For 54 years, AMD has been building processors. For the past decade, the most consequential application of that work has been in artificial intelligence — specifically in the accelerators that make large-scale AI training possible at speeds and efficiencies that previous generations of hardware could not approach. GITEX 2023 offered AMD a platform to demonstrate what that means in practice.
The clearest example is LUMI, the most powerful supercomputer in Europe and one of the fastest in the world, powered by AMD EPYC processors and AMD Instinct AI accelerators. LUMI is being used to develop ComPatAI, a system trained on millions of images of cancerous tissue to recognise the earliest signs of cancer development — dramatically improving early diagnosis rates. A second project, Dawsonia, uses LUMI to digitise decades of handwritten climate data, feeding historical records into supercomputer climate models to improve the understanding of long-term trends.
These are research-scale applications, but AMD’s AI infrastructure has also arrived in consumer hardware: the Ryzen 7040 Series mobile APUs include a dedicated on-silicon AI co-processing block, the first of its kind in a consumer notebook. And at the interplanetary scale, AMD adaptive processors are present in the Perseverance rover on Mars — enabling computer vision and AI-powered image analysis on a different planet for the first time. The MI300 — the first server APU to combine the latest AMD CDNA 3 accelerator with Zen 4 processor architecture and high-capacity HBM memory — delivers up to 8x the AI training speedup of its predecessor.
AMD’s position as the provider of AI infrastructure across supercomputers, consumer devices, and space exploration simultaneously is a consequence of decades of investment in processor design. The AI era has made that investment visible.