Advanced Micro Devices (AMD) Chair and Chief Executive Officer Lisa Su said global computing capacity remains far short of what will be required to support the rapid expansion of artificial intelligence, calling for a 100-fold increase in available compute power over the next five years.

Speaking at a keynote event ahead of the opening of the Consumer Electronics Show (CES), Su said the pace and scale of recent AI innovation had been “stunning,” but stressed that the industry was still in the early stages of a much larger transformation.

“AI is evolving very quickly from a tool into foundational infrastructure,” Su said. “While innovation over the past few years has been remarkable, we are really just getting started — and today’s global computing capacity is far from sufficient for what is coming.”

Su said worldwide demand for computing power is expected to surge from the current zetta-scale level to around 10 yottaFLOPS within five years — representing a 100-fold increase from today and a roughly 10,000-fold expansion compared with 2022.

She said such a generational leap would depend not only on breakthroughs in system architecture, but also on continued advances at the limits of semiconductor manufacturing technology.

“To support this next wave of AI, we need innovation across the entire stack — from process technology to packaging, system design and software,” Su said.

Su also previewed AMD’s future product roadmap, saying the company plans to launch its MI500 series processors in 2027. The new chips are expected to deliver performance up to 1,000 times that of the MI300 series, which AMD first introduced in 2023.

AMD has been ramping up its investments in data center and AI accelerators as it seeks to compete more directly with rivals such as Nvidia and Intel in the fast-growing market for AI infrastructure.