With GPU-powered AI breakthroughs getting the lion’s share of press (and funding) in 2025, it’s tempting to imagine that CPUs are yesterday’s information. Current predictions anticipate that GPU and accelerator installations will enhance by 17% yr over yr by way of 2030. However, in actuality, CPUs are nonetheless chargeable for the overwhelming majority of at this time’s most cutting-edge scientific, engineering, and analysis workloads. Evan Burness, who leads Microsoft Azure’s HPC and AI product groups, estimates that CPUs nonetheless assist 80% to 90% of HPC simulation jobs at this time.
In 2025, not solely are these methods removed from out of date, they’re experiencing a technological renaissance. A brand new wave of CPU innovation, together with high-bandwidth reminiscence (HBM), is delivering main efficiency beneficial properties— with out requiring pricey architectural resets.
To be taught extra, watch the brand new webcast “Powering HPC with next-generation CPUs.“
This content material was produced by Insights, the customized content material arm of MIT Know-how Evaluation. It was not written by MIT Know-how Evaluation’s editorial employees. It was researched, designed, and written by human writers, editors, analysts, and illustrators. AI instruments that will have been used had been restricted to secondary manufacturing processes that handed thorough human overview.
