Microsoft Azure Confidential Clean Rooms: status, features and alternatives

Status on 9 November 2026: limited preview. Microsoft states: "Customers should not use the preview to process personal data or other data that is subject to legal or regulatory compliance requirements" (Microsoft Learn). For live advertising data about people, it is not an option yet.

Azure Confidential Clean Rooms gives several organisations a shared, protected place to analyse their combined data. The service hides each party's raw data from the other parties and from Microsoft as the Azure operator (Microsoft Learn). Microsoft lists advertising among its target scenarios. Examples include matching an advertiser's CRM data with publisher data for targeting, and sharing audience data with measurement partners for attribution.

Microsoft also aims it at regulated sectors. Its list includes banks and insurers selling to each other's customers, banks working with retailers on offers, government departments, and hospitals studying patient outcomes. All of those use personal data, which the preview terms forbid today. The scenarios describe where the product is going, not what customers can do now.

Only Spark SQL analytics is in the preview. Teams that need machine learning training, fine-tuning or inference are asked to apply through the same form.

The engine is Apache Spark SQL, run in a confidential version. The Spark driver and executors are "fully attested policy-governed enclaves". They run as virtual nodes on Confidential Azure Container Instances inside an Azure Kubernetes Service cluster (Microsoft Learn).

Governance runs on the Confidential Consortium Framework, an open-source ledger technology. It manages who belongs to a clean room, checks that the right parties approved each query and consented to data access, and keeps a tamper-resistant audit trail.

The privacy controls are explicit:

The query author sets the minimums, and the other parties review them before approving. Microsoft publishes its clean room container images and sidecars, with source code in the Azure/azure-cleanroom repository (GitHub). Their provenance can be checked through GitHub artifact attestation.

Microsoft publishes no price for Azure Confidential Clean Rooms. The service is in a limited preview governed by the Supplemental Terms of Use for Microsoft Azure Previews, and access depends on an accepted application (Microsoft Learn). Pricing is therefore n/d until general availability.

Strengths

• Operator excluded. Enclaves and attestation keep Microsoft itself out of the data path, which is a stronger promise than access policy alone.
• Clear output rules. Column allow-lists and minimum counts before and after the query are written into the query definition, where all parties can review them.
• Open code. Container images and sidecars are open source, so a security team can inspect what runs.
• Multi-party from the start. Several data providers can share one clean room.

Limits

• Not production-ready. The preview is "intended for testing, evaluation, and feedback".
• No personal data allowed. That rules out most advertising use until GA.
• Spark SQL only. ML training and inference are not part of the preview.
• Gated access. Microsoft "might not be able to accept all requests".

• When the preview first opened. The documentation we read gives page dates, not a launch date.
• Any date for general availability, any price, or any named customer.
• The Azure regions where the preview runs.
• Whether "Azure Data Clean Room" was ever an official Microsoft product name. We found no such product; the backlog title appears to be informal.

1. Microsoft Learn, "Perform protected multiparty data collaboration on Azure" (Azure Confidential Clean Rooms; page dated 22 May 2026, updated 17 June 2026; accessed 9 November 2026). https://learn.microsoft.com/en-us/azure/confidential-computing/confidential-clean-rooms
2. GitHub, "Azure/azure-cleanroom" repository (MIT licence; accessed 9 November 2026). https://github.com/Azure/azure-cleanroom
3. LiveRamp documentation, "Configure Clean Rooms: Clean Room Types" (accessed 9 November 2026). https://docs.liveramp.com/clean-room/en/configure-clean-rooms.html
4. US Securities and Exchange Commission, EDGAR submissions record for Microsoft Corporation (accessed 9 November 2026). https://data.sec.gov/submissions/CIK0000789019.json