Dell and Intel address enterprise AI budget constraints and CPU shifts
Rising cloud token costs for agentic workloads are pushing enterprises to reconsider on-premises infrastructure and open-weight model strategies.
Key points
- Cloud token budgets for agentic AI are exhausting faster than planned, straining enterprise finances.
- Dell and Intel executives highlight a shift toward CPU utilization for orchestration and inference tasks.
- On-premises infrastructure and open-weight models are returning to deployment strategies to mitigate costs.
Dell Technologies and Intel are addressing the growing financial pressure on enterprises deploying agentic artificial intelligence workloads. According to a report by SiliconANGLE, rising cloud token costs are forcing companies to rethink where inference runs, with many exhausting monthly budgets in mere weeks. This immediate struggle is reshaping deployment strategies as organizations seek to balance frontier model usage with more cost-effective alternatives to prevent rapid budget depletion.
Helen O’Sullivan, AI business development manager at Dell Technologies, noted that customers in the Bay Area are facing severe token budget exhaustion. She stated that allocations meant for months are being consumed in weeks, prompting a return to on-premises infrastructure. O’Sullivan explained that using open-source models locally helps eliminate token stress while still leveraging frontier models for specific tasks. This hybrid approach allows enterprises to manage the heavy financial load associated with scaling agentic workloads in cloud environments.
Hardware and Infrastructure Shifts
The rise of agentic AI is altering the balance of processor demand, though many buyers remain focused on graphics processing units. Tim Wood, sales director for Intel’s Northwest region, discussed how agentic architectures span orchestration, inference, and edge computing. O’Sullivan noted that her team anticipated a heavier utilization of central processing units for agents months ago. However, she observed that many customers have not yet fully shifted their scope from GPU-centric planning to the broader CPU requirements needed for effective agent orchestration and execution.
Strategic Deployment Considerations
Enterprises are advised to evaluate deployment options before committing to production environments. Wood explained that testing open-weight models in sandbox environments can help companies assess their specific workload requirements. This step allows organizations to determine the most efficient hardware mix before scaling. The discussion, which took place at the Dell “AI Leadership Symposium,” highlighted the need for goal-oriented AI strategies. O’Sullivan emphasized that AI initiatives must align with corporate goals rather than being pursued for their own sake.
What should you do?
Organizations should audit current cloud token consumption rates to identify workloads that are exhausting budgets faster than anticipated. Security and IT leaders need to review their hardware procurement strategies to ensure they are not solely focused on GPUs. Evaluating the potential for CPU-based orchestration and inference can reveal cost-saving opportunities. Companies should also consider integrating open-weight models into their testing pipelines to reduce reliance on expensive cloud-based frontier models for all tasks.
Background: Edge computing
Edge computing processes data near its source instead of sending it to a distant data center. This approach lowers delay, saves bandwidth, and keeps sensitive information local. It shifts security responsibilities from the core network to individual devices.
Read the full guide: Edge Computing Explained: Why Processing Data Locally Changes Everything
What to do and how to stay safe: Dell Technologies
- Monitor cloud token usage metrics closely to detect rapid budget exhaustion in agentic AI workloads.
- Evaluate existing CPU infrastructure for potential use in AI orchestration and inference tasks to reduce GPU dependency.
- Test open-weight models in isolated sandbox environments to assess performance and cost before production deployment.
- Align AI project goals with broader corporate objectives to ensure strategic and financial viability.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Why are enterprises moving away from exclusive reliance on cloud-based AI models?
Rising cloud token costs are causing companies to exhaust their budgets in weeks rather than months, prompting a shift to on-premises solutions.
How is agentic AI changing hardware requirements for enterprises?
Agentic workloads are increasing the demand for central processing units for orchestration and inference, shifting focus away from solely graphics processing units.
What strategy do Dell and Intel recommend for managing AI costs?
They suggest using a hybrid approach that combines frontier models with open-weight models running on on-premises infrastructure to reduce token stress.
