The AI buildout got real
AMD says HUMAIN is now delivering production AI compute in Saudi Arabia using AMD Instinct MI355X GPUs, EPYC CPUs, and Cisco Silicon One-based networking. Translation: this is no longer the “someday, maybe” stage. The stuff is live.
Why investors should lean in
That matters because AI infrastructure deals are all about turning hype into actual watts, racks, and revenue. When a customer starts running production workloads, you’re past the demo reel and into the part where chips, servers, and networking gear actually get used.
The bigger prize
AMD, Cisco, and HUMAIN also plan to deploy up to 250 MW of AI infrastructure in the next phase of the buildout starting in 2027. That’s a chunky footprint, and it gives AMD another international AI infrastructure story to point to while it tries to keep up the momentum in a market that still loves anything with “AI” attached.
Big picture: for AMD, this is another reminder that the AI race isn’t just about training models in Silicon Valley. It’s about who gets plugged into the global buildout — and who gets paid when the lights stay on.