At CES 2026, AMD CEO Lisa Su used the industry’s largest stage to outline where the next era of AI is headed. The AI industry, he said during his keynote speech yesterday (Jan. 5), is entering the era of “yotta-scale computing,” driven by unprecedented growth in both training and guidance. The limit, Su argued, is no longer the model itself but the basis for integration under it.
“Since launching ChatGPT a few years ago, we’ve gone from about a million AI users to over a billion active users,” Su said. “We’re seeing AI adoption grow to more than five billion active users as it becomes integral to every part of our lives, like cell phones and the Internet today.”
Global AI computing capacity, he noted, is now on its way from zettaflops to yottaflops within the next five years. A yottaflop is a 1 followed by 24 zeros. “Ten yottaflops of computing power is 10,000 times more than what we had in 2022. There has never been anything like this in the history of computing, because there has never been a technology like AI,” said Su.
However, Su warned that the industry still lacks the computing power needed to support what AI will eventually be able to do. AMD’s answer, he said, is to build a foundation in the end—positioning the company as an architect of the next phase of AI rather than a supplier of isolated components.
That strategy centers on Helios, a scalable data center platform designed for multibillion-parameter AI training and prediction at scale. A single Helios rack delivers three AI exaflops, including Instinct MI455X accelerators, EPYC “Venice” CPUs, Pensando networking and the ROCm software ecosystem. The emphasis is on scalability, with systems built to grow around AI workloads rather than locking customers into closed, short-lived architectures.
AMD also previewed the Instinct MI500 Series, which is scheduled for launch in 2027. Built on the next-generation CDNA 6 architecture, the roadmap aims for a thousand-fold increase in AI performance compared to the MI300X GPUs launched in 2023.
Su emphasized that yotta-scale computing will not be limited to data centers. AI, he said, is becoming a local, everyday experience for billions of users. AMD announced the expansion of its on-device AI machine with Ryzen AI Max+ platforms, which can support models with up to 128 billion parameters using unified memory.
Beyond commercial products, Su has tied AMD’s road to important things in the public sector. Joined on stage by Michael Kratsios, President Trump’s science and technology advisor, who will be speaking at CES later this week, he discussed the US government’s Genesis Mission, an independent public initiative aimed at strengthening national AI leadership. As part of that effort, AMD’s powerful Lux and Discovery supercomputers are coming online at Oak Ridge National Laboratory, strengthening the company’s role in the nation’s scientific discovery and infrastructure.
The keynote closed with a $150 million commitment to AI education, consistent with the USAI Literacy Pledge—showing that, in AMD’s view, sustaining the yotta-scale ambition will depend as much on talent development as on silicon.
