Microsoft’s internal Azure-based Managed Cloud Labs platform supports more than 20,000 labs and 150,000 VMs across customer support, engineering, testing, and release teams. Because business leads configure CPU, memory, and storage quota limits typically based on the largest environment used, labs were often provisioned for more capacity than needed, driving unnecessary excess cost. As the annual cost of Azure resources on Managed Cloud Labs climbed to eight figures, Microsoft Digital developed an AI optimization service that processes performance telemetry, user behavior, infrastructure configuration, and cost signals across the platform. The service learns how resources are used and can right size labs to match each workload’s needs without disrupting dependent teams. The challenge was that manually analyzing the vast data—including every VM, CPU, disk, performance tier, usage pattern, and cost signal—was impractical, as noted by principal product manager Nathan Prentice. The AI approach avoids blunt solutions like cutting resources or imposing blanket quotas, which would harm user experience, and instead connects real usage telemetry with infrastructure decisions to optimize storage performance, a primary cost driver.