Dell and Intel tackle enterprise AI's runaway token costs with on-premises alternatives

The economics of enterprise artificial intelligence are forcing a reckoning. Customers deploying agentic workloads are exhausting their cloud token allocations far faster than anticipated, reshaping how organizations think about where inference computation should happen. The financial pressure is reviving interest in on-premises deployments and open-weight models as counterweights to expensive frontier models hosted in the cloud.

Helen O'Sullivan, AI business development manager and solutions specialist at Dell Technologies Inc., observed this pattern firsthand among her customer base in the Bay Area. "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. "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 and Tim Wood, sales director for the Northwest region at Intel Corp., discussed these dynamics during an appearance at Dell's "AI Leadership Symposium," where they spoke with theCUBE Research's John Furrier in an exclusive broadcast.

Rebalancing hardware for agentic workloads

Agentic artificial intelligence is shifting the processor requirements that enterprises must consider. While many organizations remain focused on graphics processing units, the nature of their workloads is evolving in ways that demand a different hardware mix.

"You can no longer do AI just for the sake of doing AI. You need to have a goal in mind that is in line with the corporation's goals," O'Sullivan said. Intel's approach to agentic AI architecture encompasses orchestration, inference and edge computing layers. O'Sullivan's team anticipated this shift in hardware utilization months before many customers recognized it, though adoption remains incomplete.

"We saw the rebalance and the heavier utilization on the CPU when it comes to agents. Customers are getting there," she said. "They're not quite there yet from the scope I'm seeing."

Testing before deployment

Organizations can experiment with different deployment models before committing resources to production environments. Sandbox testing of open-weight models allows enterprises to understand what their specific workloads actually require, according to Wood.

"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 explained.

Both Dell and Intel are adopting consultative strategies that evaluate existing infrastructure investments and compare capital expenditure models against operating expense approaches before recommending specific deployment architectures. O'Sullivan emphasized that comprehensive assessment of the entire technology stack is essential for enterprise AI scalability.

https://www.youtube.com/embed/Xg7XBk0Nweo?feature=oembed

"A lot of [customers] just see the [graphics processing unit] and [ask], 'How fast can I get the GPU?' or Vera Rubin or what have you," she said. "But it's about the data."