China may ease barriers for US AI chips

IDC

China remains one of the key but restricted markets for American chipmakers. Access to artificial intelligence technologies is increasingly becoming part of trade and political negotiations between Beijing and Washington. Chinese authorities are expected to ease restrictions on imports of US AI chips, Bloomberg reports.

Nvidia’s position and expectations for market opening

Nvidia CEO Jensen Huang said the Chinese government may, over time, reconsider its approach to importing US AI chips and ease existing restrictions. The decision will depend on Beijing’s policy toward developing its domestic semiconductor industry and achieving technological self-sufficiency. Huang previously described China as a potential $50 billion market for Nvidia, although earlier this year the company maintained its forecast of zero AI chip sales in China.

AI chip exports remain a sensitive issue, as procurement decisions in China are effectively constrained by the country’s push to develop domestic alternatives and support national manufacturers, including Huawei Technologies Co. Authorities have declined additional purchases while awaiting domestic solutions.

US policy and export restrictions

Export restrictions on advanced chip shipments to China have been in place since 2022 as part of a broader US strategy to limit the development of China’s artificial intelligence sector. These measures primarily target high-performance accelerators used for training and running large AI models.

The issue of H200 chip shipments was discussed in contacts between US and Chinese representatives. In December, a decision was made to allow shipments to customers in China, marking a notable easing of US export policy. The US Department of Commerce issued licenses that formally opened the path for such sales. Later, several Chinese companies reported an inability to complete purchases. It was also noted that orders had been placed, but transactions were not finalized.

Regulation remains inconsistent and highly sensitive to political decisions. In Washington, discussions continue on measures intended to preserve the US technological edge while not fully closing commercial channels for American companies. In May, US President Donald Trump said that Nvidia’s H200 chips had indeed come up and suggested that something might be possible with them.

AI demand and memory shortages

Taiwan remains the center of semiconductor manufacturing, although the US is seeking to expand its domestic chip production capacity. Dell Technologies CEO Michael Dell noted that the rapid rise in artificial intelligence demand, driven by the adoption of agent-based software, is increasing pressure on suppliers, especially memory manufacturers.

Experts believe that access to memory chips remains the main bottleneck for the industry, although major suppliers are increasing investments in expanding production capacity. “It takes a long time to build these factories,” Michael Dell said. “It’s a true long-term partnership, even though we would like more right now.” Jensen Huang added that it would take at least another decade to fully catch up.

AI infrastructure and semiconductor market growth

The semiconductor market is experiencing a sharp acceleration in growth driven by large-scale investments in artificial intelligence infrastructure. According to IDC forecasts, global industry revenues could increase by 50% by the end of 2026, reaching $1.29 trillion.

The main growth driver is the AI infrastructure segment, which is effectively becoming the new center of demand in the market. Data centers and computing clusters account for the bulk of activity, with rising demand for accelerators, processors, and networking solutions used for training and deploying AI models. Infrastructure investment continues to grow, and demand for computing power is becoming self-reinforcing.

The most pronounced growth is seen in the memory segment. IDC estimates that DRAM revenue could nearly triple in 2026, reaching more than $400 billion, while the overall memory market could exceed half a trillion dollars. Analysts note that memory is transitioning from a cyclical commodity into a strategically scarce resource. Limited production capacity and the technological complexity of manufacturing continue to constrain supply despite major investments by leading producers.

Conclusion

The AI chip and semiconductor market is increasingly shaped not by commercial demand alone, but by the intersection of technological constraints and geopolitics. Formal access to the Chinese market remains under discussion, but in practice it depends on political decisions and strategies of technological self-sufficiency.

At a global level, the industry is entering a phase of structural shortage, where the growth of AI infrastructure is outpacing manufacturing capacity. Bottlenecks in supply chains, particularly in memory and high-performance components, are becoming a long-term constraint on development.

For investors, this implies higher volatility in expectations and a shift in focus from individual end markets to supply chain resilience and the ability of companies to secure long-term contracts for critical components. In this environment, the key factor is not only demand, but also access to constrained production resources that define the industry’s real growth trajectory.