As the foundations for the AI economy are laid, the UK’s data center expansion has driven a mixture of headlines around its innovations and challenges.
But the ambitious IT infrastructure strategy is hitting a wall – and it isn’t power or space. The global memory shortage is starving projects of a critical hardware component as it drains from the market.
Article continues below
CTO, Park Place Technologies.
With continued demand for AI, the world's biggest memory manufacturers have been shifting production away from standard enterprise DRAM – used for servers and storage arrays – towards AI-focused high-bandwidth memory, as the products compete for the same factory space. This has resulted in a shrinking supply of components that enterprise data centers run on.
And the urgency for capacity isn’t slowing down. London's data center market added a record 193MW of new supply in 2025, and another 180MW is expected this year. With the government's AI Opportunities Action Plan in place, enterprises across the country are feeling the pressure to modernize infrastructure for AI.
Since the market shows no signs of stabilizing soon, what can organizations do now?
Understand that traditional planning is broken
Historically, IT infrastructure strategies have been centered around stable supply chains with predictable pricing, available hardware and a refresh cycle every few years where everything is ripped out and replaced.
But the move to AI workloads has dismantled these assumptions all at once. The same technology driving the demand for scaling data centers is also the thing that is cannibalizing the supply chains needed to build that infrastructure.
And for the organizations locked in multi-year digital transformation plans based on procurement assumptions from a few years ago, they are finding that the hardware is no longer available at the price they planned for – or sometimes available at all.
This needs to be treated as a reset in how leaders think about planning for enterprise infrastructure. The question shouldn’t be about waiting for supply to normalize; it should be about building resilient infrastructure at a time where uncertainty is the norm.
Shift to diversified sourcing
A growing number of UK enterprises are now acting proactively, not waiting for OEM supply chains to recover, and pragmatically diversifying how they source infrastructure components.
This has seen third-party maintenance, pre-owned enterprise hardware, and alternative sourcing channels move rapidly into the mainstream. They are no longer seen as a last resort — they are becoming a legitimate first consideration in infrastructure planning.
However, many companies are still conditioned to buy new. Speed and cost are both under pressure right now, and alternative sourcing options give organizations far more flexibility than OEM-direct procurement, with meaningful capital expenditure reductions alongside faster deployment timelines that aren't restricted to a single vendor's schedule.
This can be enough to kill AI projects in sectors that already face long procurement cycles and budget windows, or for projects where deployment is delayed and pushed beyond financial year boundaries.
Shifting to diversified sourcing isn’t the main culprit for increasing operational risk - single-vendor dependency is.
Prepare for persistent uncertainty
The most forward-thinking infrastructure teams are going further. With hardware shortages on the rise and uncertainty causing continued delays, some organizations are looking to bake supply chain resilience into the architecture of their planning cycles rather than leaving it down to emergency responses.
Often, the benefits of extending hardware lifecycles and enhancing maintenance and monitoring are ignored as a strategy. But if done well, it significantly reduces refresh frequency and the exposure to changeable procurement windows.
Lifecycle extension can also be supported by building modest inventory buffers for critical components rather than running lean configurations that assume just-in-time availability.
Enhancing monitoring and predictive maintenance management is not radical, but it is overlooked. Maximizing the operational life and reliability of existing assets is one of the most important, and most underused, levers. Businesses should be looking to it as a means to continue delivering their strategy even through the memory crisis.
Infrastructure leaders will not be able to outrun the current shortages, but those who adapt their strategies now will be better placed to keep projects moving. This crisis will undoubtedly shape enterprise IT strategy for years to come, acting as a reminder that building for uncertainty, rather than stability, is now simply the future.