AMD CEO Lisa Su links AI inference surge to potential $1.5T demand by 2030, says roughly 60% of the market will shift

AMD CEO Lisa Su said AI inference is poised to become a major driver of semiconductor and compute demand, projecting as much as $1.5 trillion in related spending by 2030. The comments, reported by Yahoo Finance, were framed as part of a broader shift in AI workloads, where models are increasingly used to generate answers, rankings, and automated actions rather than only being trained from scratch.

Su added that she expects inference-focused compute to capture about 60% of the market going forward. That share estimate suggests AMD is positioning its AI hardware roadmap not only around accelerating training, but also around serving the steady, ongoing compute needs of deployed AI systems, including what the industry often calls “AI agents” that can plan and execute tasks.

The remarks reflect a common industry thesis that inference workloads will scale rapidly as AI moves from pilot deployments to everyday use across devices, software applications, and enterprise workflows. In that framing, the performance and efficiency of AI chips for inference can matter as much as raw training throughput, because inference runs continuously and can dominate total compute consumption over time.

For AMD, the market implication is that demand tied to inference chips, networking, and systems that support end-to-end deployment could expand faster as AI adoption broadens. The CEO’s emphasis on market share also indicates AMD views inference as an enduring platform shift, not a short-lived trend limited to early adopters.

Still, the Yahoo Finance post does not provide additional detail on how the $1.5 trillion figure is calculated, which segments are included, or whether the estimate is limited to specific products or broader categories of compute spending. It also does not break down the estimate by geography, customer type, or the mix of accelerators versus CPUs and networking equipment.

There is also no disclosure in the reported remarks of timing for new products, guidance for specific revenue lines, or updates to AMD’s financial outlook. Without those specifics, the comments appear to be a directional view of the AI compute opportunity rather than an earnings catalyst tied to a particular quarter.

Sector-wide, the inference thesis matters because it changes how hardware vendors compete. If deployed AI workloads are the main volume driver, buyers can prioritize power efficiency, memory bandwidth, software performance, and total system costs, especially in data centers running large fleets of models.

What to watch next is whether AMD links the inference opportunity to concrete milestones, such as product introductions, platform upgrades, or customer deployments, and whether future investor communications quantify how much of the company’s AI pipeline is oriented toward inference and agent workloads.