Cloud: A Vital Tool for Navigating the Constrained Memory Market

Introduction: How Memory Market Constraints Are Impacting IT Now

Over the past year, the global memory market has experienced well-documented supply constraints, driven by limited manufacturing capacity expansion, dramatically increased demand from AI deployments, and manufacturers allocating precious capacity towards higher-margin products. As a result, standard DDR5 DRAM memory prices have risen meaningfully, and lead times and availability remain unpredictable for many customers. For enterprises that depend heavily on compute infrastructure, these conditions create real friction.

Many organizations are finding that infrastructure upgrade plans are no longer  straightforward decisions or executable as planned. Limited memory availability can delay refresh cycles, while higher prices increase the capital and operational cost of expansion. Deferring upgrades may feel prudent in the short term, but must be balanced against higher long-term costs through inefficient resource usage, slower application performance, and lost opportunities to modernize. In this environment, prudent evaluation of resource requirements has become a strategic imperative.

As customers seek to evaluate their options, they need data.  With memory being harder to acquire or a higher percentage of the system cost, it makes sense that they want to have a more accurate sense of how the memory they use in systems is being used.  After all research reveals that workloads have a wide variation in how much memory is needed to meet required performance levels. Of course, IT managers can get some of the basic utilization information from leading systems management platforms, and that is a great start.  But knowing how their application performs in different memory allocation scenarios can be more challenging, as few IT managers want to manually reconfigure a server for test runs.

In this situation, clouds can play two critical roles in helping to manage costs and accelerate infrastructure availability.

Role #1: AMD EPYC™ server CPU Advanced Cloud Eases Memory Optimization Analysis 

One of the most effective ways to address memory constraints is to leverage cloud as a tool to help right-size infrastructure and extract more value from every gigabyte of memory and every CPU core deployed. With high core counts, strong per-core performance, and robust memory and I/O capabilities, AMD EPYC™ server CPU processors are designed to maximize compute: helping you run more work on smaller infrastructure. 

In practical terms, higher performance on EPYC CPU-powered infrastructure enables organizations to reduce overall VM sizes, run fewer instances, or shorten application run times. AMD EPYC server CPU-powered cloud instances available from all of the leading global cloud providers provide a fast and easy way to assess alternative memory configuration scenarios.  Rather than physically altering on-premise server configurations and running workload performance tests, these tests can be easily, inexpensively and even concurrently run in appropriate cloud instance types. 

Many cloud vendors offer instance types that are “memory optimized” or “compute optimized” or “general purpose,” with each instance type varying primarily by the ratio of memory to virtual CPUs (vCPUs).   It is relatively easy for a customer to put a workload into target instance types that would mirror the options available in on-prem configurations and to get a real evaluation of what performance might look like in scenarios using less memory than currently provisioned.  This could reveal where there is excessive over-provisioning of memory and identify opportunities to equip the next batch of servers for that workload with less memory (and at less cost).  

Adopting AMD EPYC server CPU-powered compute in public cloud can enable you to:

• Reduce VM size: Higher performance per vCPU can allow you to move to smaller instances with less memory while still meeting SLAs.
• Reduce instance count: More capable instances can consolidate workloads that previously needed multiple underutilized VMs.
• Shorten application run time: Ephemeral, batch, analytics, and High Performance Computing (HPC) jobs can complete faster, reducing total compute and memory time consumed.

The second way that cloud provides help in the time of memory scarcity is even more straightforward.

Role #2: The Hybrid Cloud Hedge

A significant portion of AMD EPYC server CPU processor business growth has come via installations at the world’s leading cloud providers.  These providers have found tremendous value in the performance they can offer customers while providing leadership energy efficiency when utilizing AMD EPYC server CPUs.  As such, there is a wide variety of AMD EPYC server CPU processor powered cloud offerings that businesses of all sizes have been able to utilize as part of their hybrid IT strategies.

