Enterprises Shift AI Workloads to Private Clouds Amid Cost and Security Concerns

A significant trend has emerged where enterprises are moving AI workloads from public clouds to private infrastructure. According to Broadcom's Private Cloud Outlook 2026 report, the percentage of enterprises using public cloud for AI inference has dropped from 56% to 41% in one year. The shift is driven by the need for better cost management, security, and control over data. Notably, 83% of enterprises are considering workload repatriation, with 50% having already moved some workloads back. The report highlights that cost has now overtaken security as the primary concern regarding public cloud usage. IT leaders report that 97% believe some public cloud spending is wasted, with over half indicating that this waste exceeds 25% of their total spending. The trend reflects a broader shift in cloud strategy as enterprises prioritize private cloud for high-security and data-intensive workloads.

Key Points: • 56% of enterprises used public cloud for AI inference last year; now only 41%. • 83% of enterprises are considering moving workloads back to private clouds. • Cost management has overtaken security as the top concern for public cloud usage.