Broadcom: Enterprises Shift AI to Private Cloud for Cost Control

According to Broadcom, 56% of enterprises are either currently using or planning to use private cloud for production AI inferencing. This marks a notable change, as public cloud utilization for these same workloads saw a year over year decline from 56% to 41%.

The findings suggest that after initial experimentation phases, companies are re evaluating their AI infrastructure strategies based on real world operational demands rather than the convenience that drew them to public cloud in the first place.

AI inferencing, the process of running trained models to generate live outputs, places new demands on IT infrastructure that traditional workloads simply don’t. These include repeated queries, large data volumes, and stringent requirements for latency, governance, and operating costs.

Broadcom’s data underscores that the move toward private cloud is a strategic response to these complex needs, not a rejection of cloud computing as a whole.

Cost and Data Protection Drive the Move

Enterprises are prioritizing several key factors when deploying AI workloads. Data protection and privacy emerged as the biggest new demand created by AI, identified by 37% of IT leaders. Security and control followed closely at 36%.

Prashanth Shenoy, Vice President of Marketing for the VCF Division at Broadcom, noted that the transition from AI pilots to production exposes organizations to higher infrastructure costs, security gaps, and increased operational complexity.

Cost has also become the leading concern for public cloud users, rising from 26% in 2025 to 31% in 2026, surpassing security as the top issue. A striking 97% of IT leaders believe a portion of their public cloud spending is wasted, with 52% estimating this waste accounts for over 25% of their total public cloud budget.

Repatriation Is Accelerating

This concern is fueling a surge in workload repatriation. 83% of enterprises are considering moving workloads from public to private cloud, and 50% have already done so. Security and compliance remain the primary drivers behind these moves.

Effective cloud cost management is becoming critical for businesses scaling their AI initiatives, particularly as the gap between expected and actual cloud spending continues to widen.

Geopolitics Now Shapes Infrastructure Decisions

Geopolitical considerations are increasingly influencing IT infrastructure decisions. Four out of five IT leaders report that geopolitics now impacts their IT strategy and operations.

Data sovereignty and residency requirements were cited by 54% of respondents as a leading geopolitical factor, followed by jurisdiction specific compliance at 51%. These issues are particularly relevant in highly regulated sectors like financial services, public sector, and healthcare, where strict security and regulatory requirements compound the pressure on infrastructure planning for AI workloads. Understanding data sovereignty is vital for global operations.

Asia Pacific Leads the Repatriation Trend

Sylvain Cazard, President of Asia Pacific, Japan and Middle East at Broadcom, highlighted regional trends, noting that 82% of organizations in the region are considering workload repatriation.

He views this as a turning point in AI infrastructure thinking, with enterprises now prioritizing security, cost predictability, performance, and operational control as AI scales beyond experimentation. The full Broadcom report details these regional insights.

Broadcom’s findings indicate that private cloud is solidifying its position as the preferred deployment option for production AI workloads. This trend is especially pronounced in scenarios where organizations require enhanced control over data, predictable spending, and robust compliance with evolving regulatory landscapes.

The shift signals a more mature approach to AI deployment, one that prioritizes operational realities over the initial ease of access that made public cloud attractive during early experimentation.