% estimate. Readers should look for its market definition, baseline year, treatment of custom silicon and distinction between revenue share and computing capacity.

The second is updated shipment or procurement data from Chinese cloud providers, public-sector buyers and hardware distributors. Those figures could help test whether the reported 90% domestic-share claim describes actual deployment or only a narrower market segment.

Product-level evidence will also matter. New domestic accelerators, software compatibility announcements, production capacity and independent benchmark results would show whether local suppliers are winning through performance, availability, policy support or a combination of factors.

Finally, NVIDIA and AMD’s China-specific product announcements, regulatory filings and earnings commentary could clarify whether the companies expect a temporary supply disruption or a sustained loss of market access. Any change in export rules would also alter the assumptions behind forecasts extending to 2026.

Creati.ai perspective

The news is less a confirmed 90% market-share event than a warning about how quickly China’s AI hardware outlook is being reassessed. The gap between the 10% and 21% forecasts shows why headline percentages should not guide infrastructure decisions without knowing what is being measured.

For builders and enterprise buyers, the practical response is scenario planning: track local hardware and software maturity, validate workloads across available platforms and avoid treating one accelerator ecosystem as the only deployment path. Until the underlying research is published, the direction of the forecast is more credible than its precise number.