During a wide-ranging CNBC interview at CES, AMD CEO Lisa Su outlined the company’s strategy to capitalise on so-called physical AI—applications that power autonomous machines such as humanoid robots, industrial automation systems and self-driving vehicles. Su explained that AMD is integrating its CPUs, GPUs and adaptive silicon to deliver end-to-end compute platforms capable of real-time perception, reasoning and control. She highlighted partnerships with two leading robotics integrators, noting pilot deployments in logistics facilities that are projected to drive an incremental 15% uplift in data-centre GPU deployments by year end.

Su warned that AI compute requirements are set to jump more than 100× over the next 4–5 years as models grow in size and inference shifts to edge devices. She introduced the concept of the yottaflop—10^24 floating-point operations per second—to describe aggregate compute capacity needed across cloud, on-premise and endpoint systems. To meet this challenge, AMD plans a new generation of high-efficiency accelerators targeting 3× improvement in performance-per-watt compared with current offerings, a metric Su said will determine the pace of AI adoption in power-constrained environments.

At the Las Vegas conference, AMD unveiled four new products: the MI400 series data-centre GPU featuring 128GB of HBM3E memory; the EPYC 9004 processor family with up to 96 cores; the Ryzen AI 9000 mobile APU line for hybrid notebooks; and the Versal-inspired adaptive AI chip for industrial edge use cases. Su emphasised that this comprehensive stack enables customers to optimise performance across workloads—from large-scale training to on-device inference—supporting AMD’s goal of securing at least one design win in each of the hyperscale and automotive segments by mid-2026.

AMD confirmed it will report fiscal fourth-quarter and full-year 2025 results on February 3, 2026, after market close, with a management webcast to follow at 5:00 p.m. EST. The company also announced that CTO Mark Papermaster will present at the Morgan Stanley Technology, Media & Telecom Conference on March 3. Investors will be watching for revenue growth trends in data-centre GPUs, gross-margin progression and any updated guidance reflecting accelerated AI-related investments.