Photo: Tima Miroshnichenko / Pexels
The hardware giant projects 3,400% growth in global token consumption as enterprise agentic workloads move from pilot programs to full-scale production.
Dell Technologies just put a number on how much AI infrastructure the world is going to need, and it’s a big one. At its Dell Technologies World event, the company projected that global token consumption will grow 3,400% by 2030, driven by enterprises moving AI projects out of the lab and into daily operations.
The inference economy is coming #
Training AI models grabs the headlines, but inference, the process of actually using those models to generate outputs, is where the compute demand is quietly exploding. Dell expects inference workloads to represent nearly two-thirds of all AI computing requirements in the coming years.
The driver behind this shift is what the industry calls “agentic AI,” systems that don’t just answer questions but autonomously execute multi-step tasks. Dell projects that these enterprise agentic workloads will become the dominant category of AI compute by 2028.
The company says more than 5,000 enterprises are already running production AI workloads on Dell infrastructure. The client list includes names like Lilly, Samsung, and Honeywell.
The numbers behind the bet #
Dell’s confidence is showing up in its financial guidance. The company raised its fiscal 2027 AI server revenue forecast from $50 billion to roughly $60 billion.
Dell reported $16.1 billion in AI server revenue for Q1 2027, alongside an AI backlog of $51.3 billion.
On the hardware side, Dell introduced the PowerEdge XE9812 system, which the company says delivers up to 10x lower cost-per-token compared to previous generations for large-scale AI inference.
Dell also unveiled its Deskside Agentic AI offerings, essentially on-premises AI systems designed for enterprise use. The pitch: an 87% reduction in token spending over two years compared to public cloud APIs.
Why hybrid is winning #
Dell’s strategy revolves around its AI Factory platform, built in collaboration with NVIDIA. The approach emphasizes hybrid infrastructure, a mix of on-premises and cloud resources, rather than forcing enterprises to go all-in on one model.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our