Organizations are investing in the infrastructure needed to support AI, recognizing that success depends on the data and storage platforms that underpin it.

Wasabi's 2026 Global Cloud Storage Index found UK organizations are directing the majority of their AI budgets towards data, storage and compute. According to the research, almost two-thirds (62 percent) of AI budgets are allocated to the data, storage and processing power needed to feed and run AI applications, compared with the third (36 percent) spent on AI software and SaaS solutions. For many organizations, building the right infrastructure has become just as important as selecting the right AI tools.

That shift reflects the changing nature of enterprise AI. Models are only as effective as the data they can access. As organizations expand AI initiatives, they must manage growing volumes of structured and unstructured information, while ensuring that data remains available, governed, and accessible throughout the AI lifecycle.

Infrastructure challenges continue to affect AI outcomes

While investment in AI continues to grow, many organizations are still working to turn that investment into measurable business value. The research found that only about a quarter (25 percent) of UK organizations currently report a positive return on investment from their AI projects. However, almost half (48 percent) expect to achieve positive returns within the next year. When asked about the challenges affecting AI implementation, respondents most frequently identified data storage challenges, including cost, data access, and management.

The ability to move, manage, and access data efficiently has a direct impact on how effectively AI initiatives can be developed and operated.

AI creates new demands on storage infrastructure

Supporting AI workloads is about far more than storing increasing amounts of data.

AI is also changing the way organizations think about storage itself. Rather than supporting a single, predictable workload, storage environments now need to accommodate a mix of data ingestion, model training, inference, archiving, and governance requirements. Each stage has different performance, cost, and resilience demands, making flexibility an increasingly important consideration when designing infrastructure.

AI datasets are rarely static: training data is continually refreshed, models evolve through repeated iterations, and organizations retain artifacts and logs for governance and compliance. Managing that lifecycle efficiently places new demands on storage architecture, requiring organizations to think beyond capacity.

Wasabi's research highlights how organizations are responding to those demands. Nearly three-quarters (72 percent) of UK respondents said they are using hybrid storage, combining on-premises infrastructure with public cloud storage to support AI workflows. Respondents also identified data retrieval and aggregation, alongside model retention and archiving, as two stages of the AI workflow where they most commonly use public cloud storage.

Different AI workloads place different demands on infrastructure. High-performance data ingestion, long-term model retention, governance requirements, and cost optimization don’t always sit comfortably within a single environment. As a result, rather than relying on a single environment, many organizations are building architectures that allow workloads to move between on-premises infrastructure and the cloud depending on operational requirements.

This flexibility is becoming increasingly valuable as AI projects move from experimentation into production. Infrastructure that can support changing workloads without requiring wholesale redesign allows organizations to scale AI initiatives more efficiently while maintaining control over performance, cost, and operational complexity.

Managing growth without losing control

As AI environments expand, organizations are also managing increasingly complex storage estates. The Cloud Storage Index found that roughly half (46 percent) of UK respondents exceeded their cloud storage budgets last year. At the same time, respondents reported that 48 percent of their cloud storage spending is allocated to fees rather than storage capacity itself.

Managing costs while continuing to scale AI infrastructure is becoming an important consideration for IT leaders. As data volumes increase, organizations need storage architectures that provide predictable performance and governance while helping to keep infrastructure spending aligned with business expectations.

Alongside cost considerations, organizations are also focused on ensuring their data remains protected and available. The research found that over half (62 percent) of UK organizations are not completely confident their data would remain operational and unaltered after a cyberattack, while only 51 percent are completely confident their data would meet regulatory or compliance requirements if audited by a third party. These findings underline the importance of building infrastructure that supports both operational resilience and governance as AI deployments mature.

Looking ahead, infrastructure decisions are likely to play an even greater role in determining how successfully organizations can operationalize AI. As data volumes continue to grow and AI workloads become more diverse, storage strategies will need to balance performance, governance, and cost without sacrificing the agility needed to support evolving business requirements.

Building for the next phase of AI

Data quality, storage architecture and infrastructure design are becoming increasingly important factors in supporting AI workloads long-term. The organizations that are best positioned to realize value from AI are likely to be those that view storage as more than a capacity requirement. Building infrastructure that enables data to be stored, managed, and accessed efficiently provides the foundation on which AI initiatives can continue to grow.

Current purchasing patterns suggest that as organizations allocate more investment towards data, storage and compute, infrastructure is becoming an increasingly important part of the conversation around successful AI adoption.