Dell Pushes On-Premises AI as Token Use Jumps 320-Fold
Updated
Updated · ZDNet · May 27
Dell Pushes On-Premises AI as Token Use Jumps 320-Fold
9 articles · Updated · ZDNet · May 27
Dell Technologies World 2026 centered on moving enterprise AI from cloud pilots into on-premises and hybrid production, arguing businesses need dedicated infrastructure to scale agents reliably.
320-fold growth in AI token usage and a projected 3,400% rise in global token consumption by 2030 were cited as key reasons to shift workloads off cloud-based LLMs and cut costs.
Dell tied that push to sovereignty and governance, unveiling the Dell AI Data Platform, Deskside Agentic AI and support for Nvidia OpenShell to keep data, policies and agent actions under tighter control.
One case study said agents exhausted a full-year token budget by March, underscoring Dell's broader message that companies must balance pressure to move fast on AI with slower rollout, security checks and tool maturity.
Is Dell's on-prem AI push simply trading cloud vendor lock-in for a new dependency on hardware ecosystems?
Are hidden operational costs making the on-prem AI savings promise a myth for most businesses?
As AI agents gain autonomy, can on-premise systems truly provide the control needed to prevent catastrophic business errors?
Dell’s On-Premises AI Revolution: $9B in Server Revenue Signals Shift from Cloud Amid Token Cost Crisis
Overview
Dell Technologies is making a major shift toward on-premises AI infrastructure in response to the rising demands and costs of modern AI workloads. As token consumption for agentic AI applications surges, public cloud solutions are becoming too expensive for many enterprises. Organizations now seek more cost-effective, secure, and controlled environments for their AI operations. To address this, Dell has rapidly expanded its portfolio with solutions like the Dell AI Factory with NVIDIA, Deskside Agentic AI, new PowerEdge servers, and PowerStore Elite storage arrays. These products are specifically engineered to handle the intensive computational and storage needs of advanced AI.