I have known Sonya Vargas for some time. She built the analyst relations program at Databricks. This year was the first time I went to Data + AI Summit. The summit was huge. The expo floor was packed with partners who sell around Databricks.

Ali Ghodsi, the chief executive, spent a long stretch with the analysts. It was an open question session, close to the kind of room where nothing is on slides. His style came through in how he stayed in the conversation. The founding team is still on the board and still on stage. That is unusual once a company reaches this size. The spirit I heard from that group is simple enough to write down. Keep Databricks open. Obsess over the customer experience so people do not have to keep a second set of tools, or a second way to write the work, just to finish a job.

On the flight home I sat next to a technology leader from a finance company that processes billions of transactions. She said her days got easier after the move to Databricks because the platform could hold volumes her old stack dropped. The work she described still ends in a sheet.

I have hit that wall on marketing and finance extracts. Microsoft publishes the Excel worksheet limit as 1,048,576 rows by 16,384 columns. Load a large export and the file stops. Power Query can keep more in memory. The workbook you send a colleague is still the grid. Teams split the file or they leave the warehouse tool the business group will not open.

Finance teams can already pull Databricks data into Excel. Databricks ships Unity Catalog into the Excel ribbon. Google Sheets still gets the export. Snowflake customers can keep a Row Zero workbook on that warehouse, and Row Zero already hooked Cortex Analyst the way it hooked Genie. The purchase puts the grid inside Genie so the coworker and the sheet stop living in two files (Row Zero, 2026).

Databricks finance finished Genie work in Row Zero

Genie is the coworker Databricks built for business teams. A user can ask why margins moved or ask for a note on the sales pipeline. Ghodsi wrote on LinkedIn that large parts of Databricks began running on Genie at the start of 2026. Finance did its work there, then opened Row Zero, a cloud spreadsheet that sits on live warehouse tables.

They were combining Genie with a Live Cloud Spreadsheet called Row Zero.

That pairing is the deal. Databricks met Breck Fresen and Nick End, former Amazon Web Services and Tableau engineers who founded Row Zero in 2021, and agreed to buy the company on September 24. Terms were not disclosed. TechCrunch, citing PitchBook, put a May 2025 seed at $10 million and an estimated $40 million valuation at the time (TechCrunch, 2026). Patrick Wendell, a Databricks co-founder and vice president of engineering, said a spreadsheet inside Genie is required once finance, operations, and go-to-market teams live in the coworker (Databricks, 2026).

Exported workbooks are the copies agents will read

Teams still export a table, paste it into Excel or Google Sheets, and pass the file around. The copy has no live refresh and no clean record of who changed a cell. Databricks calls those files one of the largest sets of ungoverned data pipelines in the enterprise, and says agents cannot be allowed to work on sensitive data of unknown lineage (Databricks, 2026).

Row Zero keeps the formulas, pivots, and keyboard shortcuts people know. The workbook connects to live sources, follows each user's permissions, can refresh, can lock exports, and can write results back. Databricks says every action is auditable, including work an agent does through the same grid a person can read. The product will sit inside Genie on the web, desktop, and mobile, on Genie Ontology, Unity Catalog, and Unity Gateway. Support for sources outside Databricks continues, including Snowflake, Amazon Redshift, and Google BigQuery. In June, Row Zero shipped an integration that let a workbook call Genie and drop the answer onto a live table.

Row Zero says a workbook can work on billions of rows at interactive speed. Treat that as the vendor's claim and as the reason Databricks finance noticed the tool.

If you already use the tool, Databricks wants it inside the platform

In June I wrote that Databricks spent the summit crossing categories analysts still keep in separate binders. Row Zero is the same motion in a form finance already knows. Ghodsi keeps repeating open source, open integration, and lower cost. The purchase is the customer version of those words. You already model in a spreadsheet. Databricks would rather make that grid convenient inside its system than send you back to an export.

The woman on the plane wanted the volume to stay in one system. Databricks finance wanted the coworker and the sheet in one system. The purchase tries to serve both.

About Shashi Bellamkonda

Loading bio…

LinkedIn Profile

Get new posts by email:

Subscribe

Powered by