Half a Trillion in Silicon: Lisa Su’s AI Vision

When AMD CEO Lisa Su walked on stage at the Axios AI+ Summit, she didn’t talk about the next model drop or the latest benchmark drama. She talked about time. In her view, we are only in year two of a ten-year cycle that will dwarf today’s AI hype. Chips, data centers, and models are about to scale far beyond anything the market has priced in. And she called AI the most transformative technology of her career—on par with the internet itself.

Su wasn’t pitching the crowd on a product demo. She was sketching a construction plan for the next decade of the global economy. And she put a number on it: a $500 billion accelerator market emerging in just a few years, fueled by demand that has no precedent in tech history.

The Curve Ahead

Su framed the AI boom not as a frenzy of apps but as a long arc of infrastructure. Today’s viral chatbots and copilots, she argued, are only possible because of the industrial build-out beneath them: racks of GPUs, clusters of high-bandwidth memory, warehouses wired for megawatts of draw. That build-out is still in its infancy.

Her forecast was blunt. The next three to four years will see the accelerator market explode to half a trillion dollars, driven by hyperscalers racing to provision compute and enterprises scrambling to build in-house capacity. The demand for training and inference hardware will outpace even the most optimistic projections of model adoption.