With storage costs consuming a growing share of AI budgets, Kevin Dunn, Vice President and General Manager EMEA at Wasabi, explains why a more flexible approach to data management is needed to turn investment into meaningful returns.

Cloud storage was built on the principles of efficiency and flexibility, but as data infrastructure moves into the age of AI, those characteristics are being challenged. Data storage is more important than ever; AI applications would be rendered useless without the data on which their models are trained and from which they regularly draw.

However, when it is not managed well, cloud storage can become more of an anchor than an enabler. The wrong storage strategy can consume a significant share of an enterprise’s AI budget without delivering equivalent value.

For businesses looking to generate a meaningful return on investment (ROI) from AI, their storage strategy must be one of IT’s top priorities. Choosing cloud services that support flexibility and data movement can contribute to the effective operation of an AI programme.

Data infrastructure: the foundation of AI

The efficient operation of AI is only possible because of the extensive data networks running behind the scenes. As agents and other AI applications become increasingly important components of enterprise IT strategies, it is vital that sufficient investment and attention are given to the underlying data infrastructure.

Often, the AI data pipeline will begin and end in the cloud. In this scenario, intermediate stages such as training and processing are distributed across multiple environments, with training datasets, intermediate outputs and models regularly transferred, processed, stored and accessed across numerous on-premises and cloud sites.

The influence of data on the success of a project is reflected in how AI budgets are being distributed. In the most recent iteration of Wasabi’s annual Cloud Storage Index, survey results found that around two-thirds of AI budgets are now being spent on the data, storage and processing power that underpin AI applications, with 36% of spending dedicated to the software itself. What’s more, most British organisations surveyed (56%) plan to raise their AI-related spending further over the next year.

Increased investment, decreased ROI

However, while an enormous amount of money is being funnelled into AI data infrastructure, it is not always being converted into the results businesses had hoped for. Financial returns are too often falling short of expectations. Currently, only a quarter of UK businesses surveyed report achieving a positive ROI from their AI projects.

The disconnect between investment and value raises an important question: where is the money going?

It is easy to attribute this to deployments still being in their early stages and organisations currently focusing on set-up, but it is not simply a matter of waiting for projects to mature. There are also structural issues that can prevent companies from making their investments in AI worthwhile.

Data movement drives up costs

The way in which an organisation manages its data can have a significant bearing on the success of its AI initiatives. AI relies on immense and often unstructured datasets, which must be cleaned and prepared before they can be used. Data quality is therefore often cited as one of the key challenges organisations are working to overcome in their pursuit of AI success.

Another key challenge is cost. As the use of AI becomes more widespread, data access requests and transfers will continue to increase, amplifying the cost impact for businesses.

With some hyperscaler pricing models, a substantial proportion of cloud spending can be driven by additional charges, such as data access and API operation fees, rather than by storage capacity itself. According to Wasabi’s research, approximately half of UK cloud storage spending in 2025 was dedicated to paying fees.

This has been a recurring concern in the hyperscaler-dominated cloud market and is among the issues that regulatory bodies such as the UK’s Competition and Markets Authority have sought to address.

Data storage fees can therefore account for a significant portion of AI spending, reducing the funds available for other areas of development. Budget pressures do not end with fees, however; businesses must also increase their spending on storage as they scale their AI applications.

Cost has become a leading challenge for contemporary data management strategies. In Wasabi’s research, just under half (46%) of UK businesses reported exceeding their cloud storage budgets. These accessibility and budget-planning issues are not only financial burdens today, but also potential obstacles to long-term storage and the sustainable reuse of AI models.

Hybrid becomes the preferred approach

These challenges mean that many organisations are finding a single storage environment insufficient for their needs. The computing performance required for each AI application within an enterprise will differ, as will the frequency with which it requires access to data. Costs can therefore vary widely across a company’s AI deployments.

There is no one-size-fits-all approach to data infrastructure when it comes to enterprise-level AI. As a result, many organisations are adopting a hybrid approach. In the UK, 72% of businesses surveyed are using a combination of on-premises and public cloud storage for their AI workloads.

Typically, sensitive and frequently accessed data remains in company-owned data centres, while AI models may be stored in the cloud alongside large datasets. Dividing data storage in this way allows companies to plan strategically and move assets between environments as needed.

Cost and compliance requirements can be addressed according to the needs of individual workloads. A hybrid approach can also reduce dependence on a single provider and allow organisations to select environments based on their performance requirements.

That said, hybrid environments can also become expensive when fees are charged each time data is moved between sites. It is therefore important to understand a cloud provider’s billing structure and choose services with those costs in mind.

Flexible data management supports AI value

The multitude of workflows within the AI data pipeline places significant demands on data access speeds and data movement. Any resulting delays can directly affect performance, while the sheer volume of data transfers taking place can increase costs.

When high levels of investment in AI fail to deliver meaningful ROI, data management bottlenecks may be one of the contributing factors.

To achieve greater business value, organisations are increasingly using hybrid storage to support their AI strategies. It can provide the flexibility needed to manage growing data volumes and fluctuating access requirements while balancing complex fee structures.

Understanding the influence that data storage has on the success of AI applications is essential if enterprises are to manage their budgets and strategies effectively. With a flexible data management approach in place, businesses will be better positioned to realise the value of their AI investments.