95% of Firms Have Delayed AI Projects on Governance Fears — Singapore Is Ahead on the Hybrid Fix

Enterprise AI has hit a governance bottleneck, and it is showing up in project pipelines worldwide. Cloudera's survey of 1,500 enterprise architects, cloud infrastructure leads and data architects found that 95% of organisations have delayed or cancelled AI projects because of infrastructure and governance constraints — even as 77% report they are actively using AI. The gap between running a promising pilot and running AI across the business is where most momentum stalls.

The numbers explain why. Nearly three quarters of respondents — 73% — said AI has made data governance more complex, while 72% admitted their current data architecture needs significant change to meet future AI requirements. Data movement is compounding the problem: 97% of organisations shift data between environments at least once a month, which makes consistent controls harder to apply. And the bills are rising — 84% reported higher infrastructure costs tied to AI workloads. Unsurprisingly, two thirds of firms globally say they have moved AI workloads out of the public cloud and back toward private cloud or on-premises infrastructure over the past year.

Singapore is emerging as an early mover on that shift. The survey, which covered nine markets including Singapore, found 43% of local organisations have already moved AI workloads out of the public cloud, with a further 36% evaluating the same move. Another 35% expect to increase spending on edge infrastructure over the next two years, and 41% cite data security, governance and compliance as the main reasons for changing their AI infrastructure. Cloudera's Asia Pacific and Japan head Remus Lim describes the regional conversation as moving from selecting an AI host to deciding which environment delivers the best business results for each workload.

The findings sit neatly alongside Singapore's own governance push — the IMDA's updated agentic AI framework, the AI governance guidance now reaching company boards. They suggest governance is not just a compliance overhead but the deciding factor in where AI actually runs: firms that cannot control their data across environments are retreating from architectures that expose them, and the hybrid model — public cloud, private cloud, on-premises and edge combined — is where a quarter of respondents plan to land.

Why it matters for Singapore: The survey validates the strategy of treating governance as infrastructure rather than an afterthought. Singapore's enterprises are ahead of the regional curve on workload repatriation, and that pragmatism — matching each AI workload to the environment that best balances latency, cost and control — is precisely what scaling AI safely looks like. As more local firms move from pilots to production, expect hybrid and edge setups to keep growing, and expect governance capability to become a competitive differentiator rather than a back-office concern.