A global survey of 1,800 senior IT decision makers at organizations with more than 1.000 employees, published today, finds that well over half of respondents (56%) are deploying artificial intelligence (AI) workloads in private cloud computing environments, compared to 41% that are deploying them in public clouds

Conducted by Broadcom, the survey also finds that nearly three-quarters of respondents (72%) plan to increase spending on private cloud computing environments over the next three years, with 58% planning to make them a priority.

Overall, the survey finds 50% of respondents have also repatriated workloads from public to private cloud computing environments, with a third (33%) considering whether to migrate workloads in 2026.

Types of workloads that are candidates for migration are have high security or compliance requirements (51%), data-intensive (48%), business-critical (47%), AI (43%), productivity (40%), latency-sensitive (39%) and modern cloud-native applications (36%), the survey finds.

The more critical factors when evaluating workload placement are security and compliance: 32%, followed by data sovereignty or control (15%), performance and latency (14%), integration with existing systems and data (14%), cost predictability and financial control (12%) and speed of deployment and scalability (12%).

Despite the relatively small percentage that are concerned about data sovereignty, 80% of respondents said geopolitics and regulatory factors are affecting IT strategy and operations with data sovereignty and residency requirements (54%) and jurisdiction-specific compliance requirements (51%) cited as the top two issues.

In general, the survey finds 70% of enterprise IT organizations have either a central team or standardized policies, tooling, or automation to manage workload placement. A full 81% either fully outsource or rely on professional services for cloud-related needs.

Nevertheless, skills gaps are an issue. AI infrastructure and operations (40%), cloud and distributed security operations (38%), cloud-native and Kubernetes operations (37%), automation and infrastructure as code (35%), monitoring, observability, and performance (33%), data platform and data infrastructure (31%), cost and financial governance (31%) and design and architecture (27%) are all areas where there are skills shortages.

Additionally, nearly all respondents (97%) said they believe some of their public cloud spend is wasted, with more than half (52%) believing that waste exceeds 25%, the survey finds.

Prashanth Shenoy, vice president of marketing for the VMware Cloud Foundation Division at Broadcom, said the survey makes it clear that IT teams are struggling to effectively manage workloads in public cloud computing environments. Generally low utilization of cloud infrastructure resources is now being further exacerbated by AI workloads that don’t efficiently consume graphical processor unit (GPU) resources, he noted.

It’s not clear to what degree IT teams are investing in observability tools and platforms to better optimize consumption of IT infrastructure, but Broadcom is clearly betting that private clouds provide IT teams with more control over those resources, regardless of whether that private cloud is deployed in a hyperscaler or on-premises IT environment.

The challenge, as always, is that many applications today are first developed in a public cloud. The challenge then becomes overcoming the natural amount of inertia that builds up, which often results in workloads continuing to run where they are, versus migrating them to the IT platform that makes the most sense to actually deploy them.