Databricks made AI Runtime generally available, bringing fully managed GPU infrastructure directly into its Data + AI Platform for training and fine-tuning foundation models on enterprise data.

The company also expanded its collaboration with NVIDIA, adding support for NVIDIA NeMo, NGC containers and Nemotron models on serverless NVIDIA GPUs. The integration allows enterprises to customize open models while keeping training data governed through Unity Catalog and avoiding the need to manage underlying GPU infrastructure.

Source: databricks.com