AMD AI Roadmap 2027: Verano, MI500X & AI Rack
- AMD is speeding up its advancement cycle for CPU, GPU, and AI rack-scale solutions, aiming for yearly updates.
- Lisa Su, AMD's chief executive, highlighted the company's progress at the Advancing AI event.
- AMD's 2026 rack-scale AI plans include the Helios system, featuring the 256-core EPYC 'Venice' processor.
AMD’s AI ambitions surge toward 2027, promising groundbreaking advancements with the EPYC ‘Verano’ CPU and Instinct MI500X-series accelerators. The roadmap reveals a strategic push for improved performance and scalability in rack-scale AI solutions. News Directory 3 reports on how AMD plans to unveil its next-generation AI rack solution,setting the stage to challenge the competition. Discover the Helios system,sporting a 256-core EPYC ‘Venice’ CPU by the end of 2026,hinting at a new era in processing power. Get ready as TSMC’s A16 process technology may fuel these innovations. Discover what’s next.
AMD Accelerates CPU, GPU, and AI Rack Solutions Roadmap
Updated June 12, 2025
AMD is speeding up its advancement cycle for CPU, GPU, and AI rack-scale solutions, aiming for yearly updates. the company plans to introduce its next-generation EPYC ‘Verano’ CPU, Instinct MI500X-series accelerators, and a new rack-scale AI solution in 2027.
Lisa Su, AMD’s chief executive, highlighted the company’s progress at the Advancing AI event. She noted that development is well underway for the 2027 rack-scale solution,which will further improve performance,efficiency,and scalability using Verano CPUs and Instinct MI500X GPUs.
AMD’s 2026 rack-scale AI plans include the Helios system, featuring the 256-core EPYC ‘Venice’ processor. The EPYC ‘Venice’ processor is projected to deliver a 70% performance increase over the previous generation. The system will also use Instinct MI400X-series accelerators, which are expected to double AI inference performance compared to the Instinct MI355X, and Pensando ‘Vulcano’ 800 GbE network cards.
the second-generation rack-scale system, slated for 2027, will be powered by EPYC ‘Verano’ processors, Instinct MI500X-series accelerators, and pensando ‘Vulcano’ 800 GbE NICs.
While AMD has not released specific details about the second-generation rack-scale solution’s performance, a company-provided image suggests increased compute blades, which should boost performance density. This hints at higher performance and power consumption, positioning it to compete with Nvidia’s NVL576 ’Kyber’ system.
The production timeline for EPYC ‘Verano’ CPUs and Instinct MI500X-series accelerators aligns with TSMC’s A16 process technology rollout in late 2026. TSMC’s A16 process is its first production node to offer backside power delivery, a technology beneficial for data center CPUs and GPUs. It remains unconfirmed whether AMD’s 2027 products will use TSMC’s A16.
What’s next
Looking ahead, AMD’s focus remains on continuous innovation in the CPU, GPU, and AI accelerator space. The company’s commitment to yearly updates signals a dedication to pushing the boundaries of performance and efficiency in rack-scale solutions.