Moreover, customers have found AMD EPYC server CPU processor-power instances as a reliable vehicle for driving consolidation of their cloud footprint to drive savings. The foundation is simple: when your CPUs do more work per core, your applications require fewer resources overall, including memory, especially given fixed memory to vCPU ratios typically available from cloud providers.  Customers can use EPYC CPU powered instances to either raise performance at a similar footprint or maintain performance with fewer resources.

Having access to the same AMD EPYC server CPU infrastructure in the cloud as they deploy on prem also provides a “pressure-release valve” that can help in time of system and memory constraints.  Rather than being gated by hardware availability or cost, customers may simply choose to shift some incremental workloads into cloud deployment to keep their IT refresh on-track.  In general, cloud vendors are not yet reflecting hardware constraints in their instance availability and pricing, but this too could change, so flexibility is the key.  As is knowing what you really need more than what you might have typically purchased in the past.

Here to Help: AMD and Partner Ecosystem Assessment & Execution Support

To help customers systematically identify these “right sizing” opportunities, AMD offers our own tooling for infrastructure assessments and recommendations, as well as tooling and transformation services through our partner ecosystem.

AMD offers the EPYC Instance Advisor (EIA), a free, data-driven right-sizing assessment tool. EIA captures and analyzes a lightweight telemetry fingerprint of existing cloud workloads—including CPU, memory, network, and storage utilization. Using real usage data rather than assumptions, EIA recommends right-sized modern instances, whether that means downsizing, upsizing, or migrating to more efficient EPYC server CPU powered VM types where there is clear benefit.

While tools are valuable, successful optimization often requires experience and structured execution. AMD works closely with an ecosystem of partners that specialize in cloud assessments, workload modernization, and large-scale migrations. These partners help customers translate right-sizing insights into transformation execution with measurable outcomes.

AMD and our Partners can help you to:

• Run large-scale telemetry assessments across your cloud accounts or data centers.
• Interpret detailed CPU, memory, network, and storage profiles.
• Design a phased migration plan focused on risk reduction and business impact.
• Execute and validate transitions to AMD EPYC server CPU-powered instances.

Customer Success Story with HCLTech 

A leading Network Equipment Provider (customer) was facing escalating cloud infrastructure costs and declining price performance efficiency driven by large, memory intensive workloads running on legacy x86 instance types across both Azure and AWS. With a highly distributed footprint and thousands of instances in production, customer needed a low risk way to optimize costs without disrupting performance critical telecom workloads.

HCLTech partnered with AMD to conduct a cloud modernization and memory rightsizing assessment using the AMD EPYC™ Infrastructure Advisor (EIA), analyzing over 2,000 instances on AWS and 400 instances on Azure. The assessment identified opportunities to modernize and downsize to AMD EPYC server CPU powered instances while preserving workload stability.

By executing a phased modernization strategy, HCLTech helped customer overcome performance risk concerns and change management complexity, delivering significant annual savings and performance uplift. The engagement enabled customer to significantly reduce cloud spend while simultaneously improving workload performance, creating a scalable blueprint for ongoing cost and performance optimization across their global cloud estate.

Next Steps

Given today’s memory market constraints, organizations should look to leverage cloud.  First, as a testing environment to understand true workload CPU/memory needs.  And beyond that, look for savings by assessing their current cloud estate with a specific focus on downsizing opportunities. In many environments, memory—not CPU—has now become the dominant cost and availability driver.

AMD and our partners can help with this effort through structured assessments and transformation programs that identify where EPYC CPU-powered instances can deliver immediate benefits. By right-sizing now, enterprises can regain control over memory costs, do more with existing cloud infrastructure, improve performance, and continue modernizing—even in a constrained market.

Please reach out to AMD for assistance as you begin your journey to optimize your infrastructure. AMD can work directly with you to assess right-sizing opportunities, or connect you with trusted partners that specialize in cloud assessments, modernization, and large-scale transformations. Either way, you don’t have to navigate this optimization effort alone.