AMD · Earnings call · 2026Q3T · Full report

AI Infrastructure Demand

ADVANCED MICRO DEVICES INC · 2026-11-03 · Importance 58 · Surprise 42

Frontier labs, hyperscalers, and cloud providers expanded Instinct deployments for inference and training, driving record data center AI demand. AMD began volume shipments of MI455X, and Anthropic, Meta, Microsoft, OpenAI, and Oracle brought their first Helios racks online and started running models and workloads. Management said Helios deployments would increase as production ramps and OEMs including HPE, Dell, Lenovo, and Super Micro begin shipping systems; customers are also planning multi-gigawatt deployments around future MI500 products.

Key facts

• AMD said Helios MI450 revenue shipments began in Q3 2026, with a significant step-up planned in Q4 and continued ramp through 2027.
• AMD said customer engagements for MI500 were progressing very well, with customers planning multi-gigawatt deployments.
• AMD said it had good visibility into 2027 data center infrastructure needs, including land, power, and shells, and was working hand-in-hand with customers.
• AMD said enterprise AI deployment was developing more slowly than deployments by frontier labs and hyperscalers, and it had introduced enterprise-specific GPUs to grow its enterprise footprint.
• AMD said Helion’s CPU-only racks target hyperscalers and large cloud providers, while enterprise deployments could combine CPUs and GPUs.
• AMD began shipping MI455X in volume in Q3 2026, and lead customers brought their first Helios racks online.
• AMD expected Helios deployments to increase substantially in Q4 2026 as production ramped and OEM and ODM partners began shipping systems.