As enterprise AI moves into production, Dell’s infrastructure strategy is taking shape around a practical constraint: autonomous agents require more than computing capacity. To run those agents at scale, businesses need governed data, an environment built for sustained operation and a deployment model they can afford. Technology leaders increasingly need to weigh where AI tools will run alongside which ones to acquire.
Sam Grocott, senior vice president of product marketing at Dell Technologies Inc., views execution, rather than insufficient ambition, as the challenge for enterprises. Dell’s approach reflects that assessment, connecting existing business data to AI-ready infrastructure and giving customers more control over where workloads run and what they spend.
Two decisions now shape that investment:
- Where agents operate: Data location, latency and governance requirements determine whether a workload belongs on a PC, in a data center or in the cloud.
- How capacity expands: Separating storage from compute lets enterprises adjust each independently, without a full infrastructure upgrade every time demand grows.
Enterprise Data Determines Where AI Infrastructure Belongs
Existing enterprise data is central to Dell’s strategy. Most data remains on-premises, Travis Vigil argues. Moving processing, management and automation closer to it can reduce expense and improve efficiency, he says. His position makes data locality an architectural choice with a direct business consequence: organizations may be able to extract value without first relocating the information that supports their operations.
Dell’s AI Data Platform addresses that challenge by bringing storage, security and data management together for enterprise AI workloads. Its role is to connect business information with the infrastructure that applications and agents need. Dell has also introduced further capabilities intended to provide more consistent data access, extending the platform’s focus beyond storing information toward making it usable in production.
Consistency becomes more consequential as applications leave the pilot stage. For Paul Nashawaty, principal analyst for theCUBE Research, reliable access to trusted, governed data during development and operation gives enterprises a competitive advantage. He sees enterprises that link data infrastructure with production-ready application delivery as better placed to generate measurable returns from AI spending.
Even a capable agent depends on the information it can reach and the controls governing access, a dependency executives need to account for. Investment in models and computing power can therefore outpace business results if data readiness remains unresolved.
Hybrid deployment follows from this constraint. Companies must incorporate AI into environments that already contain essential operational information, while deciding which capabilities should sit alongside that information and which should run elsewhere. Dell’s platform strategy attempts to make that connection manageable without requiring every workload to follow the same deployment path.
Autonomous Agents Need a Dedicated Operating Environment
Agentic AI can automate complex processes that run over extended periods, reducing the work employees must handle directly. That operating pattern also changes infrastructure requirements. Greater system autonomy puts continuous monitoring, security controls and orchestration at the center of operations.
John Roese, global chief technology officer and chief AI officer at Dell, argues for an enterprise environment built specifically to develop and operate agents. In his view, legacy infrastructure should supply that environment with data and services rather than determine its entire architecture. Adding isolated AI features to an existing estate leaves the deeper operating requirements unresolved.
This is a significant distinction for adoption planning. Enterprises may retain established systems while creating a separate foundation for autonomous execution. The transition then becomes a question of integration and control, rather than simply extending familiar infrastructure with another application.
Security architecture must support that separation. An agent environment needs oversight of autonomous activity, alongside governance of the information it uses. Local deployment can address some concerns about sending data outside the organization, but location alone does not resolve the need for monitoring and operational management.
Dell’s AI Data Platform fits into this broader design as the connection between enterprise information and AI infrastructure. Its strategic value depends on how well that connection supports reliable operations. Access to data is necessary; maintaining control as agents use it is the production requirement.
Workload Placement Becomes a Financial Decision
Production introduces a different cost discussion from experimentation. Businesses must account for token usage, inference expense and information movement while deciding among public cloud, private cloud, edge devices, endpoints and data centers. The appropriate location depends on the workload rather than a universal preference for local or cloud execution.
Dell’s collaboration with Nvidia Corp. on Deskside Agentic AI targets enterprises that want an alternative to running agents entirely through cloud services. Local operation can keep information within the business and reduce dependence on external AI services. Those are potential advantages, not a blanket guarantee of lower spending.
Jon Siegel, senior vice president of Dell portfolio marketing, presents the PC as a starting point for employee experimentation. He describes locally generated tokens as free and says workloads can later move into a data center or the cloud for production. Businesses should distinguish that token-level proposition from the broader cost of acquiring and operating infrastructure.
The placement options within Dell’s strategy serve different purposes:
| Offering | Intended role | Business consideration |
|---|---|---|
| AI Data Platform | Connect enterprise information with AI workloads through storage, security and data management | Whether applications can access governed data consistently |
| Deskside Agentic AI | Support local agent use rather than exclusive cloud execution | Whether in-house processing suits security and spending requirements |
| Azure Local | Enable hybrid deployment while retaining selected workloads on-premises | Which workloads should remain close to enterprise data |
| Dell Private Cloud | Provide a disaggregated environment for managing compute and storage | Whether independent resource scaling improves infrastructure economics |
Latency also shapes placement, alongside security, governance and sovereignty requirements. In healthcare, those requirements can influence where workloads are deployed. Whatever the sector, managers need to match a workload’s operating requirements to a platform that can meet them economically.
Kenny Lowe, technical staff member and cloud platforms evangelism and enablement lead at Dell, questions whether earlier deployment approaches still fit emerging demand. He advocates matching capacity and platform choice more closely to workloads, and describes a movement away from hyperconverged designs toward greater separation of infrastructure resources.
Disaggregated Infrastructure Changes the Scaling Equation
Dell’s disaggregated approach separates compute from storage so each can grow independently. That distinction gives enterprises another way to manage AI expansion: a workload requiring more processing need not automatically dictate an equivalent storage increase. Resource allocation becomes more granular.
Within Dell Private Cloud, Dell positions PowerStore Elite as providing storage performance and data reduction capabilities. Combined with the platform’s disaggregated architecture, the company says those capabilities give customers greater flexibility in allocating compute and storage.
Caitlin Gordon, senior vice president of product management for private cloud and AI solutions at Dell, cites a 6:1 data-reduction ratio for PowerStore within Dell Private Cloud and acquisition-cost reductions reaching 65%. She also attributes operating savings to the private-cloud software. These remain Dell’s claims; procurement teams should evaluate them against their own workload characteristics and purchasing assumptions.
Financing is another part of the company’s response to expansion. David Kennedy, executive vice president and chief financial officer at Dell, reported $64 billion in booked AI demand during the prior year, including $34 billion in Q4. He points to Dell’s financial-services business as a way to help fund infrastructure investment, while emphasizing due diligence and the durability of demand.
Those figures indicate the scale of appetite Kennedy describes, but booked demand does not establish customers’ eventual returns. Financing addresses access to capital. Enterprises still need to determine whether the resulting environment can deliver useful agent operations at a sustainable cost.
Dell’s strategy brings data readiness, autonomous operation and infrastructure economics into the same purchasing discussion. That alignment is timely because production deployments force enterprises to resolve all three together. The practical next step is to assess agent workloads against data access, operating controls and placement costs before committing to the infrastructure that will support their expansion.
