The Main Highlights
AMD Advancing AI 2026
AMD Advancing AI 2026
CEO Dr. Lisa Su’s keynote address at Advancing AI 2026 in San Francisco marks AMD’s strategic shift toward becoming a full-fledged infrastructure provider for the age of Agentic AI. The focus is on rack-scale systems, a shift away from pure training models toward mass inference, and close integration of CPUs and GPUs.
The new flagship system combines Instinct MI455X GPUs, EPYC CPUs (‘Venice’ with Zen 6 cores) and Pensando DPUs/NICs into a unified racing infrastructure. Helios offers up to 15 per cent more computing power and 50 per cent more HBM4 memory capacity than comparable platforms.
With the launch of the MI455 accelerators and the prospect of the MI500 and MI600 architectures, AMD is making massive strides in efficiency. The MI500 generation aims to increase inference throughput by a factor of 2,000 within four years.
Complex AI agents require powerful orchestration. AMD forecasts that the server CPU market will grow to over US$200 billion by 2030, driven by the 6th generation of EPYC processors.
The global market for artificial intelligence has reached a critical tipping point. Whilst in previous years model training consumed the lion’s share of computing capacity, in 2026 the world will, for the first time, use more energy and computing power for inference – with execution accounting for around 60 per cent of the total AI compute volume.
This development is being driven by agentic AI: autonomous agents no longer act merely as simple chatbots, but solve complex, multi-stage tasks in medicine, science and enterprise processes. This multiplies the demand for scalable, highly efficient data centres.
AMD extends the Ryzen AI Embedded processor family combining CPU, GPU and NPU acceleration to power robotics, industrial automation and intelligent embedded systems.
AAI 2026: AMD Delivers Leadership Heterogeneous Compute for Physical AI - AMD Newsroom
AMD Kria AI solutions extend robotics capabilities from body to brain and combine the Kria AI SOM, robotics developer platform and open software ecosystem to accelerate physical AI.
AAI 2026: AMD Introduces Open, Turnkey Integrated Platform for Physical AI - AMD Newsroom
AMD Ryzen AI platforms help developers, enterprises and creators run powerful AI locally, with more secure, responsive and cost-effective client systems.
AAI 2026: AMD Brings Personal AI Compute Closer to Users - AMD Newsroom
New AI infrastructure combines leadership compute, networking and open software, delivering rack-scale performance for at-scale AI.
AAI 2026: AMD Launches AMD Helios Rackscale Solution for Frontier AI - AMD Newsroom
Die Server-CPUs der AMD EPYC 9006-Serie bieten speziell entwickelte Lösungen für agentenbasierte KI, Unternehmensanwendungen und HPC-Workloads.
AAI 2026: 6th Gen AMD EPYC Server CPUs Power the Agentic Data Center - AMD Newsroom
To meet these enormous demands, AMD is adopting a holistic co-design approach combining CPU, GPU and networking. By combining proprietary hardware with the enhanced ROCm.AI software platform, developers and data centre operators are provided with a scalable alternative that reduces the total cost of ownership (TCO) and accelerates implementation.
Further information & industry insights
- Strategic partnerships: Leading companies such as OpenAI and Anthropic are working closely with AMD to integrate the new architectures directly into their infrastructures.
- Outlook: With a clear annual release cycle for data centre systems, AMD offers businesses reliable planning certainty for the coming years.
- What is the AMD Helios system?
Helios is a fully integrated rack system for enterprise AI that combines Instinct GPUs, EPYC CPUs and Pensando network components to deliver maximum computing power and memory density for large language models.
- Why is the focus shifting from training to inference?
As billions of people and autonomous AI agents use pre-trained models in their everyday lives, the main focus at present is on execution (inference) rather than on building the base models.
- What role do CPUs play in the age of AI accelerators?
Although GPUs handle the computationally intensive model calculations, high-performance server CPUs such as AMD EPYC are essential for data orchestration, agent sandboxes and the management of workloads in the data centre.