AMD

AMD in the enterprise

As of November 5, 2026. Every figure links to its source; company-reported numbers are labeled as such.

AMD sells AI to businesses mainly as hardware: Instinct GPUs (the chips that train and run AI models), EPYC server processors, and since 2026 complete "Helios" racks that bundle the two with networking and its ROCm software. Its buyers are a short list of very large AI spenders. OpenAI (up to 6 gigawatts of GPUs, agreed Oct 2025), Meta (up to 6 gigawatts, Feb 2026, a deal the Wall Street Journal valued at more than $100 billion) and Anthropic (up to 2 gigawatts, Jul 2026) have each signed multi-year commitments. Oracle plans a 50,000-GPU cluster. To win these deals AMD has handed out equity: OpenAI and Meta each received warrants for up to 160 million AMD shares, and AMD agreed to invest up to $5 billion in Anthropic.

The business is now large and growing fast. In Q3 2026, reported Nov 3, AMD's Data Center segment brought in $8.5 billion, up 97% from a year earlier, on total revenue of $13.4 billion (company figures). AMD said Anthropic, Meta, Microsoft, OpenAI and Oracle are already running workloads on the first Helios systems. It guided Q4 revenue to about $17 billion, roughly 66% year-over-year growth, as Helios and its 6th-generation EPYC "Venice" chips reach volume production.

What to watch: whether that ramp lands on time. Analysts at TIKR call the timing of Helios shipments, and customers' ability to bring data-center capacity online, the biggest risk to 2027 forecasts. Futurum notes that power availability limits how fast gigawatt-scale deals can be deployed, and expects OpenAI to keep relying heavily on Nvidia for its more mature software and networking. Revenue is concentrated in a few AI labs and hyperscalers, and the share warrants dilute existing AMD shareholders if their milestones are hit.