GMI Cloud plans to build a $500 million AI data centre in Taiwan, marking one of the company’s most significant infrastructure investments to date. Powered by Nvidia’s newest Blackwell-generation chips, the facility is designed to meet soaring global demand for high-performance AI computing.
The data centre is expected to go online by early 2026 and will feature roughly 7,000 GPUs arranged in high-density racks optimized for large-scale model training and inference. This setup will give GMI Cloud a major boost in processing capacity, enabling faster throughput and more efficient handling of advanced AI workloads. The project will draw about 16 megawatts of power, reflecting the scale and cooling requirements of next-generation accelerated computing.
GMI Cloud’s leadership has emphasized that the facility will serve both local and international AI customers. Demand for GPU compute from sectors such as cybersecurity, cloud services, manufacturing, and model-training platforms is already high, and the company reports that current utilization is near capacity. The new centre will act as an “AI factory,” providing on-demand compute resources for enterprises building or deploying AI tools.
Once operational, GMI Cloud expects the Taiwan data centre to generate substantial long-term revenue and strengthen the region’s growing AI ecosystem.