Databricks (AI Runtime training)

Active, privately held. Databricks states its managed Foundation Model Fine-tuning product has reached end of life; it recommends AI Runtime instead.

Sources 1, 2, 3, 4 · checked 2026-11-08

• AI Runtime: serverless GPU compute (on-demand A10s and H100s) for training and deep learning
• Run your own training code, including fine-tuning workloads, on serverless GPUs
• Submit distributed GPU training jobs with YAML job configs; Ray Train supported
• Stream training data from Unity Catalog volumes
• Deploy models trained on AI Runtime to Databricks Model Serving

Sources 1, 2, 3, 4 · checked 2026-11-08

• Teams whose training data already lives in Databricks and Unity Catalog
• ML teams comfortable writing their own fine-tuning code

Sources 1, 2, 3, 4 · checked 2026-11-08

Contact vendor (usage-based Databricks pricing; no fine-tuning-specific rate found)

Sources 1, 2, 3, 4 · checked 2026-11-08

• SOC 2 Type II (report via account team)
• ISO 27001:2022

Listed as stated by the vendor. Scope can differ by product or subsidiary; request current reports.

Sources 1, 2, 3, 4 · checked 2026-11-08

• Managed Foundation Model Fine-tuning (formerly Mosaic AI Model Training) is end of life; databricks_genai package and UI no longer available
• AI Runtime is general GPU compute: you supply the fine-tuning code
• Certification scope varies by cloud, region, and compliance profile

Sources 1, 2, 3, 4 · checked 2026-11-08