Advanced Micro Devices IncAMD launches Helios architecture, new processors and accelerators, and long-term roadmap to 2030.
AMD has announced the launch of its next-generation AI infrastructure portfolio, with the rack-scale AMD Helios architecture as the core solution, which is now entering commercial production and will be deployed in gigawatt-scale data centers of leading global technology companies. Dr. Lisa Su, CEO of AMD, stated that AMD is committed to delivering open computing platforms that enhance performance and cost-effectiveness, helping customers scale AI infrastructure from data centers to edge devices. At this launch, AMD announced strategic partnerships with major players. Anthropic is set to deploy up to 2 gigawatts of AMD Instinct MI455X on the Helios platform and engage in a long-term collaboration to bring Claude models to ROCm software development. Meanwhile, OpenAI is advancing joint AI stack development using the Triton framework with ROCm to optimize GPT-class workloads on Helios, with installation expected to begin in the fourth quarter of 2026 and scaling rapidly through 2027. Meta is currently testing gigawatt-scale systems featuring both 6th Gen AMD EPYC processors and the Helios platform in preparation for broad deployment. Cerebras is combining its low-latency computing with Helios to boost performance and cost efficiency. On the hardware front, AMD introduced the 6th Gen AMD EPYC Venice processor and the AMD Instinct MI455X and MI430X accelerators. The Helios architecture delivers up to 30% better cost per unit compared to competitors, alongside the launch of the ROCm.ai developer platform. Additionally, AMD revealed a long-term roadmap extending to 2030, including EPYC processors on Zen 7 and Zen 8 architectures, Instinct MI500 and MI600 Series cards, and an entry into the Physical AI market with AMD Kria AI solutions for autonomous robotics, as well as collaborations with AT&T and Cisco in the enterprise sector.
Advanced Micro Devices IncAMD launches Helios architecture, new processors and accelerators, and long-term roadmap to 2030.
Meta Platforms Inc.Meta is testing gigawatt-scale systems with AMD EPYC and Helios for broad deployment.
Cerebras Systems Inc. Class A Common StockCerebras combines its low-latency computing with Helios to boost performance and cost efficiency.
Cisco Systems Inc
AT&T Inc.Anthropic will deploy up to 2 gigawatts of AMD Instinct MI455X on Helios and collaborate long-term.
OpenAI advances joint AI stack development using Triton with ROCm to optimize GPT workloads on Helios.