AMD expands AI strategy and targets 20x energy efficiency increase

AMD has announced a comprehensive AI strategy focused on end-to-end deployment, spanning cloud infrastructure, enterprise AI, robotics, and embedded systems. At its Advancing AI 2026 event in San Francisco, the company emphasized a unified development experience that allows developers to build applications on local systems and deploy them across various environments using a common software framework.

A central component of this strategy is the evolution of the ROCm platform into a broader AI software stack and continued investment in open-source software. AMD reported that over three million Hugging Face models now run natively on its platforms.

Financially, AMD reported record second-quarter 2026 revenue of $11.5 billion, a 50% year-over-year increase, driven largely by surging demand for EPYC processors and Instinct AI deployments.

In terms of sustainability, AMD pledged to achieve a 20x increase in rack-scale efficiency for AI training and inference by 2030. The company aims to reduce electricity usage and carbon intensity through tighter co-optimization of silicon, memory, interconnects, and software. By 2030, AMD predicts its configured server racks will perform tasks using significantly less physical infrastructure compared to 2024 levels.