When US President Donald Trump met Chinese President Xi Jinping in Beijing between May 13 and 15, 2026, Taiwan was not part of the summit. Yet the island, widely known for its semiconductor dominance, remained central to nearly every strategic issue surrounding the talks, from AI infrastructure and semiconductor controls to the future of advanced chip manufacturing.

After the summit ended, Taiwan’s representative to the US on May 17 sat for an interview with CBS News. Ambassador Alexander Yui was pressed on Trump’s remarks during his Beijing trip, including a potential call with Taiwan President Lai Ching-te, pending US arms sales, and Taiwan’s position on declaring independence.

“We don’t want a war. We want peace and stability,” Yui told Face the Nation. He added, “We are sovereign, independent away from the Chinese People’s Republic of China’s attempt to swallow us as one of their own. They have never ruled or controlled Taiwan, ever.”

After the summit concluded, Taiwan’s Ministry of Foreign Affairs (MOFA) released two statements dated May 15, 2026.

Taiwan’s MOFA said it had noted Trump’s comments in a media interview and aboard Air Force One. It said the Trump administration, including Secretary of State Marco Rubio, had repeatedly stated that long-standing US policy toward Taiwan had not changed.

MOFA described China’s military threat as the only regional source of insecurity. It said Taiwan’s task was to maintain the status quo and strengthen self-defense.

In a second statement the same day, MOFA cited Rubio’s May 14 interview in Beijing, where he said that US Taiwan policy had not changed across administrations and that he opposed any forced or coercive change to the status quo. Foreign Minister Lin Chia-lung thanked the US for repeatedly explaining its support for peace and stability across the Taiwan Strait.

What Taiwan Controls in the AI hardware Supply Chain

Yui’s interview and the two MOFA statements land at a moment when Taiwan’s centrality to the global AI hardware landscape can be read directly from the earnings filings of the company at the centre of it all, Taiwan Semiconductor Manufacturing Company (TSMC).

In the first quarter of 2026, TSMC reported revenue of US$35.9 billion, according to its April 16 earnings release and Securities and Exchange Commission (SEC) filing.

Advanced nodes at 7 nanometers and below, which TSMC classifies as advanced, accounted for 74% of the company’s wafer revenue. High performance computing, the category that includes AI accelerators, was 61% of total revenue in the quarter.

In the fourth quarter of 2025, TSMC reported revenue of US$33.7 billion. Advanced technologies were 77% of wafer revenue in that quarter. 3- nanometer chips alone contributed 28% of wafer revenue, while five nanometer contributed 35%. The share of advanced nodes climbed through 2025, and for the full year, advanced technologies accounted for 74% of wafer revenue and HPC was 58%, according to TSMC’s Q4 2025 earnings release and SEC filing.

TSMC said N2, its most advanced process node, entered high-volume manufacturing in the fourth quarter of 2025 with good yield. N2 is now ramping in phases at TSMC’s fabs in Hsinchu and Kaohsiung. The company told investors it is expanding 3 nanometer capacity because of demand from AI customers, according to the Q4 2025 earnings transcript.

A look at the company’s capital spending tells the same story. TSMC said its 2026 capital budget would be between US$52 billion and US$56 billion. Of that, 70% to 80% would go to advanced process technologies. The company noted that process complexity had increased and that the capital required for N2 capacity was substantially higher than what N3 had required.

Why this Capacity Cannot be Moved Elsewhere

The US is trying to expand domestic fabrication capacity, but decoupling from Taiwan will not be easy. Currently the world’s most advanced semiconductor manufacturing expansion is still happening at Hsinchu and Kaohsiung.

TSMC said its first Arizona fab entered high-volume production in the fourth quarter of 2024. The second Arizona fab has completed construction. Moving in tools and installation has been planned for 2026 and volume manufacturing using 3 nanometer technology is expected in the second half of 2027.

The first Arizona fab produces 4 nanometer chips, according to TSMC’s 2025 annual report, while Taiwan’s lead remains its leading-edge N2 nanosheet manufacturing.

In Japan, TSMC’s second fab is now planned for 3 nanometer technology with volume production scheduled in 2028, the company told investors during its Q1 2026 earnings call.

However, the timeline for the most advanced nodes outside Taiwan stretches into the second half of this decade.

This delay is not merely capital related. In May 2025 TSMC declined invitations from India, Qatar and Singapore to build fabrication plants, according to DigiTimes.

The company, however, cited a shortage of skilled labor and absence of a mature supply chain ecosystem.

Any new fab will need chemicals, gases, ultra pure water, equipment suppliers and engineering talent. Taiwan accumulated these over a 50-year timeline. The Hsinchu Science Park for instance packs fabs, suppliers, and research institutions into a radius no other country has managed to replicate.

ASML, the Netherlands-based company, remains the sole supplier of extreme ultraviolet lithography machines. Without Extreme Ultraviolet (EUV) tools, economically competitive production at TSMC’s most advanced nodes becomes extremely difficult. China remains restricted from purchasing ASML’s most advanced EUV systems under export controls.

