Nvidia pitches RTX Spark as the chip that finally makes local AI agents practical on Windows devices

Nvidia unveiled RTX Spark, a Windows-focused version of its DGX Spark hardware for laptops and compact desktops.

Built on Grace Blackwell, it pairs a Grace CPU and Blackwell RTX GPU with up to 128GB of unified memory and claims up to 1 petaflop FP4 performance.

The system is designed for local AI inference and agent workflows, relying on Nvidia’s CUDA stack and related software support.

Nvidia also showed Windows security and agent-management features, plus Adobe and OpenShell Runtime integrations for creative and developer workflows.

A larger DGX Station for Windows was also announced for enterprise-scale local model execution.