Nvidia doubles down on Taiwan with AI supercomputer as US trade pressure mounts

Nvidia CEO Jensen Huang reaffirmed the company’s deep ties to Taiwan on Monday, unveiling plans for a new Taipei headquarters and a powerful AI supercomputer powered by 10,000 of its cutting-edge Blackwell chips. The announcement came during his keynote address at the Computex technology show, as Nvidia continues to navigate geopolitical tensions, US trade policy, and global demand for its AI hardware, the Financial Times reported.

The supercomputer, developed in collaboration with Foxconn’s Big Innovation Company and the Taiwanese government, will be used by key players in Taiwan’s technology ecosystem — including Taiwan Semiconductor Manufacturing Company (TSMC), Nvidia’s principal chip manufacturer. The project is expected to cost hundreds of millions of dollars and reflects the company’s ongoing reliance on Taiwan’s advanced electronics infrastructure.

“Taiwan is the centre of the computer ecosystem,” Huang said, calling it “the largest electronics manufacturing region in the world.”

New hub and growing footprint

Huang also revealed that Nvidia will build a futuristic new office complex — called Constellation — in Taipei’s Beitou district, signalling an expansion beyond its current footprint in Taiwan. “We are growing beyond the limits of our current office,” he told attendees, unveiling a video that imagined the building as a spacecraft landing in the city.

The announcement underscores Nvidia’s dual-track strategy: maintaining deep ties to Taiwan while responding to US demands for more domestic manufacturing. The company has committed $500 billion over the next four years for US AI infrastructure projects, including partnerships with TSMC, Foxconn, and Wistron.

Tariffs, China, and the Trump administration

The move comes at a time of heightened trade friction. President Donald Trump’s administration recently revoked a Biden-era rule that restricted AI chip exports to dozens of countries — a shift that benefits Nvidia’s global ambitions. At the same time, Trump has tightened restrictions on exports to China, specifically targeting Nvidia’s H20 chip, a product designed to comply with earlier export controls.

During his Taiwan visit, Huang acknowledged the balancing act. While the company is still evaluating how to serve the Chinese market under new US restrictions, he made clear that further modifications to the H20 are no longer viable. Nevertheless, Nvidia plans to open a new research centre in Shanghai, signalling continued interest in China despite geopolitical hurdles.

Asked about reports of possible AI chip diversion to China, Huang denied any wrongdoing: “There’s no evidence of any AI chip diversion,” he said, responding to US lawmakers’ recent concerns.

Expanding AI dominance

At Computex, Huang introduced a new initiative, NVLink Fusion, which will allow Nvidia’s GPU and networking technologies to interface with competing chips. The goal is to embed Nvidia’s infrastructure at the core of an increasingly modular and fragmented AI landscape — even as companies like Google and Amazon build proprietary AI chips to reduce dependency on Nvidia.

“Nothing gives me more joy than when you buy everything from Nvidia,” Huang joked. “But it gives me tremendous joy if you just buy something from Nvidia.”

The company’s international push continues, following last week’s tour of the Gulf with President Trump, during which Saudi Arabia and the UAE pledged multibillion-dollar investments in AI infrastructure powered by Nvidia.

As Nvidia races to meet skyrocketing global demand while balancing strategic alliances and trade restrictions, Huang’s latest moves underscore both the company’s unrivalled position in the AI hardware market — and the complex politics that now shape its future.