Databricks has published a walkthrough showing how financial institutions can join on-premises portfolio holdings with FactSet data in the cloud, without copying the holdings into Databricks. The same combined data can feed dashboards and natural-language questions through its Genie tool, though the example is a company demonstration, not an independent assessment of performance or security.
Databricks Watch analysis
What happened
The walkthrough connects an on-premises SQL Server database to Databricks Lakehouse Federation, where its tables appear in Unity Catalog as a foreign catalogue. It then adds FactSet financial data from the Databricks Marketplace as a shared catalogue. Databricks says queries run against the on-premises source in real time, with no data copying or persisted intermediate tables in its example.
A single SQL query joins the two sources for an investment dashboard covering holdings, earnings, risk and recommendations. Databricks also wires the same data into Genie, which translates natural-language questions into queries. The company’s Lakehouse Federation walkthrough provides the example setup.
Why it matters
For organisations with sensitive records kept behind a corporate firewall, joining them to cloud-based data without first building an ingestion pipeline could simplify some analytics workflows. The practical proposition is that the data stays at its source while teams query it alongside a subscribed dataset, rather than creating another copy to maintain.
That is not the same as proving every query is fast, every recommendation sound, or every governance concern solved. Databricks presents a worked example, and the real-world experience will depend on the source systems, permissions and query workloads involved.
Our read
The useful change here is the combination: federated first-party data, a cloud data share and a natural-language interface in one analytics workflow. It is a credible option to investigate if your organisation already uses Databricks and needs to combine data without building another pipeline. But keep the boundary clear: Genie can make asking easier; it does not make the underlying data or recommendations automatically trustworthy.
What to watch
- Whether Databricks publishes measured performance for queries against live external sources.
- How teams manage permissions and audit access across federated and shared catalogues.
- Whether the approach works as smoothly with other databases and real production workloads.
Discussion spark: Would you let staff query sensitive on-premises records through a natural-language interface, or keep that data behind SQL-only controls?
Sources and evidence
- Unlocking Investment Insights Without Migrating Your Data: Lakehouse Federation (27 October 2026, 18:00 UTC)
not affiliated with or endorsed by Databricks