Where China Stands, For Now

Public analyses from the Council on Foreign Relations (CSIS), the American Enterprise Institute, and the American Affairs Journal differ on the pace of China’s progress in domestic chip production. However, there is broad agreement on one point, China has made gains despite constraints in leading-edge fabrication, advanced packaging, high-bandwidth memory, and yield.

CSIS noted in an April 2025 analysis that China had already demonstrated export controls would not impede its capacity to innovate over the longer term.

Huawei continues to face major challenges producing enough advanced GPUs because US controls restrict Semiconductor Manufacturing International Corporation (SMIC), China’s leading domestic foundry partly owned by the Chinese government.

In November 2024, Reuters reported that SMIC’s yield was around 20% for Huawei’s Ascend 910C and around 50% for the Ascend 910B, citing sources familiar with the results. To be commercially viable, advanced chips generally need to yield above 70%.

Estimates on Huawei’s production capacity also vary. The US Commerce Under Secretary for Industry and Security, Jeffrey Kessler, told lawmakers in June 2025 that Huawei’s 2025 Ascend production capacity would be “at or below 200,000”. CFR cited Semi Analysis as estimating that Huawei could produce up to 1.5 million AI chip dies in 2025 but only completed 200,000 to 300,000 due to high-bandwidth memory constraints.

Some TSMC chips were also making their way to Huawei. In a 2024 Reuters report, TechInsights, an information platform for the semiconductor industry, found a TSMC chip in Huawei’s Ascend 910B. TSMC duly notified the US Commerce Department. Reuters later reported that the chips had been routed through Sophgo, a Chinese semiconductor solutions company co-founded by Chinese billionaire Micree Zhan. Sophgo is currently heavily restricted by US trade controls.

China has been trying to compensate for the chip gap through scale. The American Affairs Journal described Beijing’s national computing network as an effort to pool computing resources across the country and use scale to offset technological weaknesses. Meanwhile, the US Bureau of Industry and Security (BIS) claimed in May 2025 that Huawei’s Ascend 910B, 910C and 910D chips were likely developed or produced in violation of US export controls. While China’s cluster strategy can reduce disadvantages created by weaker chips, it doesn’t close the fabrication gap at the leading edge, for now.

The Unresolved AI and Taiwan Tensions

The May 2026 summit in Beijing was the first US presidential visit to China since Trump’s 2017 trip. As for the outcome, several disputes remain unresolved, including rare earth export controls, AI semiconductor restrictions, as well as Taiwan arms sales.

According to a White House fact sheet released after the summit, China had committed to buying at least US$17 billion in US agricultural products annually in 2026, 2027 and 2028. It also said both governments would set up a US China Board of Trade and a US China Board of Investment. The White House fact sheet makes no mention of Taiwan.

Chip export controls were not presented as a summit breakthrough. US Trade Representative Jamieson Greer told Bloomberg TV that semiconductor export controls were not a major topic in the bilateral meeting. “We did not talk about chip export controls at the meeting,” Greer said, according to Reuters.

The H200 issue remains where it stood before the summit. On December 8, 2025, Trump announced the US would allow H200 and similar chip sales to approved buyers in China in exchange for a 25% export fee. The Bureau of Industry and Security implemented that decision through a final rule on January 13, 2026, shifting export license applications for Nvidia H200, AMD MI325X and similar chips from a presumption of denial to case-by-case review.

Bureau of Industry and Security (BIS) said applicants would need to show that the exports would not reduce global semiconductor production capacity currently available to US customers. They would also need to demonstrate that Chinese buyers had adopted export compliance procedures, including customer screening, and that the chips had undergone independent, third-party testing in the US to verify their performance and security.

China has instead increasingly shifted toward domestic alternatives including Huawei’s Ascend chips.

For Jensen Huang, Nvidia’s CEO, who joined Trump’s China trip as a late addition, what he has described as a roughly US$50 billion market opportunity in China remains out of reach in practice. Washington has cleared around 10 Chinese firms, including Alibaba, Tencent and ByteDance, to buy H200 chips, but Beijing has not approved the purchases and no shipments have been made.

The Taiwan Deputy Foreign Minister Chen Ming-chi said Trump’s remarks had caused “some unnecessary concern,” but that Taipei believed “nothing has changed.”

Trump’s comments, however, did create ambiguity around the future of Taiwan. Trump told media that he was undecided on new arms sales and suggested he might speak with Taiwan President Lai Ching-te. He added that the United States was not looking for anyone to declare independence. Chen said Taiwan’s defence needs were urgent, and that Taiwan needed US arms “as soon as possible”.

Conclusion

For Taiwan, the moat is the semiconductor ecosystem built over five decades, one that has helped the island hold strategic relevance despite China’s pressure. Taiwan is still where the world’s most advanced AI chips are being made, and where the next semiconductor ramp is already underway.

Despite billions in US semiconductor investment and China’s push for domestic AI chips, neither side has yet displaced Taiwan at the leading edge of advanced semiconductor manufacturing. That is why every discussion about AI power, export controls, and compute infrastructure eventually returns to the island.

Also Read: Trump-Xi AI Talks Test Washington’s Strategy to Restrict China’s AI Growth