Databricks and OpenAI have formed a multi-year partnership to make OpenAI’s models, including GPT-5, natively available on the Databricks Data Intelligence Platform and its Agent Bricks product for building AI agents, the companies said on Saturday (Sep 28).

The collaboration, described by the firms as a US$100 million agreement, will provide OpenAI models to more than 20,000 Databricks customers across cloud providers. Databricks said GPT-5 will be a flagship model on its platform, enabling organisations to develop, evaluate and scale production-grade AI applications and agents on governed enterprise data.

Enterprise AI agents on governed data

Databricks said businesses need AI agents that can analyse and act on enterprise data while meeting standards for accuracy, security and governance. By offering OpenAI models directly within the Data Intelligence Platform and Agent Bricks, customers can build agents without moving data off the platform, with access via SQL or API and “high-capacity” processing for workloads.

Use cases cited include earlier disease detection, energy-efficiency analysis, fraud detection and faster application development.

“This partnership makes it easier for enterprises to securely leverage their data and OpenAI models at scale with best-in-class governance and performance,” said Ali Ghodsi, co-founder and CEO of Databricks.

OpenAI COO Brad Lightcap said enterprise demand for frontier AI is accelerating. “With Databricks, we’re making its deployment even simpler … bringing our most advanced models to where secure enterprise data already lives.”

Mastercard is among customers endorsing the approach. “For any use case, AI agents come down to three things: quality, scale and trust,” said Greg Ulrich, the firm’s chief AI and data officer, adding that the partnership allows Mastercard to build “trusted AI agents” on Databricks with the latest OpenAI models.

Governance, observability and evaluation

Databricks said Agent Bricks measures model accuracy with task-specific evaluation and LLM judges, and can tune models such as GPT-5 and gpt-oss for domain-specific outputs. Unity Catalog provides end-to-end governance and responsible-AI controls, while built-in observability is designed to help teams move agents into production with confidence.

The companies also plan collaborative optimisation work between their technical teams to improve model performance for enterprise use cases.

Databricks noted that OpenAI already uses Databricks for AI-data processing to generate product insights, and that Databricks was among the first to host gpt-oss, OpenAI’s open-weight model.

The firms will host “The Future of AI: Build Agents That Work” on Nov 11 (AMER) and Nov 12 (APJ/EMEA), featuring Databricks’ Ali Ghodsi and OpenAI’s Sam Altman, to discuss how enterprises can deploy AI agents at scale.