Dell Bets on On-Prem as the New Frontier for Corporate AI
In Las Vegas, the manufacturer announced the rack-scale PowerRack platform and expanded support for models from Google, OpenAI, Palantir, and SpaceX running locally. AI Factory with NVIDIA reaches 5,000 clients.
Dell opened its annual conference in Las Vegas on 18 May by positioning on-premises as a strategic destination for corporate AI workloads. This move addresses a growing frustration among hyperscaler customers: unpredictable costs in inference, regulatory constraints on data sovereignty, and the perception that open models can be served with significant savings in proprietary hardware.
PowerRack and the Rack-Scale Bet
The flagship of the announcements was the PowerRack, a platform that integrates computing, networking, storage, and cooling into a single scalable system. The aim is to eliminate the complexity of integrating individual components, which has historically delayed AI deployments in corporate data centres. The product directly competes with pre-configured solutions from NVIDIA and equivalent offerings from hyperscalers.
The company also presented the Dell Deskside Agentic AI, aimed at companies that need to run AI agents locally without exposing sensitive data to external environments. The focus is clear: regulated sectors, government, and companies with critical intellectual property.
Expansion to Google, OpenAI, Palantir, and SpaceX
Dell has expanded its on-premises deployment programme to include models from Google, OpenAI, Palantir, and SpaceX. This move consolidates the thesis that the cloud-first strategy for generative AI, dominant in 2024 and 2025, is being recalibrated by a significant segment of the corporate market.
The AI Factory with NVIDIA, launched in 2024, gained 1,000 clients in the last quarter and reached a total of 5,000 deployments. Eli Lilly, Honeywell, and Samsung are among the adopters cited by the company. For IT consultancies, the expansion of the installed base reshapes the services pipeline: implementation, integration with proprietary models, and operation of hybrid environments become concrete opportunities, with reduced dependency on cloud-native skills.
However, the most relevant takeaway from the event lies in what was not said. Dell did not announce a specific partnership with hyperscalers for the PowerRack, signalling that it views on-prem as an autonomous market rather than a complement to the cloud. For clients still defining architecture, the question shifts from which cloud to use to where to process each type of workload.