NVIDIA is constructing Taiwan’s “first nationwide AI supercomputer” and opening a brand new workplace within the nation, CEO Jensen Huang stated throughout his Computex 2025 keynote speech on Monday. The world’s main provider of AI chips will likely be collaborating with Foxconn, TSMC, and the Taiwanese authorities for the supercomputer facility, which can home 10,000 of its newest Blackwell GPUs. Foxconn will present the AI infrastructure, whereas TSMC will use the system for superior analysis and improvement.
“Taiwan doesn’t simply construct supercomputers for the world,” Huang stated in his speech, delivered the day earlier than Taiwan’s Computex tech expo. “ We’re additionally constructing AI for Taiwan. Having a world-class AI infrastructure in Taiwan is admittedly necessary.”
“By constructing this AI manufacturing unit with NVIDIA and TSMC,” Foxconn CEO and chairman Younger Liu stated in a press release, “we’re laying the groundwork to attach folks in Taiwan in addition to authorities organizations and enterprises reminiscent of TSMC to speed up innovation and empower industries.”
Huang stated his firm’s engineering workers had grown “past the bounds of (its) present workplace” in Taiwan. In consequence, NVIDIA intends to accumulate a bigger workplace in northern Taipei dubbed the “NVIDIA Constellation.”
These bulletins come at a time of turbulence within the AI business after President Donald Trump imposed sweeping tariffs on nations with which the U.S. has a commerce deficit. Whereas his said intention was to encourage home manufacturing, such insurance policies could disrupt tech provide chains within the brief time period. As well as, the U.S. is obstructing gross sales of sure superior chips to China with no license, which NVIDIA predicts might price it as much as $5.5 billion.
In response, NVIDIA has sought to diversify its technique by increasing operations in Taiwan whereas additionally reinforcing its presence within the US. Sooner or later after Trump warned that semiconductor tariffs are coming within the “very close to future,” the corporate stated it will be constructing supercomputers within the U.S.
Final week, the U.S.-based chipmaker unveiled plans to promote 18,000 of its high-performance Grace Blackwell AI processors to Saudi Arabia as a part of a nascent alliance that mixes American tech with Saudi capital. In April, NVIDIA additionally confirmed that it’ll not hit the brakes on its plans to construct extra AI information facilities, even after the tariffs sparked issues that rising tech costs will scale back demand for AI.
SEE: U.S. & China Slash Tariffs in ‘Very Good Deal,’ Markets Soar
NVIDIA additionally unveils an open AI ecosystem, up to date robotic software program, and a compute market
The supercomputer venture and new Taipei workplace have been solely a part of NVIDIA’s broader slate of bulletins this week. The corporate additionally launched NVLink Fusion, a brand new open server platform that enables tech firms to co-design AI chips and infrastructure. The system provides firms better flexibility to construct customized AI techniques utilizing NVIDIA’s high-speed NVLink interconnects, broadening the corporate’s attain in information heart deployments.
NVIDIA additionally unveiled the primary replace to Isaac GR00T, its basis AI mannequin for humanoid robots, designed to assist them be taught and adapt in actual time. The newest model, Isaac GR00T N1.5, consists of GR00T-Desires, a simulation software that creates digital coaching environments to speed up robotic studying.
Lastly, NVIDIA launched DGX Cloud Lepton, a compute market that enables builders to hire and scale highly effective AI infrastructure on demand. NVIDIA additionally introduced a brand new line of DGX private AI supercomputers, launched in partnership with Acer, Asus, Dell, and different main {hardware} producers.