AMD Advancing AI 2026 Keynote Live Coverage - ServeTheHome

What the report says

ServeTheHome reported live from AMD’s Advancing AI 2026 event in San Francisco, describing it as AMD’s largest version of the enterprise and AI-focused conference to date. The publication said the event filled Moscone Center West and has expanded from a press-style announcement forum into a broader trade show with developer sessions and product displays, drawing comparisons to NVIDIA’s GTC.

The keynote, led by AMD Chair and CEO Dr. Lisa Su, focused on the company’s data center AI roadmap and the launch of Helios, AMD’s rack-scale AI compute system. According to the live coverage, Helios combines MI455X accelerator modules, EPYC “Venice” CPUs and Pensando Vulcano networking hardware. AMD said the MI455X uses 2nm compute chiplets, includes 432GB of HBM4 memory, and is intended to deliver higher memory capacity and FP4 performance versus prior systems.

ServeTheHome reported that AMD said Helios is now in full production, with shipments beginning in Q3 and a ramp into the second half of 2027. The company positioned Helios as its answer to NVIDIA-style rack-scale AI systems, a key category as frontier AI models increasingly require full-rack infrastructure rather than individual servers or accelerators.

The keynote also highlighted AMD’s view of rising demand from agentic AI and inference workloads. ServeTheHome cited AMD projections including a $1.4 trillion AI accelerator market by 2030, a $220 billion CPU market in the same year, and a $2 trillion silicon total addressable market growing at a 40% compound annual rate. Anthropic’s Tom Brown appeared during the event, and AMD said Anthropic plans to buy 2GW of Helios hardware, underscoring AMD’s push to win major AI infrastructure customers.