AMD Targets 20-Fold Gain in AI Rack Efficiency by 2030

The rapid expansion of generative AI is making electricity, cooling and data-center infrastructure as important as raw chip performance. AMD’s latest roadmap addresses efficiency at the rack level, combining advances across processors, accelerators, memory, networking and system design. The strategy is intended to reduce the power and operating burden of deploying increasingly large AI models and computing clusters.

AMD expects its rack-scale AI solutions to deliver four times the energy efficiency of its 2024 baseline by 2026, with the improvement rising to 20-fold by 2030. If the roadmap is achieved, the company says two racks in 2030 could provide as much computing capacity as 570 racks did in 2024, sharply reducing the space, electricity and cooling needed for large-scale AI workloads.