Broadcom: Enterprises move AI workloads to private cloud as costs rise

Broadcom found that 56% of enterprises use, or plan to use, private cloud for production AI inferencing, while public cloud use fell to 41%.

Enterprise AI workloads are moving further into private cloud environments as organizations place more focus on cost control, data protection, and infrastructure governance, according to Broadcom's Private Cloud Outlook 2026 report.

Broadcom said the findings point to a change in how enterprises are deploying AI after initial experimentation. Public cloud remains part of AI development and training activity, but production inferencing is placing different demands on infrastructure.

The report found that 56% of enterprises are running, or plan to run, production AI inferencing on private cloud. Public cloud use for the same workloads fell from 56% to 41% year over year.

AI inferencing refers to the process of running trained models to generate outputs in live environments. These workloads can involve repeated queries, large data volumes, and tighter requirements around latency, governance, and operating cost.

According to the report, 37% of IT leaders identified data protection and privacy as the biggest new demand created by AI. Security and control followed at 36%.

Prashanth Shenoy, vice president of marketing for the VMware Cloud Foundation Division at Broadcom, said enterprises moving from pilots to production AI are facing higher infrastructure and operational costs, security gaps, and added complexity. He said the research showed a growing preference for private cloud in production AI deployments.

“Across Asia Pacific and Japan, organizations are moving beyond AI experimentation and focusing on production deployment. As AI workloads scale, enterprises are prioritizing security, cost predictability, performance and operational control, factors that are accelerating the shift towards private cloud. With 82% organizations in Asia Pacific and Japan considering workload repatriation to private cloud, this marks a turning point in how organizations are thinking about AI infrastructure,” explained Sylvain Cazard, President, Asia Pacific, Japan & Middle East, Broadcom.

Cost becomes the main public cloud concern

Cost has also become the leading public cloud concern in the report. It rose from 26% in 2025 to 31% in 2026, overtaking security as the top issue cited by IT leaders.

Cloud waste remains a common concern among respondents. Broadcom said 97% of IT leaders believe part of their public cloud spending is wasted, while 52% estimate that waste accounts for more than 25% of their total public cloud budget.

The report also found that workload repatriation is under review at many organizations. Eighty-three percent of enterprises are considering moving workloads from public cloud to private cloud, and 50% have already repatriated at least some workloads.

Security and compliance remain the top reason for repatriation, cited by 51% of organizations. Cost predictability and performance were each cited by 39%.

Private cloud spending plans are also rising. Broadcom said private cloud spend intent increased by 21 points over a three-year outlook, compared with a 10-point increase for public cloud.

Building new workloads on private cloud was named a top priority by 58% of IT leaders, up from 53% a year earlier. The report links this to the operational requirements of AI workloads, including cost predictability, data protection, and infrastructure control.

Data sovereignty shapes infrastructure decisions

Geopolitical considerations are also affecting infrastructure decisions. Four out of five IT leaders said geopolitics now influences their IT strategy and operations.

Data sovereignty and residency requirements were cited by 54% of respondents, making them the leading geopolitical factor shaping infrastructure decisions. Jurisdiction-specific compliance followed at 51%.

The report said these issues are especially relevant in sectors with strict security and regulatory requirements, including financial services, public sector, healthcare, and life sciences. In these industries, AI-related data volumes and cross-border data governance requirements are adding pressure to infrastructure planning.

Broadcom's findings show that private cloud is becoming a more common deployment option for production AI workloads, particularly where organizations need tighter control over data, spending, and compliance obligations.