Most CIOs have a cloud strategy. Very few have a strategy for keeping it simple. That gap shows up where enterprises can least afford it: cost and risk. It also slows the business down, as teams spend more time managing where data lives and how to access it than putting the data to work.
When companies first moved to the cloud, the promise was real: more speed, more agility, and faster access to services that would have taken years to build. Then cloud estates expanded. Different providers served different workloads, while crown-jewel data and workloads requiring greater control or performance remained on-premises. That choice is continuing to expand as enterprise data services become more deeply integrated across the major public clouds.
The cost is not only in the cloud but in the operational overhead: moving and copying data, managing different tools and policies, and reconciling governance across multiple locations. Every new silo adds another location where data needs to be secured and managed. The administration of this data introduces another level of risk and cost that needs to be managed.
More Options Do Not Automatically Mean More Value
The cloud market keeps growing, but the value it delivers doesn’t always follow the same trajectory. More choice can also mean more complexity: more things to evaluate, more systems to connect, and more to configure, secure, and reconcile across environments.
The problem is not a lack of cloud tools. Most companies have plenty. The problem is the friction created when data, applications, and governance work differently depending on where a workload runs. A data team waiting to access or move the data it needs is not productive, and time is money. When storage and control teams manage separate rules across multiple clouds, that introduces friction. When moving a workload from one place to another means rebuilding it from scratch, that introduces friction.
That friction compounds in ways that show up directly in AI outcomes. According to S&P Global Market Intelligence’s 2025 survey, 42% of companies abandoned most of their AI initiatives that year, a dramatic rise from just 17% in 2024. On average, those organizations scrapped 46% of AI proofs-of-concept before they ever reached production. The failure pattern, examined across hundreds of these initiatives, points consistently to the same upstream cause: not the models, but the data infrastructure beneath them. Companies that cannot move data cleanly, govern it consistently, or make it accessible where AI services actually run are compounding the problem and paying for AI investments that never fully land.
Simplify the Data, Not the Cloud
There is a leadership dimension here that does not get enough attention. Cloud decisions are often made by teams focused on what a technology can do, rather than on what the business is striving to achieve. The result can be a portfolio of capable tools sitting on top of a data layer that was never designed to move easily among them.
To be direct: simplicity here does not mean fewer clouds. Most enterprises will run across on-premises systems and multi-cloud setups for years, and that makes sense for the variety of workloads that they run. Nor should every workload run everywhere: performance, economics, SLAs, and native services all affect where it belongs. The goal is to make those workload decisions without creating a new data silo every time. Recent advances in native cloud services and data mobility are making that model more practical, allowing organizations to expand workload choice while preserving familiar data management and operational practices.
That requires a common data layer across environments. File, block, and object data should be governed and protected through a common operating model, wherever workloads run. Data should be able to move when a workload requires it without rebuilding the surrounding storage and needing specialized skills to pull it off. It means AI services should be able to reach data where it lives rather than only where it has been copied.
When the data layer removes friction rather than creating it, cloud choices become easier to manage, operations become more consistent across the estate, and AI projects can more easily move from pilots to production. The next step is not simply connecting more environments, but giving teams a unified way to see, govern, protect, and increasingly automate how data is managed across them.
