Editor's note: Lin G. is a CGTN economic commentator. The views expressed in this article are the author's own and do not necessarily reflect those of CGTN.

When NVIDIA CEO Jensen Huang boarded US President Donald Trump's plane during a stop in Alaska en route to China this week, the image became one of the defining symbols of the visit.

Not because another business executive had joined the trip, but because the moment revealed that the center of gravity in China-US relations has now shifted from tariffs toward artificial intelligence, semiconductors, and the broader technological architecture of the next era.

The unintended consequence of US tech containment

Before Trump's second term, the United States had already started repositioning its strategy around technological containment. In October 2023, during the administration of Joe Biden, the US Bureau of Industry and Security expanded export controls targeting advanced AI chips and semiconductor manufacturing equipment destined for China.

Under Trump's return to office, these restrictions have become even more central to Washington's broader China strategy.

One of the most symbolic moments came on April 9, 2025, when the Trump administration prohibited NVIDIA and AMD from exporting the H200 and MI300X to China. Ironically, the H200 itself had already been designed specifically as a downgraded chip tailored for the Chinese market after earlier rounds of restrictions.

However, what followed revealed an increasingly visible contradiction inside the American technology sector itself. Executives such as Jensen Huang repeatedly argued that excessive restrictions on China were not serving long-term American interests.

The decisive factor of artificial intelligence computing is not whether a single chip is the most advanced. Its power is distributed across enormous numbers of chips operating in parallel. In other words, even if access to the highest-end chips is restricted, AI systems can still achieve strong performance by using larger quantities of less advanced processors.

This is precisely why American chip export controls have failed to halt China's AI development. Ironically, Washington's attempt to preserve technological dominance has accelerated the emergence of competing ecosystems. Following successive rounds of US restrictions, demand for domestic alternatives such as Huawei's Ascend AI chips has surged across China.

The paradox of AI decoupling

By mid-2025, NVIDIA successfully lobbied for partial relaxation of export rules on the H200. Later, in early 2026, restrictions on the more advanced H200 chips were loosened further.

But by then, something fundamental had changed inside China's AI industry. More Chinese AI companies started shifting toward domestic architectures. DeepSeek's new version — V4 models placed emphasis on compatibility with Huawei's CANN architecture.

That's exactly why NVIDIA's Jensen Huang publicly expressed concern over the possibility of Chinese frontier models moving away from CUDA entirely.

In fact, after DeepSeek optimized newer models around domestic architectures, NVIDIA moved rapidly to ensure compatibility from its own side. This was a symbolic reversal: Instead of Chinese firms rushing to adapt to NVIDIA, NVIDIA found itself needing to adapt to China's evolving AI ecosystem.

This also indicates that full technological decoupling in artificial intelligence remains structurally unrealistic.

Unlike traditional industrial sectors, artificial intelligence is built on deeply shared theoretical foundations. Research papers circulate globally. Open-source communities transcend borders. Developers collaborate across continents on platforms like GitHub. Open-source models can be downloaded, modified, and redistributed worldwide.

Many Chinese models are open-source. API access is often dramatically cheaper. This creates a paradox for Washington. Restricting the use of Chinese AI systems inside certain American institutions does not erase the growing international appeal of open-source and lower-cost Chinese AI services.

In practice, AI decoupling would likely inflict enormous costs on both sides — especially on the United States.

The United States may still maintain an edge in frontier-model capabilities. But without Chinese competitors, America would face higher AI costs and a more fragmented innovation ecosystem. China, meanwhile, would lose short-term access to some of the world's most advanced frontier technologies, but would likely accelerate self-reliance even further.

Why the AI era requires controlled competition

The deeper issue is that artificial intelligence is not simply another commercial industry. It is a general-purpose technology that can be turned into an instrument of confrontation and potentially destructive escalation.

This risk is serious because AI systems are shaped by the principles and incentives guiding their development. Under conditions of manageable competition, major powers still share broad incentives to ensure AI remains beneficial to society. But under conditions of escalating fragmentation, the incentives change. AI development risks becoming subordinated to strategic hostility.

If the United States and China were to move toward full-scale AI decoupling, the global technology ecosystem could fracture into rival blocs with incompatible standards. Under such conditions, artificial intelligence would no longer function primarily as a tool for improving productivity or advancing human welfare.

The world has already experienced a Cold War. Few people truly want to see a second one emerge around artificial intelligence.

This is why the symbolism surrounding Trump's visit to China matters. Jensen Huang's appearance was not simply about chip exports or commercial lobbying. It represented a pivotal moment in which there is still an instinctive recognition among Americans that the future of AI cannot rest on separation between the world's two largest technological powers.

China and the United States will continue competing in the AI era. Healthy technological rivalry can accelerate innovation, improve efficiency, and push scientific development forward. The real question is whether both sides can prevent that competition from sliding into a technological war whose consequences would reach far beyond either country alone.