Cloudera survey finds enterprises rethinking data architecture for AI

Cloudera found that 72% of organizations say their data architecture needs major changes to support future AI requirements.

Enterprise AI deployments are changing how organizations manage data infrastructure, with companies reporting higher costs, governance challenges, and changes to where workloads are run, according to a global survey from Cloudera.

The survey, which covered 1,500 enterprise architects, cloud infrastructure leads, and data architects, found that 77% of organizations are actively using AI. At the same time, 95% said they had delayed or canceled at least one AI initiative during the past year because of data governance, compliance, or regulatory issues.

Those constraints are also influencing infrastructure plans. Globally, 72% of respondents said their current data architecture requires substantial changes to support future AI requirements, compared with 68% in Asia Pacific.

AI deployments are already affecting how organizations store and manage data. Three-quarters of respondents said AI integrations had changed their data storage and architecture practices, while 84% reported higher infrastructure costs linked to AI workloads.

Cloudera CTO Sergio Gago said existing architectures were often designed around traditional analytics rather than current AI workloads.

"Many enterprises are discovering that the architectures built for traditional analytics weren't designed for the scale, governance, and flexibility AI demands today," Gago said. "Success will depend on building a data foundation that gives organizations the freedom to run AI wherever it makes the most sense, without compromising control or security."

Governance and workload placement reshape AI infrastructure

Governance issues are also delaying AI deployments. Cloudera's earlier Data Readiness Index found that 75% of organizations said AI was exposing limitations in their existing governance processes.

In the latest survey, 73% of global respondents said AI had made data governance more complex. The figure stood at 61% among respondents in Asia Pacific.

More than half of organizations globally, or 55%, said they had delayed or canceled more than six AI projects during the previous 12 months because of governance, compliance, or regulatory requirements. In Asia Pacific, 92% had delayed or canceled at least one project for those reasons.

Data security, governance, and compliance were also the most commonly cited reasons for changing AI infrastructure in Asia Pacific, selected by 46% of respondents.

The survey also found that 97% of respondents move data between environments at least once a month, including across public cloud, private cloud, on-premises systems, and edge infrastructure. Moving data across multiple environments adds another consideration for organizations applying governance and compliance controls to AI workloads.

Infrastructure decisions are also extending to where individual AI workloads are run. Globally, 66% of respondents said their organizations had moved AI workloads from public cloud platforms back to private cloud or on-premises infrastructure during the past year.

The figure was 64% in Asia Pacific. Respondents also reported plans to invest across cloud, on-premises, edge, and hybrid infrastructure rather than concentrate AI workloads in a single environment.

One-quarter of respondents globally said they expect to prioritize a hybrid-first architecture over the next two years.

In Singapore, 43% of respondents said their organizations had moved AI workloads out of the public cloud, while another 36% were evaluating such a move. About 35% of local respondents also said they expect to increase spending on edge infrastructure during the next two years.

Data security, governance, and compliance remain part of those infrastructure decisions. In Singapore, 41% of respondents identified these requirements as the main reason for changing their AI infrastructure.

Remus Lim, Cloudera's senior vice president for Asia Pacific and Japan, said organizations in the region are assessing infrastructure according to the requirements of individual workloads.

"Singapore reflects this evolution as enterprises collectively evaluate latency, performance, expenditure, and governance when defining their AI deployment strategies," Lim said.