Enterprise AI workloads are shifting toward private cloud, according to Broadcom‘s latest research. The company’s Private Cloud Outlook 2026 report, titled “The AI Tipping Point,” is based on a global survey of 1,800 senior IT decision-makers.
The report notes that production AI workloads are changing how organizations evaluate cloud architecture, cost, security, and governance. 56% of enterprises surveyed are running or planning to run production AI inferencing on private cloud.
Production AI workloads introduce sustained compute demand, sensitive data flows, governance requirements, and performance expectations that can expose limits in a purely public cloud approach. Public cloud use for the same workloads fell 15 percentage points year over year, from 56% to 41%. Public cloud remains part of enterprise IT strategy for experimentation, elastic capacity, and specialized services.
But AI production workloads require a more controlled environment. The shift is also reflected in repatriation data. 83% of enterprises are considering or have already repatriated workloads from public cloud to private cloud.
Related: AI’s role in specialty medicine grows
50% have already repatriated some workloads. In the past, the corresponding figures were lower. AI appeared as a repatriation category for the first time in the 2026 study. 43% of organizations repatriating workloads are moving AI training, large language models, and inference from public cloud to private cloud.
Broadcom said cost has overtaken security as the top public cloud concern in the 2026 study. According to the report, 31% of respondents cited cost management as a leading public cloud challenge, up from 26% in 2025.
97% of surveyed IT leaders believe some portion of their public cloud spend is wasted. 52% said that waste exceeds 25%. They linked those findings to AI infrastructure pressures, including compute, storage, bandwidth, GPU pricing, data movement fees, and unpredictable usage patterns.
As enterprises move AI workloads into production, they are looking for a more controlled and secure environment. Private cloud is emerging as the preferred deployment environment for AI inference among surveyed organizations, according to Broadcom‘s research.
Related: WashU Medicine AI institute bridges research fields
One reason for this shift is that private cloud offers better control over costs. With AI workloads requiring significant computational resources, the cost of running these workloads in the public cloud can be prohibitive.
Prashanth Shenoy, vice president of marketing for the VMware Cloud Foundation Division at Broadcom, noted that enterprise AI “has found its infrastructure home. And it is private cloud.”
The trend towards private cloud for AI workloads is likely to continue, driven by the need for better control over costs, security, and governance. They expect to see more enterprises moving their AI workloads to private cloud as the use of AI becomes more widespread.
