A quick look at the current headlines surrounding AI spend continues to paint a clear picture of huge IT infrastructure investment.

In April alone, for example, there were stories about a $100bn Amazon-Anthropic deal alongside further multi-billion-dollar announcements from the likes of Nvidia, Oracle and Disney, among various others.

To say that this is causing increased pressure on supply chains and component availability, is putting it very mildly. GPUs remain the most visible victim of market distortion, but they also rely on high volumes of supporting components to operate effectively.

In particular, the supply of RAM and fast-access storage, such as SSDs, NVMe drives, and other low-latency, high-throughput technologies, is under the spotlight. As one piece of analysis recently put it, “AI giants are hoovering up the world's memory supply, and everyone else will pay for it in higher prices, delayed products, and canceled launches.”

Supply chains for these components are finite, with manufacturing and distribution capacity under pressure and, generally speaking, unable to keep pace. Supply-demand imbalances are emerging quickly, including reduced availability and significant upward pressure on prices.

A big part of the problem is that these issues are not isolated to AI-specific environments; the same components underpin general enterprise infrastructure, consumer electronics and various other relevant product categories.

High demand and limited supply

As always, the devil is in the detail. Suppliers are allocating limited capacity towards larger, higher-margin AI infrastructure projects. Almost everyone else is having to pivot from predictable procurement to strategies characterized by short-termism and delay.

Where infrastructure planning was typically based on expected workload growth, the key driver is now component availability. In this context, procurement timing becomes a critical factor, with organizations needing to secure capacity ahead of need. Flexibility is also more important, as rigid architectures are harder to adapt to under these market conditions.

For some organizations, overprovisioning has re-emerged as a risk mitigation strategy. Where additional capacity would previously have been minimized to control costs, some businesses are now maintaining headroom to protect against procurement delays and the risk that infrastructure components are simply not available when needed.

Consider this increasingly common scenario: An organization plans a routine infrastructure refresh to support its growth objectives and core business applications. This requirement is not AI-driven, just standard compute, memory and storage expansion. When they go to procure hardware, lead times are significantly longer than expected and pricing has increased significantly compared to previous refresh cycles.

What’s more, their preferred configurations are not immediately available, necessitating compromises in specifications or timing. The organization must decide whether to delay the project, accept higher costs or redesign the deployment. In some cases, existing infrastructure is kept in place for longer than planned to avoid disruption. Whichever route they take, the delay has a knock-on effect on infrastructure investment and, eventually, bottom-line performance.

Don’t get too excited

These issues show no concrete signs of abating any time soon, so what options are left for IT leaders?

Firstly, organizations need to place greater emphasis on securing access to capacity rather than assuming availability. From the outset, infrastructure planning should factor in supply constraints alongside workload and performance requirements.

For example, there are still service providers out there that hold available inventory or can provision capacity quickly to reduce exposure to delays. Organizations that maintain some level of readily deployable capacity can help avoid disruption when demand changes or their preferred components are unavailable.

Central to the overall planning process is maximizing visibility into current and expected future usage levels to support more accurate forecasting and reduce the risk of under- or over-provisioning. As part of this approach, reducing reliance on single environments or fixed configurations can improve resilience as conditions change. At present, there is emerging talk of a correction in RAM pricing, but as the analysis points out, in the context of a 2,200% increase, buyers shouldn’t “get too excited”.

Indeed, analysts such as IDC argue “knock-on effects for the device manufacturers and end users that could persist well into 2027.” For consumers and enterprises alike, they explain, “this signals the end of an era of cheap, abundant memory and storage, at least in the medium term.”

This is difficult to dispute. What we are witnessing is not a short-term disruption, but a structural shift in how infrastructure is supplied and consumed. With AI demand set to remain elevated, pressure on component availability is indeed likely to persist, continuing to influence infrastructure decisions across the market.

For businesses, current market dynamics still strongly suggest shifting the focus from maximizing efficiency to ensuring continuity and predictability where possible.

We've reviewed the best cloud storage.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit