At CES 2026, AMD CEO Lisa Su announced that AI’s enormous growth is pushing the world toward “yottascale” computing, a level of processing power almost unimaginable just a few years ago.
How Big Is Yottascale?
Think of it this way:
- 1 exaflop = 1 billion calculations per second (the level of today’s fastest supercomputers)
- 1 zettaflop = 1,000 exaflops
- 1 yottaflop = 1 million exaflops
In simpler terms: 1 yottaflop = 10²⁴ operations every second, equivalent to running millions of today’s top supercomputers at once.
Why Does AI Need This Much Power?
It’s not just about making AI models bigger. AI is now being used constantly, by billions of people, in real-time applications, and inside everything from smart devices to self-driving systems. This shift from occasional AI use to “always-on” intelligence is creating explosive demand for computing power, especially for AI inference (using models, not just training them).
What Could Yottascale AI Enable?
- Healthcare: Real-time diagnostics, personalized medicine, faster drug discovery.
- Climate Science: Hyper-detailed weather and climate modeling.
- Autonomous Systems: Smarter robots, factories, and vehicles.
- Personal AI Agents: Always-available assistants that can reason and act on your behalf.
How Do We Get There?
AMD says yottascale won’t be one supercomputer, it will be a global network of AI systems working together across:
- Data centers (for heavy training and inference)
- The edge (in vehicles, sensors, robots)
- Personal devices (PCs, phones)
AMD is positioning its portfolio, from Instinct GPUs and EPYC CPUs to Ryzen AI processors, to power this “AI everywhere” vision.
Bottom Line
AI’s explosive growth is forcing the tech industry to think bigger than ever, moving from exascale to zettascale and now toward yottascale computing. According to AMD, this isn’t science fiction, it’s the necessary next step to make AI truly integrated into our daily lives.
