California-based GMI Cloud said it will launch a $500M AI data center in Taiwan with Nvidia’s (NVDA) chips.
GMI said the AI factory will serve as a key pillar of the region’s AI infrastructure, enabling enterprises to train and deploy AI models at unprecedented scale.
Powered by 7,000 Nvidia Blackwell Ultra graphics processing units, or GPUs, across 96 high-density GB300 NVL72 racks, the 16-megawatt Taiwan AI Factory can be expected to process close to 2 million tokens per second, the company added.
Built on Nvidia Blackwell architecture, the GMI Cloud Taiwan AI Factory supercomputer will feature Nvidia’s NVLink, Quantum InfiniBand, and Spectrum-X Ethernet networking as well as Nvidia BlueField DPUs.
“This data center is intended to become the blueprint for the heart of Asia’s AI future,” said GMI’s CEO Alex Yeh. “With thousands of next-generation NVIDIA GPUs running in synchrony, our AI infrastructure helps turn the world’s AI visions into reality.”
GMI Cloud is a GPU-as-a-Service provider and one of Nvidia’s cloud partners.
Besides the Taiwan project, GMI Cloud plans to build a new 50-megawatt U.S. data center, and is looking to seek an IPO in two to three years, Reuters reported.
The project with Nvidia is expected to generate about $1 billion in total contract value once fully operational, said Yeh, the report added.
Nvidia did not immediately respond to a request for comment from Seeking Alpha.