U.S. cloud services provider GMI Cloud is preparing a major expansion in Asia as it moves forward with a $500 million artificial intelligence data centre in Taiwan, backed by leading chipmaker Nvidia NVDA.O. The company confirmed the development on Monday, underscoring Taiwan’s growing importance as a global AI hub.
The facility is scheduled to go live by March 2026 and will be equipped with Nvidia’s new Blackwell GB300 chips. According to the company, the centre will house roughly 7,000 GPUs distributed across 96 high-density racks, enabling nearly 2 million tokens per second in processing capacity. The data centre will consume around 16 megawatts of power, highlighting the scale of infrastructure required to support modern AI workloads.
GMI Cloud Founder and CEO Alex Yeh described Taiwan as in need of more data centres that function as strategic assets for its fast-accelerating AI ecosystem. While acknowledging the island’s power-supply constraints, Yeh said the challenges are solvable. He noted that demand for AI compute has remained intense, with GMI Cloud’s GPU utilisation nearly at full capacity.
“You want to promote local ecosystems; you have to build the data centre first, you have to build the AI cluster first,” Yeh said.
The investment arrives at a moment when global technology leaders are racing to expand AI infrastructure. Nvidia continues to benefit from this surge, generating substantial revenue from high-performance chips used in data centres worldwide.
Nvidia CEO Jensen Huang often refers to these massive clusters as “AI factories.” Over the past year, Nvidia has secured high-profile deals to supply top-tier GPUs to large-scale projects in Saudi Arabia and South Korea. U.S. President Donald Trump has also previously said that the most advanced chips, including Nvidia’s Blackwell line, should be reserved for U.S. companies.
Taiwan has become an active site for AI infrastructure investments. Earlier this year, Foxconn 2317.TW and Nvidia announced plans for a 100-megawatt AI data centre in the region, signaling rising industry momentum.
GMI Cloud already operates facilities across the United States, Taiwan, Singapore, Thailand and Japan, positioning itself as a major player in the GPU-as-a-Service space and an established Nvidia cloud partner. Beyond the Taiwan expansion, the company also plans to build a 50-megawatt data centre in the United States and is considering an initial public offering within the next two to three years.
Yeh expects the collaboration with Nvidia to generate about $1 billion in total contract value once the Taiwan AI factory becomes fully operational. Early customers include Nvidia, cyber-security leader Trend Micro 4704.T, electronics manufacturer Wistron 3231.TW, Chunghwa System Integration, data-infrastructure specialist VAST Data and industrial technology firm TECO 1504.TW.
Reporting by Wen-Yee Lee. Editing by Brenda Goh and Lincoln Feast. Our Standards: The Thomson Reuters Trust Principles.
