With demand for storage, memory and compute becoming increasingly unpredictable, Mark Molyneux, Field CTO North Europe at Commvault, explains why cloud strategy can no longer stand still.

The business pressures associated with AI adoption have taken another significant turn. The technology has fundamentally changed the scale and speed of enterprise data generation in data centres, particularly as organisations move from experimentation into production-scale deployments.

One of the many consequences of this trend is the scramble to build AI infrastructure, and the associated component grab is now extending far beyond GPUs. Recently, storage systems, flash memory, DRAM and supporting hardware supplies in data centre environments have all come under increasing pressure from demand created by AI-related workloads.

Across the board, data centre storage pricing is becoming increasingly unpredictable, a challenge that quickly translates into budgeting and capacity-planning issues for IT leaders. This is unfamiliar territory and is being compounded by wider supply-chain pressures, where shortages in one part of the semiconductor ecosystem can ripple across the broader infrastructure market.

If that wasn’t difficult enough, many organisations are now finding themselves competing for infrastructure capacity against hyperscalers and AI-native companies with significantly larger spending power.

Granted, it’s an overused term, but this is transformational. Data centre storage is shifting from a routine operational and financial line item into a more strategic business concern tied directly to scalability, resilience, long-term infrastructure planning and a range of other associated issues.

Horses for courses

For organisations everywhere, something has to give. One way many are choosing to mitigate the impact is to reassess how workloads are distributed across multi-cloud environments, rather than relying too heavily on a single provider or architecture.

Indeed, multi-cloud adoption is increasingly being driven by operational pragmatism. This is characterised by an emphasis on balancing performance requirements, controlling costs, improving resilience and accessing specialised AI capabilities across different providers.

This is helping to shape the market, with different cloud providers increasingly associated with different strengths across AI tooling, compute availability, analytics capabilities, geographic presence, pricing structures and sovereignty options within data centre ecosystems. The list goes on, but the underlying point is that many organisations no longer want to commit entirely to a single ecosystem because infrastructure conditions are changing too quickly for inflexible long-term architecture decisions.

In this context, multi-cloud can provide organisations with greater flexibility to move workloads in response to the uncertainties currently determining data centre infrastructure costs. It can also offer IT leaders more options as they look to balance innovation priorities with wider risk-management and operational-resilience requirements.

Managing multi-cloud environments

This isn’t simply a matter of diversification, however. Distributing workloads across multiple environments inevitably changes the management challenge for IT teams. For instance, multi-cloud environments can quickly become operationally fragmented because each provider has its own set of tooling, interfaces, management models, policy frameworks and other idiosyncrasies.

One consequence is reduced visibility across the wider data centre infrastructure estate, making it harder for IT teams to maintain a clear understanding of where data resides and how it is being used. Access governance also becomes more difficult as workloads and datasets move between environments with different policy structures and security controls.

The use of disparate platforms and environments accentuates cyber resilience challenges. When data and workloads are spread across public cloud, private cloud and on-premises systems, it becomes harder to maintain consistent security, backup and recovery processes. Different tools and policies across environments can create gaps in visibility and protection, making it more difficult for IT teams to respond quickly to cyber incidents.

AI introduces an additional layer of complexity because organisations must now manage and secure not just enterprise data, but also models, pipelines, training datasets, prompts, outputs and associated metadata within data centre environments. In many settings, these environments are growing very quickly and, consequently, the operational challenge changes from simply deploying AI workloads to maintaining control at scale.

As for overall cost management, multi-cloud doesn’t automatically solve this problem, particularly if organisations lack a unified understanding of resource consumption, storage utilisation and data lifecycle management across their various providers. Questions such as what data genuinely needs to remain in high-performance environments versus what can be archived or deleted have, arguably, never been more relevant.

Many organisations are now looking for ways to establish a more unified operational view across cloud and on-premises environments, rather than managing each platform in isolation.

Consistent policy enforcement is becoming a key priority, particularly around access governance, compliance requirements, data protection and workload resilience.

From a data centre storage perspective, organisations also need a clearer understanding of how storage resources are being consumed across environments as infrastructure costs rise. AI environments amplify the importance of these decisions because training datasets and associated metadata can quickly scale storage consumption, particularly if left unmanaged. The broader, and increasingly urgent, challenge for organisations is no longer simply how to adopt AI, but how to do so sustainably while maintaining operational control and financial predictability.