AMD Partners With Cerebras for Enhanced AI Inference Solutions
WHO:
Techmeme - AMD is collaborating with Cerebras, a rival AI server chip developer.
ServeTheHome - AMD CEO Dr. Lisa Su is leading the keynote at the Advancing AI event.
WHAT:
Techmeme - AMD is working with Cerebras to connect AMD's server racks to Cerebras wafers for an AI inference solution to make workloads faster.
ServeTheHome - AMD's Helios is in full production, with shipments expected to start in Q3 and ramping into the second half of 2027.
ServeTheHome - The combined solution of Helios and Cerebras's wafer scale engine will allow Cerebras to deliver ultra-low latency (ULL) with Helios providing heavy lifting, achieving 5x the performance of the wafer scale engine alone.
ServeTheHome - The combined solution will be available in Cerebras’s cloud service later this year.
Techmeme - AMD is combining AMD Helios and Cerebras wafer-scale chips for an AI inference solution.
ServeTheHome - AMD's Helios is a rackscale system that integrates MI455X GPUs, EPYC “Venice” CPUs, and Pensando DPUs.
ServeTheHome - AMD expects the total addressable market for AI accelerators to reach $1.4 trillion by 2030.
ServeTheHome - AMD anticipates a 40% compound annual growth rate (CAGR) for a total addressable market for silicon of $2 trillion by 2030.
ServeTheHome - Helios is in full production, with shipments expected to start in Q3 and ramping into the second half of 2027.
ServeTheHome - Anthropic announced they will purchase 2GW of Helios hardware from AMD.
ServeTheHome - MI455X architecture features 2nm compute chiplets and 3nm elsewhere, with 432GB of HBM4 memory.
ServeTheHome - MI430X is set to ship in the first half of 2027.
WHERE:
ServeTheHome - The Advancing AI event is taking place at the Moscone Center West in San Francisco.
WHEN:
ServeTheHome - The Advancing AI 2026 keynote is scheduled to start at 9:30 AM PT/12:30 PM ET/4:30 PM UTC.
ServeTheHome - Helios shipments are expected to ramp into the second half of 2027.
ServeTheHome - MI430X is expected to ship in the first half of 2027.
WHY:
ServeTheHome - AMD aims to address the growing demand for agentic AI, which requires more computing resources than earlier AI models.
ServeTheHome - AMD's strategy focuses on leadership computing products and powering AI everywhere.
HOW:
ServeTheHome - AMD is enhancing its software release pace for ROCm, now every six weeks, to improve developer interaction with AMD platforms.
ServeTheHome - AMD is collaborating with Cerebras to create a disaggregated solution combining Helios and Cerebras's wafer-scale engine for ultra-low latency applications.
(UPDATED July 24, 2026 00:18 UTC)
AMD is collaborating with Cerebras to integrate AMD's server racks with Cerebras wafers, aiming to enhance AI inference solutions for faster workloads. The Helios system from AMD is now in full production, with shipments set to begin in the third quarter and ramping up into the second half of 2027. This partnership promises to deliver ultra-low latency by combining Helios with Cerebras's wafer scale engine, achieving five times the performance of the engine alone, and the solution will be available in Cerebras’s cloud service later this year.
--
(UPDATED July 23, 2026 23:19 UTC)
AMD is making headlines with a collaboration alongside rival AI server chip developer Cerebras, as they work together to create a disaggregated solution that combines AMD Helios and Cerebras wafer-scale chips for advanced AI inference. At the Advancing AI event in San Francisco, AMD CEO Dr. Lisa Su is set to deliver the keynote address, where the company will highlight its Helios system, which integrates cutting-edge MI455X GPUs and EPYC “Venice” CPUs, and is expected to ramp up shipments in the second half of 2027. With a projected total addressable market for AI accelerators reaching $1.4 trillion by 2030, AMD is positioning itself to meet the increasing demand for agentic AI by enhancing its software release pace and focusing on leadership computing products.