Gary Sims discusses ARM’s upcoming 2026 CPU, codenamed “Travis,” which will feature the new Scalable Matrix Extension (SME) to significantly boost AI and machine learning performance on Android devices by accelerating matrix operations directly within the CPU. ARM is also revamping its CPU branding, introducing new categories like Lumix for mobile and moving away from the traditional Cortex naming, with an expected announcement likely between August and September.
In this video, Gary Sims discusses the anticipated but yet-to-be-announced new ARM CPU cores for mobile devices in 2026. Traditionally, ARM reveals new cores around May or June each year, but as of mid-July, no official announcements have been made. Despite the silence, information from a recent Competitex speech reveals that ARM is working on a new CPU codenamed “Travis,” alongside a new GPU called “Drager” and another GPU named “Marley,” continuing their naming conventions from previous years.
ARM is also restructuring its CPU naming scheme. Instead of the familiar Cortex and Cortex X series, ARM plans to introduce new branding categories: Neoverse for servers, Neva for PCs (mainly laptops), and Lumix for mobile devices. This means Travis might be part of the Lumix family rather than a Cortex CPU. Additionally, ARM is moving away from the traditional A-series and X-series labels, opting for terms like Ultra, Pro, Nano, and Pico, aligning with trends seen in companies like Apple and Dell.
A significant highlight of the upcoming Travis CPU is the inclusion of ARM’s Scalable Matrix Extension (SME), specifically version two. SME is crucial because it accelerates matrix multiplication and manipulation, which are fundamental operations in AI workloads. While GPUs have traditionally handled these tasks efficiently, integrating SME into the CPU allows for faster on-device machine learning without relying solely on GPUs, NPUs, or software-based solutions.
Gary draws a parallel with Apple’s M4 processor, which also incorporates scalable matrix extensions, demonstrating the growing importance of this technology in modern CPUs. The integration of SME in ARM’s upcoming CPUs promises to enhance AI performance on Android devices by enabling efficient matrix operations directly within the CPU, thus improving machine learning tasks’ speed and efficiency.
Finally, while the exact announcement date for Travis remains unknown, Gary speculates that ARM will likely reveal it before major chip announcements from companies like MediaTek later in the year, possibly between August and September. He encourages viewers to stay tuned to his channel for updates and detailed coverage once official information becomes available.