Databricks and OpenAI have announced a $100 million partnership to make OpenAI’s latest models available inside the Databricks Data Intelligence Platform. The agreement means more than 20,000 Databricks customers will be able to run models such as GPT-5 directly on their data.
The deal allows companies to use GPT-5 within SQL or as an API, without extra setup or the need to move data. It also gives them access to Agent Bricks, Databricks’ flagship tool for building AI agents. These agents can analyse enterprise data, automate complex tasks, and support teams with reliable insights.
Databricks co-founder and chief executive Ali Ghodsi said enterprise demand for AI has been increasing. He explained that the partnership makes it easier for businesses to use OpenAI’s models securely at scale, while meeting strict governance standards. OpenAI’s chief operating officer Brad Lightcap added that bringing advanced models into Databricks makes deployment simpler without compromising on quality or security.
How Will Enterprises Use It?
Agent Bricks is designed to help businesses create production-ready AI agents that can be adjusted to specific needs. The system measures model accuracy with task specific evaluations and then optimises results for higher quality outputs. This continuous cycle allows the agents to adapt to domain specific work.
Examples of possible uses include reasoning agents for healthcare, risk analysis and contract review. Productivity agents may help with summarisation, customer support or knowledge management. Coding agents can debug complex systems, modernise legacy applications and generate new code.
OpenAI said GPT-5 delivers stronger instruction following and can handle multi-step requests with more reliability than previous versions. This includes adapting to changing contexts and coordinating across different tools. Databricks builds on this by ensuring all access is secure and monitored through its governance framework.
Who Is Backing The Partnership?
Large companies are already showing interest. Mastercard’s chief AI and data officer Greg Ulrich said the firm is using AI agents to strengthen commerce, personalise services and automate internal systems. He added that quality, scale and trust are the three elements they look for when working with agents, and this partnership helps meet those needs.
Experian has also welcomed the move. Its vice president of engineering Connie Cheung said the agreement enables them to build and deploy trusted applications with speed and security. She added that the models can now be integrated with their data on Databricks, allowing for more confident deployment.
The partnership builds on earlier cooperation between the two companies. OpenAI already used Databricks for processing AI data to improve ChatGPT. Databricks was also one of the first firms to host OpenAI’s open-weight model gpt-oss.
More from News
- Snapchat Through The Years: How The Social Media App Became What It Is Today
- Meta Will Be Using Your AI Chatbot History To Target Ads
- What Is The Government’s Digital Hospital Plan?
- You Can Now Make Direct Purchases Through ChatGPT
- Are AI Chatbots Being Tricked Into Giving Away Dangerous Information?
- China Launches New K Visa To Attract Global Tech Talent
- Why 8,000 Entrepreneurs Traveled to Malaysia for QNET’s 2025 Convention
- Gen Z Is The Most Digitally Literate, But Falls For Scams The Most. Here’s Why
What Technical Features Are Included?
The deal makes it that customers have access to dedicated high-capacity processing across OpenAI’s latest models. This means large volumes of enterprise data can be analysed without bottlenecks. Unity Catalog, Databricks’ governance system, brings compliance and ethical standards into the process, covering security, access as well as audit requirements.
Databricks’ Mosaic AI Gateway enforces access controls and monitoring for every request. This gives enterprises confidence that GPT-5 can be used at scale without compromising reliability. The models can be called directly through SQL, allowing tasks such as ticket classification or document summarisation to be embedded into existing workflows.
Agent Bricks adds an evaluation loop, measuring the accuracy of models on specific workloads and then tuning them for higher quality. This means production-ready agents are built with performance standards already tested.
What Will This Partnership Bring?
The collaboration is set to expand through ongoing engineering work between both firms. Teams from Databricks and OpenAI will co-develop feedback loops, model improvements and platform optimisations. The goal is to keep models like GPT-5 performing at their best for enterprise use.
A public preview is opening soon, with access available through account teams. Customers will also be able to try GPT-5 in the AI Playground, which lets users test features quickly. OpenAI models such as GPT-5 mini and GPT-5 nano will follow.
To mark the partnership, Sam Altman, chief executive of OpenAI and Ali Ghodsi of Databricks will host a virtual event called “The Future of AI: Build Agents That Work” on 11 and 12 November.
The $100 million deal is bringing a deeper integration of AI into everyday business systems, with Databricks positioning itself as a secure home for enterprise-ready AI agents built on OpenAI’s latest technology.
