'Same compute, fewer resources - More compute, same energy': AMD says it is on track to make AI four times more energy efficient
AMD says two of its 2030 racks will match 570 from 2024 using twenty times less power and reducing carbon intensity by twenty-eight times.
AMD's new rack-scale AI systems are now about four times more energy-efficient compared to its 2024 baseline, surpassing the company's threefold interim target by 33%. This achievement can be observed in two ways: either using fewer racks for the same compute power or achieving twenty times more compute with the same power consumption. If reversed-engineered using AMD's figures, this implies that each 2030 rack draws approximately 14.25 times more power than its 2024 counterpart.
AMD has published a progress report detailing its efficiency goals, which were established in June 2025. The numbers revealed in the report exceed the company's initial projections. Compared to the 2024 baseline, AMD's rack-scale AI systems are currently about four times more energy-efficient, exceeding the 3x efficiency target that had been anticipated for this stage. This represents more than double the historical industry trendline.
This development aligns with AMD's six-year commitment to deliver a twenty-fold improvement in rack-scale energy efficiency for AI training and inference by 2030. The focus on efficiency comes as AMD seeks to attract more AI customers.
AMD has a history of setting performance and efficiency goals. Its previous target, known internally as 30x25, aimed for a thirtyfold node-level improvement between 2020 and 2025, and the company surpassed expectations with a 38x performance gain. If the current efficiency gains hold as projected, AMD's offerings could deliver 20x more compute with the same power consumption in 2030.
This could be advantageous for various stakeholders, including AI startups and hyperscalers, who are increasingly pushing to increase compute power. By leveraging AMD's efficiency gains, these entities might benefit from the same performance while using less energy. Moreover, considering global data center electricity demand is projected to more than double by 2030, AMD's efficiency improvements could be particularly appealing as they align with the growing need for energy-efficient AI solutions.
Written by urgent.news from TechRadar's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
