AMD introduced a broad portfolio of new AI infrastructure products at its Advancing AI 2026 event, unveiling the Helios rack-scale AI platform alongside new EPYC processors, Instinct accelerators and AI software designed to support the growing demands of enterprise and agentic AI workloads.
The announcement positions AMD to compete across nearly every layer of AI infrastructure, from cloud data centers and enterprise servers to embedded systems and robotics, as organizations continue expanding AI deployments.
Helios Becomes AMD’s Flagship AI Platform
At the center of the announcement is AMD Helios, the company’s first rack-scale AI platform built specifically for large-scale AI training and inference.
Helios combines 72 AMD Instinct MI455X GPUs with 18 sixth-generation AMD EPYC “Venice” processors, Pensando networking technology and the ROCm software platform into a single integrated AI system.
AMD said the platform delivers up to 30% more AI inference tokens per dollar than competing rack-scale solutions, giving enterprises and cloud providers an opportunity to improve AI performance while lowering infrastructure costs.
Several major AI companies have already committed to deploying Helios, including OpenAI, Anthropic, Meta, Microsoft and Oracle, with systems expected to become available through manufacturers including HPE, Lenovo, Supermicro and others.
New AI Chips Expand Data Center Portfolio
AMD also introduced its latest generation of AI processors designed for cloud, enterprise and high-performance computing environments.
The new sixth-generation EPYC server processors are built to support agentic AI, traditional enterprise workloads and GPU host systems with higher core counts, increased memory bandwidth and improved efficiency.
The company also launched the Instinct MI400 Series GPU family, including the MI455X for AI inference and training workloads, the MI430X for scientific computing and sovereign AI deployments, and the MI350P accelerator for organizations upgrading existing AI infrastructure.
Software Push Focuses on Open AI Development
Alongside new hardware, AMD expanded its software ecosystem with the introduction of ROCm.ai, an AI-assisted development platform intended to simplify GPU software development.
The platform enables coding assistants including Claude, Codex and Cursor to understand AMD hardware and the ROCm software stack, helping developers build and optimize AI applications more efficiently.
AMD said leading AI frameworks including PyTorch, Hugging Face, vLLM and SGLang already support the latest Instinct hardware.
AI Expands Beyond the Data Center
The company also announced new embedded AI products aimed at robotics and edge computing.
These include Ryzen AI Embedded X100 processors, new Kria AI system-on-modules and the Kria AI Robotics Developer Platform, which combines CPU, GPU, NPU and FPGA technologies into a single platform for autonomous robotics and physical AI applications.
AMD said the products are designed to reduce vendor lock-in while accelerating development of intelligent robotic systems.
Long-Term Roadmap Extends Through 2030
Looking ahead, AMD outlined an aggressive infrastructure roadmap that includes next-generation EPYC processors, future Instinct GPU families and new Helios rack-scale systems through the end of the decade.
The company also said accelerating AI adoption across data centers, PCs, embedded systems and edge devices is expected to expand AMD’s total addressable market to approximately $2 trillion by 2030.
The Bottom Line
AMD’s Advancing AI 2026 announcements represent one of the company’s largest AI product launches to date, spanning hardware, software, networking and robotics. By introducing the Helios rack-scale platform while expanding its processor, accelerator and AI software portfolios, AMD is positioning itself as a full-stack AI infrastructure provider capable of competing across enterprise, cloud and hyperscale deployments.