Dell sees AI inference demand overtaking training as agentic workloads accelerate

Dell $DELL says AI inference demand has now surpassed training demand and is becoming a major source of infrastructure consumption across the industry.
The company expects inference token demand to increase roughly 87-fold to 3,600 quadrillion tokens by 2030, compared with approximately 5x growth in training demand.
Dell also expects enterprise agentic AI to become the single largest workload by 2028.
The forecast suggests the next phase of AI infrastructure spending could increasingly be driven by running models at scale rather than simply training increasingly powerful systems.