AMD presentó sus apuestas en servidores, PC y robots para ubicarse en el centro de la revolución de IA

AMD showed its new line of processors EPYC Venice for servers dedicated to AI, the GPU Instinct MI455X and the Helios rack solution. Lisa Su, CEO of AMD, presented new products for servers and AI including new EPYC Venice processors, Instinct MI455X GPUs and the integrated Helios rack solution at the Advancing AI 2026 event. AMD EPYC runs in 60% of Fortune 500 companies. The EPYC 9006 series, code-named Venice, features Zen 6 architecture with up to 256 cores and 512 threads, frequencies up to 5 GHz and PCIe Gen 6 connectivity. It offers 1.8x more performance than the previous generation (Turin). AMD Helios is already in full production: a complete rack with 72 GPU Instinct MI455X, sixth-generation EPYC processors, Pensando networking and ROCm software. Each rack delivers 2.9 exaflops of compute and 31 terabytes of HBM4 memory. Against NVIDIA's Vera Rubin NVL72, Helios promises up to 15% more peak compute, 50% more memory and up to 30% more tokens per dollar. The new GPU Instinct MI400 family includes MI455X for frontier AI and MI430X for scientific supercomputing. AMD presented ROCm.ai, the evolution of its development platform, now AI-native, with a new command line (ROCm CLI), AMD Skills integrated into programming assistants, and Hyperloom, an open-source agent system. The Gorgon Halo mini-computer (Ryzen AI Max 400 family) can run models up to 300 billion parameters locally with up to 192GB unified memory. AMD also presented Ryzen AI Embedded X100 processors for robotics and the AMD Kria AI platform for autonomous robotics, capable of making over 8000 control decisions per second.