Databricks as data-control plane

Databricks started with a speed problem. In 2009, Matei Zaharia’s Apache Spark project at the University of California, Berkeley turned slow “big data” jobs into something fast enough for machine learning and interactive analysis, and on April 8, 2026, the Association for Computing Machinery gave him its $250,000 ACM Prize in Computing for that line of work. (acm.org) (techcrunch.com) The prize was not just for one tool. The Association for Computing Machinery said Zaharia’s open-source stack includes Apache Spark for distributed computing, Delta Lake for making cloud data stores reliable, and Machine Learningflow, usually written as MLflow, for tracking machine learning experiments and deployments. (acm.org) That trio solves a very old corporate mess. One system stores raw data, another cleans it, another trains models, and another serves dashboards, so companies spend years moving the same information between boxes that do not agree with each other. (acm.org) (zdnet.com) Databricks is now selling itself as the place where those boxes get pulled into one control layer. TechCrunch reported on April 8 that the company has raised more than $20 billion, is valued at $134 billion, and has reached $5.4 billion in revenue while pushing from cloud data storage into what Zaharia called a data foundation for artificial intelligence and agents. (techcrunch.com) You can see that pitch in the customers it is chasing. On April 9, Tata Power said it would adopt the Databricks platform across all business clusters to build one enterprise-wide data and artificial intelligence system for grid management, power planning, billing, rooftop solar operations, and renewable forecasting. (tatapower.com) That is not a chatbot story first. Tata Power said the platform will combine edge data, operational data, and enterprise data on one governed system, which is the boring plumbing work every large utility needs before any artificial intelligence model can safely touch live operations. (tatapower.com) The shortcut Databricks is offering is natural-language access on top of that plumbing. Tata Power said it plans to use Databricks Genie so employees can ask questions in plain language and get answers, dashboards, and analytics from company data without digging through separate tools. (tatapower.com) That matches what other data teams are saying they want right now. ZDNET reported on April 9 that executives are using artificial intelligence to replace spreadsheet-based mapping and automate pipeline and integration work, with some reporting workload reductions of up to 40 percent. (zdnet.com) So the current Databricks bet is simple: if a company controls the layer where data is stored, cleaned, governed, queried, and handed to models, it becomes harder to rip out than a standalone model vendor. Zaharia’s award recognized the software that made that layer possible, and Tata Power’s deal shows how Databricks is trying to turn that academic infrastructure into the operating system for industrial artificial intelligence. (acm.org) (tatapower.com) (techcrunch.com)