Databricks Clean Rooms: GCP, Multi-Party Collaboration & Self-run of Notebooks
Summary
- Multi-Cloud Clean Rooms: Databricks Clean Rooms now run on GCP, AWS, and Azure, enabling cross-cloud partnerships in a single secure environment.
- Multi-Party Collaboration: Up to 10 organizations can work together in one clean room using Delta Sharing for live data access without replication.
- Notebook Self-Service: Partners can upload and run their own privacy-safe notebooks with explicit approvals, enabling joint analysis with governance.
Databricks Clean Rooms: GCP, Multi-Party Collaboration & Self-run of Notebooks

This demo shows how cross-cloud partners (on GCP, AWS, and Azure) can collaborate in a single shared clean room. It highlights multi-party collaboration, seamless data sharing via Delta Sharing, and the ability for partners to upload and run their own notebooks, with explicit approvals. The result: secure, auditable joint analysis across clouds and organizations, with no raw-data exposure.
Chapters
0:00Introduction to new features0:54Demo scenario setup1:33Creating the clean room2:13Adding data to the environment2:58Authoring notebook analysis3:48Approval and collaboration4:39Running analysis and results
FAQs
Can partners in different clouds use the same clean room?
Yes. This video demonstrates a brand on GCP, a data provider on AWS, and a publisher on Azure all collaborating in a single clean room to identify customer segments for advertising activation.
How many organizations can participate in a clean room?
The multi-party feature supports up to 10 total organizations working together in one central environment, as shown in this video's demo.
What capabilities do Databricks Clean Rooms provide for cross-company collaboration?
Databricks Clean Rooms enable multi-party collaboration across clouds with up to ten organizations participating simultaneously. Partners upload their own analysis code directly into the clean room without exposing raw data. Metadata sharing and query-based data access patterns ensure organizations maintain control while enabling joint insights on the Databricks Data and AI platform.
Can partners upload their own analysis code?
Yes. This video showcases the new ability for clean room collaborators to author their own privacy-safe notebooks and run them with explicit approvals from other partners.
Full transcript
0:00 Introduction to new features
Hi everybody, I'm DJ Shy, tech lead for data bricks clean rooms and today I'll be giving a demo of clean rooms highlighting some of our recently launched features. In this demo I'll specifically be highlighting the launch of clean rooms on GCP. This launch furthers data brick support for a truly differentiated multicloud offering. Now customers on GCP can create data bricks clean rooms and securely collaborate with partners across platforms such as AWS and Azure.
I'll also be showcasing support for multi-party clean rooms. This recent feature launch allows up to 10 total organizations to work together in one central environment. Lastly, I'll be showcasing the new ability to run your own notebooks by explicit approvals. With this feature, clean room collaborators can author their own privacy safe notebooks to run in the clean room. To help bring these features to life, I'll be showing a brand, a data provider, and a publisher working together to identify the best possible segment for the brand to activate.
0:54 Demo scenario setup
As you can see, the brand is in GCP. They have transactions level data representing who has purchased things from their e-commerce store and what has been purchased. The data provider is on AWS. They have demographic and psychoraphic data about the entire US census. Lastly, the publisher is on Azure. The publisher has exposure and impressions data such as who has seen or interacted with the brand's various advertising campaigns.
1:33 Creating the clean room
To start the demo, I'll be acting as a brand creating the clean room. First, I give the clean room a name. Then I add the data provider and publishers sharing identifiers which they have shared with me offline. I click create clean room. Soon the clean room is created and the data provider and publisher can both see it from within their accounts. Here I switch into the data provider account and I can see the clean room.
The same is also true for the publishers account which I'm now in. Now that the clean room is created, each organization can add data to it. Here you can see me operating as the brand and I'm adding transactions data to the clean room. I now switch tabs to the data providers account. I can see the data that's been added, but since this is a clean room, I can't actually query it. All I can see is m relevant metadata. The same is also true for the publisher.
2:13 Adding data to the environment
To keep this demo quick, the data provider and publisher add their data behind the scenes. Back in the brand's account, it's time for the magic to happen. The brand wants to combine their transactions data with the data provider census data and information from the publisher on which users saw ads from their past campaigns. Combining all this info together will let the brand figure out which common characteristics of users that are most likely to convert into a transaction after seeing their ads and thus which segment they should target next.
2:58 Authoring notebook analysis
The brand has written a notebook to do this analysis by joining their transactions data with the data provider census data and the publishers impressions data. The brand uploads this notebook to the clean room and sets themsself as the runner. By setting themselves as the runner, they're effectively asking the data provider and publishers permission to run the notebook on the data that's been added into the clean room.
Note, when the brand tries to run the notebook, they actually see a message saying that all collaborators must approve before the notebook can be ran. This is crucial to the clean room's privacy guarantees. In a data bricks clean room, only fully approved notebooks can be ran. This gives each collaborator an opportunity to define exactly what is allowed on their data. I've now switched to the data provider account. A user in the data provider account can now review the notebook and make sure all of the expected privacy controls are respected.
3:48 Approval and collaboration
They can then approve the notebook. Behind the scenes, I've done the same in the publishers account. Back in the brand's account, I can now see that the notebook is fully approved and the brand can run it. A few seconds later, the brand can see the results of their analysis. Thanks everyone for tuning in and please reach out if you have any questions about data bricks clean rooms.
Learn more about the Databricks Data and AI platform.
The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.