technologyScore 55/100Watch
Lisa Su: ~60% of global AI compute capacity will run inference in 2026, overtaking training for the first time
“We can say that in 2026 for the first time the world is using more AI compute to run models than to train them and we've seen that to a point where we think this year roughly 60% of the global AI compute capacity will be used for inference.”
Why it matters
A first-party quantification of the inference-vs-training mix shift, with inference becoming the dominant AI compute workload. Inference favors lower-cost, high-throughput accelerators, which is the core of AMD's MI-series AI thesis.
Investment implication
Supports AMD's positioning for inference workloads against NVIDIA, and broadly favors compute buyers (hyperscalers) and power/efficiency-oriented silicon as inference economics dominate AI capex allocation.