AI Data Growth is Driving Unpredictable Storage Costs: 2026 Wasabi Global Cloud Storage Index

As AI projects move from proofs of concept into production workflows, organizations are under pressure to show measurable returns, even as many are still working toward that goal.

The 2026 Wasabi Global Cloud Storage Index reports that while AI investment continues to accelerate, organizations are navigating a more complex set of challenges around cost, performance, and control. To understand what’s driving that shift, our research combined global survey data with in-depth interviews from IT leaders actively building and scaling AI workloads.

The takeaway is consistent: AI is making costs harder to predict, harder to control, and more consequential when they go wrong, especially when it comes to storage.

AI budgets are rising, but so is the pressure to show return

According to Cloud Storage Index survey respondents, 59% of organizations expect AI budgets to increase over the next twelve months. On average, two-thirds of that spend is going toward infrastructure (data, storage, and compute) rather than software or SaaS tools. That split means most AI investment is tied up in the systems that power it, not just the applications that showcase it, and that’s where cost pressure shows up first.

At the same time, AI is no longer treated as experimental spend and teams are being asked to justify it. Only 32% of organizations report seeing positive ROI from AI today, even as 51% expect to get there within the next year. That gap is creating a new kind of pressure: organizations need to keep investing to move AI forward, while also tightening control over how that investment scales.

What’s making that difficult is how quickly AI budgets can change. As new data sources, tools, and use cases are introduced, costs don’t scale gradually. As one IT leader described:

That volatility narrows the margin for error. When infrastructure accounts for the majority of spend, small decisions like adding a new dataset, enabling a new workflow, or increasing access can compound quickly across the organization.

Why AI makes storage costs harder to control

For 47% of organizations, AI data storage is the most cited challenge when implementing AI and managing costs. The issue is that AI continuously consumes, generates, and reuses data across the lifecycle. As models are trained, fine-tuned, and deployed, data is constantly moving and expanding. Some of the biggest storage cost drivers include:

• Transferring large training datasets to compute environments

• Storing multiple versions of data, including cleansed and compliant copies

• Retrieving data to support inference and downstream applications

• Retaining logs, checkpoints, indices, and other metadata for retraining and performance tuning

At the same time, data quality requirements are increasing. Organizations can only continue driving positive ROI by improving their models, and better models are built on higher-quality data.  Since model performance depends on well-prepared data, organizations often create and maintain multiple versions of the same datasets, adding to both storage footprint and management complexity.

Compliance adds another layer. Organizations need to document how models are trained, what data they use, and how outputs are generated. In regulated environments, that means maintaining audit trails that can stand up to scrutiny.

The result is a growing volume of data that must be stored, managed, and kept accessible over time, making storage a central factor in controlling AI costs.

The cloud fee problem gets worse under AI pressure

Cloud cost overruns are already common, with 49% of organizations exceeding their cloud budgets in 2025. AI is accelerating that problem by exposing a deeper discrepancy between how AI workloads behave and how cloud storage is priced.

The challenge is unpredictability. AI systems don’t access data in consistent or easily forecasted patterns. To keep models and applications functioning, organizations often need large datasets to remain readily accessible, whether or not they’re being used continuously.

At the same time, a significant portion of cloud storage spend isn’t tied to capacity. On average, 50% goes to fees. Each time data is retrieved for inference, moved for retraining, or accessed across environments, API request fees and egress charges apply.

For many AI use cases, it’s difficult to predict how often data will be accessed or how it will move. As a result, 42% of organizations report higher-than-expected data operations, and 38% report the same for API call fees.

Why hybrid and multicloud are becoming part of the answer

As AI costs become harder to predict and control, organizations are rethinking how their infrastructure is designed. Instead of relying on a single provider, many are turning to hybrid and multicloud approaches to regain flexibility across performance, availability, and cost.

This shift is already widespread. Eighty-one percent of organizations use multiple cloud providers for object storage, and 64% use hybrid storage deployments to support AI workloads.

Avoiding vendor lock-in is a growing priority. When data is tied to a single provider, it becomes harder to adapt as costs, performance needs, and usage patterns change. Maintaining the ability to move data and choose where it lives gives organizations more control over how those tradeoffs are managed.

That flexibility also helps reduce dependency. Rather than being locked into a single pricing model, organizations can shift data and workloads as needs change, giving them more leverage as AI usage scales.

Companies cite several drivers behind this approach, including:

• Wider performance options (49%)

• Application availability (46%)

• Improved cost of ownership (42%)

What IT leaders should prioritize now

AI data storage requirements will continue to expand as initiatives scale, usage increases, and new workflows generate more data that must be retained. That makes storage strategy an early decision, not something to optimize later.

The teams getting ahead of this are prioritizing a few key capabilities:

• Predictable pricing to avoid cost volatility as usage scales

• High-performance access to support training, inference, and real-time workloads

• No egress or API fees that penalize data movement and retrieval

• Flexibility across environments to adapt as requirements change

• Freedom from vendor lock-in to maintain long-term cost control

These are a response to how AI workloads actually behave: unpredictable, data-intensive, and constantly evolving.

Wasabi addresses these requirements with high-performance object storage and a pricing model designed for predictability, with no egress or API fees, helping organizations maintain cost control as AI workloads grow more complex and less predictable.