Dell and CoreWeave push AI reliability beyond the rack

Dell Technologies and CoreWeave are focusing on AI infrastructure reliability by aligning compute, networking, power and cooling across engineering, manufacturing and operations. Dell’s work with CoreWeave has helped adapt rack-scale systems to varied facility requirements, including different cooling layouts and power whip sizes, while supporting modular infrastructure that can be configured more quickly for each data center.

The collaboration starts before systems reach production. CoreWeave works with Dell on firmware settings, mechanical changes and deployment needs, while Dell uses operational feedback to improve validation and diagnostics across server and rack assembly. Dell has shifted complex hardware-failure diagnostics further left in the manufacturing process to reduce the risk of expensive fixes later in deployment.

Liquid cooling is adding new demands. CoreWeave’s Racky rack manager combines power, cooling and environmental telemetry in a single control interface, and Dell integrates Racky in its L11 factories so systems are tested before shipment. The companies have also worked on stronger leak detection, which Dell described as a critical reliability concern.

For Nvidia’s Vera Rubin systems, the operating model is expanding beyond individual racks to rows, data halls and full data centers. CoreWeave said reliability now depends on tight systems integration across compute trays, switches, data processing units, network fabrics, facility power and liquid cooling.