GMI Cloud to build a $500M NVIDIA-powered AI data center in Taiwan as US demand queue hits 134 GW
GMI Cloud will build a $500 million AI data center in Taiwan stocked with NVIDIA chips, while US data center power demand is projected to grow from 76 GW in 2026 to 134 GW by 2030. The buildout is globalizing — and the binding constraint everywhere is the grid, not the silicon.
A half-billion-dollar NVIDIA facility in Taiwan is the buildout following the supply chain home. Locating AI capacity next to the fabs that make the accelerators shortens every logistics loop and hedges the geography of demand — and it extends the neocloud land grab beyond the US and Europe into the Pacific, where power and proximity now trade against each other.
The demand curve behind it is unforgiving. Seventy-six gigawatts of US data center draw in 2026 nearly doubling to 134 by 2030 is a grid problem masquerading as a tech story; utilities measure interconnection queues in years, and the industry's response — co-packaged optics to cut data-movement power, custom inference silicon, gigawatt campuses sited on generation — is all engineering aimed at the same wall.
Put alongside China's self-supplied gigawatt, the pattern is a global compute race with two supply chains. One runs through NVIDIA, TSMC, and allied grids; the other through domestic Chinese silicon and state-backed power. Both are constrained by electricity before chips, which is why every serious player is becoming an energy developer whether it wanted to or not.
MarketScreener — GMI Cloud to build $500 million AI data centre in Taiwan with NVIDIA chips → · EnkiAI — Semiconductors and AI chips — TSMC's $100B US expansion →