Dell, Intel Executives Describe Rising Enterprise AI Costs as Agents Scale

Some enterprise AI deployments are facing cost pressure as agentic applications perform more planning, tool use, code execution and validation, prompting consideration of hybrid infrastructure that combines cloud models with local computing.

Helen O’Sullivan, an AI business development manager and solutions specialist at Dell Technologies, said customers she supports in the Bay Area used token allocations planned for months within weeks.

“A lot of our customers I support in the Bay Area, they’re dealing with this real immediate struggle of running through their token budget in a matter of weeks that they had allotted for months,” O’Sullivan said.

O’Sullivan said enterprises could continue using frontier models while shifting some workloads to open-source models and on-premises infrastructure to reduce pressure on cloud token capacity.

“You still use those frontier models, but you use the open source and infrastructure on-prem to at least eliminate some of the token stress and token maxing going on right now,” O’Sullivan said.

Tim Wood, Intel’s sales director for the Northwest region, said companies may already have unused CPU capacity in their data centers that could support parts of newer AI workloads.

“A lot of enterprises already have CPUs in their data center. There may be existing resources and headroom in the infrastructure they already have that’s capable of doing some of these next-generation apps and workloads,” Wood said.

Intel’s September testing found that an instance built around its Xeon 6 processor delivered up to 1.66 times the agent-task throughput of an instance based on a fifth-generation AMD EPYC processor during a 60-minute healthcare workload evaluation. In a separate test, Intel said a Xeon 6990E+ system remained below a 200-millisecond sandbox-creation latency target until approximately 313 concurrent provisioning requests.

Those figures come from Intel’s vendor-produced tests and describe specific evaluation conditions rather than performance across all enterprise workloads.

Dell has promoted local and on-premises deployments to address cloud costs, latency and data-sovereignty constraints. Dell’s comparison methodology estimates that its Deskside Agentic AI systems can reduce spending by up to 87% compared with cloud APIs over two years.

The discussion comes as Intel emphasizes enterprise infrastructure for AI inference, with workloads assigned among CPUs, GPUs and other accelerators according to the task.

The focus for enterprises is how to divide workloads among cloud frontier models, open-weight systems and infrastructure already installed in their data centers as agentic applications consume computing resources more intensively.