AI datacentres are inordinately hungry, requiring colossal amounts of Random Access Memory (RAM) to power their servers. They’re expected to monopolise 70% of all high-end memory chips in 2026, according to the WSJ, and this has led to mass memory shortages that threaten to destabilise supply chains.
It’s a situation now being further exacerbated by manufacturers such as Samsung, SK Hynix and Micron diverting production away from dynamic random access memory (DRAM) and NAND drives used in conventional computing devices towards high-bandwidth memory (HBM) meet AI datacentre demand. Add to that companies such as Cisco choosing to exercise their right to cancel orders 45 days before shipment and HPE halving its quote validity window from 30 to 14 days and it’s clear these are unprecedented times, where the usual rules of supply and demand simply don’t apply.
From a business perspective, these shortages are now having a significant impact on budgeting, project timelines and operational resilience and could even jeopardise viability.
Elastic budgets
Budgets are being stretched thin but are also having to flex to meet changing timelines which is requiring them to be almost elastic. Usually, shortages equate delays but with manufacturers reducing order windows, businesses are now under pressure to adjust i.e. shorten order/delivery timeframes. As a consequence, there has been a shift from ‘buy now’ to address the issue to ‘what needs replacing’ as the business focuses on business-critical memory.
Where memory is a primary component for server builds, expansions or upgrades, it’s also likely to cause delays that will impact the ability of the business to evolve. As data centres begin to feel the pain, we can expect capacity planning to become more uncertain, for instance, resulting in longer lead times that could see the business under-provisioned. And the scaling of RAM-intensive workloads such as virtual desktop infrastructure (VDI), databases, analytics and AI inference could also suffer, reducing architectural flexibility.
IT overload
IT teams will become overloaded. If replacement parts or devices aren’t available when needed, this has an adverse effect on refresh programs, office moves, new site opening or security upgrades. So, IT will spend more time on keeping old equipment in service through break-fix support and seeking extended warranties. Even standard end user devices will become more difficult to maintain as preferred memory tiers dry up and, laden with ageing equipment and less resource to secure and patch them, businesses will become more exposed and susceptible to attack.
Ultimately, the shortages will hamper the ability to operate efficiently and to transform, because in addition to these internal pressures, the costs associated with training and deploying LLMs will go up, which will then curtail the adoption of AI in the enterprise. However, there is a silver lining to these memory shortages in that this will force a radical rethink of procurement practices and it’s this that could place the Channel in a prime position to advise, assist and act on their customers’ behalf.
Leveraging channel connections
Technology resellers and systems integrators (SIs) are ideally placed to advise businesses on how to stay ahead by focusing on inventory hold and lifecycle planning. Procurement timelines, for example, can be altered to focus on critical projects rather than waiting for the deployment window. Alternative approaches, such as capacity tiers, module types and approved OEM options to reduce the risk of last-minute changes, can also be considered, while stock hold can see assets retained for high failure or high urgency components.
As technology providers often enjoy a closer relationship with manufacturers, it can leverage these to provide real-time supply chain insights and in-stock inventory. Some may even be in the position to offer financing to help customers juggle budgets and timelines. Or they could offer cloud-based alternatives such as Desktop-as-a-Service as a substitute for user hardware or Infrastructure-as-a-Service or Platform-as-a-Service for workloads to keep projects on track.
Inevitably, organisations will need to explore other options and SI’s are particularly well placed to help test alternative options and experiment with specifications in labs or via benchmarking tools. This could even include road-testing AI workloads over these devices to inform purchasing decisions for future device fleets. SIs should also be willing to review plans, design solutions and offer forecasting support.
Intelligent refresh
From a strategic point of view, the Channel can help businesses to develop intelligent refresh programs (IRPs). These identify the most efficient refresh paths and are likely to become essential as organisations focus not just on replacement but refurbishment and redeployment. The reality is that these businesses will also be managing a far more mixed OEM environment, which will add complexity. But an IRP can simplify matters by providing a shared view of refresh timelines, supply risks and budget triggers as well as a runbook on what to accelerate, what to delay and what to standardise, easing pressure on procurement.
Channel partners can also leverage their expertise in IT asset management (ITAM), software asset management (SAM) and FinOps to uncover cost reduction opportunities, rationalise applications, ensure right-size licensing and to free up additional budget. And asset recovery can even allow the business to extract, resell or redeploy memory from end-of-life devices.
While it’s not yet clear how long the shortages will last, reports suggest that suppliers will be bringing new plants online towards the end of 2027 and into 2028 and of course, technological innovation such as chip stacking could help alleviate the issue. But in the meantime, Channel partners can ease the pain by helping organisations adapt to these market conditions and in so doing provide a range of valuable services.
Ian Foddering
Ian Foddering is VP of Europe at SHI
