By now we’re all too familiar with the memory chip shortage. Some semiconductor industry experts expect the crunch to last through 2027. If you’re not feeling the effects of it yet, in higher prices and longer lead times, you will soon. Gartner estimates prices for DRAM and solid-state drives (SSD) will increase 130% by the end of the year.
The reason for the shortage is simple: demand for AI infrastructure is growing in every industry, driving a huge investment in data centers that could exceed $800 billion in 2026. As a result, the global semiconductor market is shifting, with manufacturing capacity increasingly allocated to high performance GPUs and high bandwidth memory.
Enterprises could experience far-ranging effects: higher server and PC prices, more expensive network gear, higher storage prices, higher cloud infrastructure costs, and more expensive (or delayed) AI deployments.
With memory supply short and prices going up, here are five ways to mitigate the impact on your organization:
1. Rightsize existing memory
IT teams often provision more resources than a workload actually needs to hedge against execution issues—outages, latency spikes, out-of-memory errors, and more. When RAM was cheap, this was not a big problem. But with rising costs and supply chain disruption, deliberate overprovisioning becomes too expensive to continue. It’s better to optimize memory utilization before purchasing more capacity. Some ways to do that include:
- Reducing overprovisioned virtual machine memory
- Adjusting Kubernetes memory requests and limits
- Consolidating underutilized databases
- Tuning JVM, cache, and application memory settings
- Archiving cold data from high-performance storage
- Moving noncritical workloads to lower-cost storage tiers
- Eliminating idle development and test environments
To help identify and act on these opportunities, ePlus offers a Memory Analysis and Optimization Assessment — an intelligent decisioning platform that analyzes real workload demand and surfaces actionable reclamation recommendations. It's a good place to start.
2. Avoid niche hardware configurations
With more manufacturing capacity devoted to supporting AI infrastructure demands, chipmakers want to use what is left to churn out profitable, easy-to-build products. This means high volume, standard configurations. Non-standard orders—those with niche memory densities—are pushed down the priority list and placed in backorder status.
Standardization helps procurement teams, too. A fragmented server estate with many unique configurations is harder to source, especially during a shortage, which could put key infrastructure refresh projects and data center modernization initiatives at risk.
Review your hardware orders. Reconfigure your requests to avoid non-standard configurations where you can.
3. Engage vendors earlier and leverage strategic sourcing
One option is to extend the life if existing assets —deferring new purchases until prices stabilize But, delaying purchases may not be feasible or wise if the shortage runs into 2026 and beyond.
For critical infrastructure, engage your vendors early, and if possible, negotiate long-term supply agreements. Hyperscalers and large enterprises are already doing this. You may not have the leverage they do, but you can still reduce risk by consolidating demand forecasts, pre-committing to critical configurations, and avoiding last-minute orders.
4. Classify workloads into tiers
Not every workload needs equal access to compute resources. High priority workloads take precedent over everything else. If you haven’t done so already, classify workloads based on value to your organization. A simple example might be
- Tier 1: customer-facing, revenue, security, and regulatory workloads
- Tier 2: modernization projects that improve resilience, efficiency, and operating cost
- Tier 3: discretionary projects—non-essential upgrades, pilots, etc
When memory is scarce (or very expensive), workload classification helps ensure you are getting maximum value from constrained resources.
5. Use cloud capacity strategically
Cloud providers are subject to the same supply chain disruption as every other enterprise, although their size and scale afford them options that most companies don’t have. Still, cloud providers can’t absorb price increases and shipment delays forever—at some point, those higher costs and limited availability will get passed down to customers. This may show up as higher prices or could take the form of limited capacity, longer wait times for large reservations, and less discounting.
Make sure to plan cloud use strategically. Reserve capacity for critical workloads, use multiple regions where feasible, and don’t assume memory-optimized instances will always be available on demand.
What can you do now?
The shortage is out of your control, but your response to it isn’t. There are things you can do to soften the blow. Rightsizing your existing memory — reclaiming unused capacity before buying more — is the most immediate great step.
The ePlus Memory Analysis and Optimization Assessment gives you a data-driven view of where memory is being wasted and what can be reclaimed. For more information, go to ePlus Memory Optimization and Reclamation Assessment.