U.S. based cloud services provider GMI or General Machine Intelligence Cloud has stated it will build a $500 million artificial intelligence data center in Taiwan with the support of U.S. chipmaker Nvidia.
The data center will come online by March 2026 and will run on Nvidia’s new Blackwell GB300 chips. This facility will include about 7,000 GPUs across 96 high-density racks, capable of processing nearly two million tokens per second. It will draw around 16 megawatts of power.
GMI Cloud Founder and CEO Alex Yeh said Taiwan needs more data centers as “strategic assets” to support its AI development, adding that the island’s power-supply challenges can be remedied. He said AI demand has been strong with the company’s GPU utilization “almost full.”
READ: Meta set to finalize $30 billion financing deal for Louisiana data center (
“You want to promote local ecosystems — you have to build the data center first, you have to build the AI cluster first,” he said. He also said that the project is expected to generate about $1 billion in total contract value once fully operational.
GMI also said that 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.
This comes at a time when tech companies are investing massively in AI infrastructure. Meta recently signed a $27 billion financing deal with Blue Owl Capital to fund the Hyperion data center in Louisiana, which will be the company’s biggest data center till date. Blackrock recently signed a deal with Spain’s ACS worth $26.8 billion to develop data centers.
Nvidia CEO Jensen Huang has previously referred to such clusters as “AI factories” and has in the past year also announced deals to sell its most advanced GPUs to projects in Saudi Arabia and South Korea. President Trump has said he wants the top AI semiconductors such as Nvidia’s Blackwell chips reserved for U.S. companies.
Other AI infrastructure projects recently announced in Taiwan include a 100-megawatt AI data center project announced by Foxconn and Nvidia in May. GMI could, which is a GPU-as-a-Service provider and one of Nvidia’s cloud partners, already operates data centers in the United States, Taiwan, Singapore, Thailand and Japan.
GMI also plans to build a new 50-megawatt U.S. data center. The company is also planning to seek an initial public offering in two to three years.