While US-China semiconductor rivalries have long been a battlefield for technical advantage, recent changes in export control policies are reshaping the landscape. As the Trump administration readjusses its approach to chip exports, investors need to tackle crucial questions. Will relaxing restrictions unlock new opportunities for US technology giants like AMD and Nvidia, or will it accelerate China's independence in AI semiconductors? The answer lies in understanding the interactions of policy, innovation and market dynamics.
Policy Pendulum: From restrictions to strategic flexibility
For years, the US has been putting strict export controls on advanced semiconductors to China, aiming to reduce Beijing's access to AI and supercomputing capabilities. These measures include blacklisting Chinese entities, limiting high-bandwidth chips, and enforcing strict licensing of AI hardware. But the withdrawal of the Biden-era AI proliferation rules in May 2025 and the Trump administration's emphasis on a “bold and comprehensive” AI strategy signal the pivot. The new approach will prioritize collaboration with “trusted allies” while increasing scrutiny of hostile access, particularly China's AI model.
This shift is a readjustment rather than a retreat from containment. The US recognizes that blanket restrictions have unintended consequences. They innovated Chinese companies and created alternatives such as Huawei's Ascend 910D and RISC-V-based architecture. Meanwhile, US companies such as AMD and Nvidia are facing revenue losses due to restrictions on access to China's $16 billion AI chip market. The recent relaxation of export controls to Saudi Arabia explains that strengthening transactions, such as AMD's 500MW AI contract, are strengthening transactions with Humain, focusing on diversifying the market while maintaining pressure on China.
High-tech companies in the US: Navigating regulatory headwinds and innovation
Advanced Microdevices (AMD) and Nvidia are at the forefront of this transition. AMD's second quarter results highlight tensions between regulatory constraints and market resilience. Despite a $800 million hit from US export restrictions on the MI308 chip to China, AMD reported record revenue of $7.7 billion due to strong demand for EPYC processors and Ryzen CPUs. The company is currently pivoting to the MI350 and MI400 series, with the MI355X already surpassing NVIDIA's GB200 in token efficiency per doll.
Nvidia, meanwhile, is navigating complex paths. The H20 chip has been downgraded to comply with our restrictions, but still accounts for 12.5% of the company's revenue. However, Chinese companies such as Bytedance and Alibaba have spent $16 billion stockpiling these chips, creating a shadow market. The potential approval of the Trump administration's new license could rekindle demand, but Nvidia will also compete with Huawei's CloudMatrix 384 system, integrating the 384 Ascend 910C chip to provide raw computational power of 300 PFLOPS.
China's innovators: fill gaps and expand ecosystems
Huawei's Ascend 910d and CloudMatrix 384 illustrate the rapid advances of China. Built on Smic's 7NM N+2 process, the Ascend 910D offers FP16 performance of 1.2 PFLOPS and 12% better energy efficiency than the NVIDIA H100. Although individual chip performance is still behind, Huawei's system-level integration and open-source Pangu AI model create a robust ecosystem. The company's 2025 revenue was $369.9 billion, highlighting its growing influence in AI-As-As-As-As-a-Service (AIAAS) and industry-specific AI deployments.
However, Huawei's success is not without its challenges. US export control of EUV lithography machines and etching tools relies on smuggling and illegal procurement. Huawei acknowledges that it raised the logic of two million people in 2024 highlights supply chain vulnerabilities. However, investments in full-stack AI solutions that combine hardware, software and cloud services are positioned to compete globally despite tackling the technical hurdles of software ecosystems such as Cann and Mindspore.
Investment Opportunities: Balance between Risk and Reward
For investors, the key is to identify companies that can thrive in both a restricted, more open trade environment. Here are three conviction opportunities:
-
AMD (AMD): With the next-generation MI350 and MI400 series, AMD is poised to regain market share in its AI infrastructure. The company's focus on strategic pivots to the Middle East and its full-stack solutions (such as Helios Rack-Scale AI) mitigates China-related risks. Investors should monitor the US Department of Commerce licensing decisions and AMD's third quarter guidance.
-
Huawei Technologies: Despite US restrictions, Huawei's ecosystem-driven approach has gained traction. The company's open source Pangu model and partnership with China's Hyperscalar could drive adoption. However, investors should weigh supply chain vulnerabilities against long-term R&D investments.
-
nvidia (nvda): Nvidia's dominance in AI infrastructure remains unchanged, but dependence on the Chinese market creates volatility. The recent focus on enterprise and automotive AI (such as drivetroll) has diversified revenue streams. Potential mitigation of export controls could unlock a big upside down, but the competition with Huawei's CloudMatrix system is a wild card.
The Way to Begin: The New Equilibrium
The US-China semiconductor dynamics are entering a new phase. Export control remains a tool for strategic impact, but the Trump administration's emphasis on “comprehensive” AI collaboration suggests a more nuanced approach. For US businesses, this means balancing innovation compliance. For Chinese players, it means promoting independence while navigating supply chain constraints.
Investors who position themselves at the intersection of these trends (support companies that can adapt to regulatory changes and technical breakthroughs) can benefit from the next chapter in the Global AI Semiconductor Race. The key is to stay agile as the rules of the game continue to evolve.
