On July 9, 2026, AMD said it will help build the next generation of "gigascale" AI campuses in partnership with 5C, a North American AI-infrastructure provider, centered on AMD's Helios rack-scale platform. Investors liked it: AMD shares rose roughly 7-8% on the day, extending a run that has lifted the stock well over 100% this year.
Moving up the stack
The pitch is that frontier AI can no longer be assembled from parts bought separately. AMD and 5C will produce "validated, repeatable rack-scale reference designs" in which "compute, power, cooling, networking, and operations are planned together." It is a bid to sell not just chips but the blueprint for an entire AI factory — the bundled approach Nvidia has used to lock in customers.
What Helios is
Helios is AMD's flagship rack-scale system, built around its next-generation Instinct MI400-series accelerators and detailed earlier in 2026, with availability targeted for the second half of the year. Notably, the announcement itself named no specific GPU model, capacity or contract terms — the reference to Helios is the technical anchor, not a spec sheet.
No numbers, yet
This is a strategic, qualitative deal. AMD and 5C disclosed no committed gigawattage, GPU count, dollar value or timeline for the campuses. A widely cited figure — "over 1.5 gigawatts of roadmap capacity" and the ability to "power hundreds of thousands of GPUs" — describes 5C's business overall, not this collaboration. Initial deployments are underway in Ohio and Memphis, each serving different "neocloud" customers, the specialized GPU-cloud operators driving much of the current buildout.
The Nvidia contest
The move plants AMD deeper in the "AI factory" fight. "The AI industry has reached a point where performance at scale requires deeply integrated partnerships across all areas of the AI factory," said Andrew Dieckmann, who leads AMD's data-center GPU business, arguing that Helios "delivers leadership performance and TCO for the next generation of agentic AI." 5C CEO Jonathan Ahdoot framed the campuses as "tightly integrated ecosystems" rather than "chips or data centers in isolation."
