Microsoft inks $9.7B deal with Australia’s IREN for AI cloud capacity
What Happened
The deal will give Microsoft access to compute infrastructure built with Nvidia's GB300 GPUs, which will be deployed over phases through 2026.
Our Take
Same playbook as UAE. Australia can't rely on US exports for training compute, so Microsoft builds it locally with GB300s. Looks like infrastructure, actually a workaround.
The pattern's obvious: US companies + foreign governments + Nvidia GPUs = localized AI capacity that sidesteps export restrictions. Australia's copying the UAE move, and it'll work because there's no law against it—Nvidia makes chips, Microsoft builds the center, Australia provides dirt and power.
Don't expect US-based model training to stay centralized.
What To Do
Plan for regional inference endpoints; don't bet on centralized US-hosted training staying viable.
Cited By
React
