PepsiCo, one of the world's largest food and beverage companies, is replacing a 27-year-old fragmented business intelligence estate by moving to a single, governed analytics and AI platform built on Azure Databricks SQL and Unity Catalog, with the early win being an 80% cost reduction on a key sales-analytics workload and roughly 50% faster processing across financial and commercial reporting served through Power BI and Tableau.
"Now, we have a clear direction and technology to support it. With open source data lakehouse, we can centralize all of our global data in one place. Adding in the benefits of a serverless infrastructure, we can now perform various AI analytics at scale super efficiently," said Joshua Lee, Lead Global Solution Architect for Data Analytics & AI at PepsiCo.
The platform was built because PepsiCo's existing BI landscape had grown through decades of acquisitions and regional data-warehouse deployments, leaving finance, sales, and field teams querying slightly different versions of the same numbers and slowing down the company's ability to roll out AI-driven analytics at global scale. Leadership selected Azure Databricks SQL because it offered a serverless SQL engine with the cost profile of a cloud data lake and the governance surface needed to retire legacy Synapse and Teradata estates. Unity Catalog was a deciding factor because it gives PepsiCo one place to manage cross-region access, lineage, and data-product ownership across the entire enterprise data foundation.
The technical architecture centralises data from multiple older warehouses and sector-specific data lakes into one Databricks lakehouse. DBSQL serverless handles the heavy analytics workloads that previously ran on provisioned warehouse capacity, with cost-down driven by the serverless pricing model and improved query plans. Core financial and commercial tables are built once in DBSQL and served downstream to Power BI and Tableau for business users. The team is now preparing to layer AI-driven analytics on top of the same governed warehouse, using the same data products that already serve traditional BI.
The operational impact is that a key sales-analytics workload that previously cost about $500K per year on provisioned warehouse infrastructure now runs on DBSQL serverless at about $175K per year — roughly 80% lower — with similar performance. Processing of core financial and commercial tables improved by around 50% end-to-end. PepsiCo is now methodically retiring Synapse and Teradata estates onto the unified lakehouse so that all enterprise reporting runs against a single governed source, with the same data foundation ready to host AI-driven analytics workloads as the company expands its use of generative and predictive AI.
Details