The data center memory shortage is not transitory

Reading Time: 7 minutes

What is the memory shortage?

As organizations lay the architecture foundations for AI, they face an inconvenient reality. While they may need to upgrade memory and storage—resources that are critical for data-intensive applications such as AI—rising equipment prices and limited supply can hamper their plans.

Ironically, the DRAM shortage to fuel enterprise AI stems, in part, from increasing AI demand.

Rising AI data center demand for high-bandwidth memory (HBM) has prompted manufacturers to deprioritize DRAM production in favor of higher-margin HBM AI chips. With constrained supply given less production of DRAM and hyperscaler players gobbling up HBM, DRAM memory prices have risen dramatically and supply remains tight.

The new normal for the IT infrastructure supply chain

This new normal in memory capacity—and indeed IT equipment more generally—is expected to persist beyond 2026. Gartner estimates that that DRAM pricing will increase 199% in 2026 and 2% in 2027  given a 6.9% and 3.6% undersupply in these years, respectively. And some predict shortages through 2030. Companies need to recognize that rising memory pricing is not a transitory problem that will resolve in a few months’ time.

Some believe that constrained capacity will be resolved by new manufacturers entering the market, such as China’s CXMT. But additional production won’t ease the supply constraints as quickly as many expect. Existing supply agreements, regional trade restrictions, and continued investment in higher-margin AI memory products can all constrain the impact of this new capacity has on enterprise availability and pricing.

Organizations that need to upgrade memory can’t just buy their way out of the problem. Longer procurement cycles, advanced ordering, higher equipment prices, and enduring supply shortages will persist, despite enterprise efforts to modernize IT environments.

In what follows, we explore how organizations can navigate the shortage—with some caveats.

With constrained memory supply given less production of DRAM and hyperscaler players gobbling up HBM, memory prices have risen dramatically and supply remains tight.

Five memory shortage insights

1. The financialization of the semiconductor supply chain

Organizations need to grasp that the memory supply chain has undergone lasting behavioral change. Increasingly, it behaves like other commodity markets, such as crude oil. Access to resources increasingly depends on organizations committing months or years before deployment, not simply on real-time pricing when the hardware is needed. That creates new dynamics for memory purchasing.

Data center memory is now secured through forecasting and allocation strategies rather than traditional purchasing. Large hyperscalers and AI infrastructure providers are locking in future capacity through LTAs, which reduces the amount of supply available on the open market. As a result, enterprises can’t assume that memory will be readily available when a project receives budget approval. In turn, infrastructure planning now requires closer alignment between IT, finance, and procurement teams to ensure access to future supply.

Data center memory is now secured through forecasting and allocation strategies rather than traditional purchasing.

2. More manufacturing capacity doesn’t equal lower prices

More DRAM production can’t automatically restore the old pricing model. The location, accessibility, contractual status, and economics of existing capacity matter as much as the headline volume, particularly when existing supply agreements and higher-margin AI memory products continue to influence manufacturer behavior.

Indeed, this is why predictions that new fabs that generate additional memory capacity will quickly resolve the shortage are probably overly optimistic. Much of the new production coming online is intended to support AI-oriented memory technologies or is already committed to strategic customers. Even as capacity expands, manufacturers remain incentivized to prioritize the most profitable products and customers. As a result, enterprises should prepare for an environment in which supply gradually improves while elevated pricing persists.

More DRAM production can’t automatically restore the old pricing model.

3. Why moving to public clouds won’t solve the problem

Some organizations may believe that they can circumvent infrastructure shortages by shifting workloads to public cloud providers. But the truth is, public cloud providers are hardly immune from these structural changes.

In reality, hyperscalers purchase memory and infrastructure from the same supply chain as enterprises, and often secure capacity first through long-term supply agreements. As memory prices rise and AI demand absorbs a growing share of available capacity, those costs increasingly flow through to cloud services as well. Organizations may change how they consume infrastructure, but they cannot escape the underlying economics driving the shortage. Cloud can still be the right platform for some workloads, but it should not be viewed as a workaround for constrained memory supply and rising infrastructure costs.

4. Supply-chain competence is a competitive edge

In a constrained market, the ability to navigate the IT supply chain becomes an increasingly essential skill and competitive advantage. It allows organizations to act earlier, distinguish durable signals from market noise, and protect critical modernization programs from avoidable disruption. This increasingly requires disciplines more commonly associated with commodity procurement and financial planning than traditional infrastructure purchasing.

Organizations that understand market trends and maintain strong vendor relationships will be better positioned to secure capacity as needed. Conversely, companies that continue to rely on reactive purchasing may find themselves delaying projects, paying significantly higher prices, or accepting alternative configurations due to unavailable components. Managing supply-chain risk is therefore becoming as important as managing technology risk itself.

The ability to navigate the IT supply chain becomes an increasingly essential skill and competitive advantage.

5. Buying nothing is not a strategy

Demand collapse is not the same as demand resolution. Delaying projects may reduce near-term purchases, but pushing out purchases does not eliminate the underlying business, security, and modernization needs of an organization.

In fact, postponing equipment upgrades layers on additional operational challenges over time. Aging infrastructure can consume more power, degrade performance, and expose organizations to security and reliability concerns. While some purchase delays may be appropriate, enterprises shouldn’t assume the market will return to pre-shortage conditions. Prioritizing business-critical investments, optimizing existing assets, and developing a phased modernization roadmap that accounts for continued supply constraints.

Procure strategically to navigate the memory shortage

To adjust to the changing operational model, enterprises should move away from just-in-time infrastructure purchasing toward proactive procurement. For many organizations, the risk of waiting for equipment supply to ease is increasingly viewed as greater than the risk of purchasing early.

Indeed, some 27% of respondents are seeking alternate vendors to secure supply, according to Omdia research.
For those with budget and forecasting ability, they may opt to order 6 to 12 months ahead of required deployment dates. Organizations should also consider securing long-term agreements (LTAs) with memory distributors to guarantee supply allocations. The goal is not simply to buy earlier, but to reduce the risk associated with price volatility and supply disruption.

Companies may opt to work with vendors that can help cost-optimize IT infrastructure. Vendors like HPE have declared a commitment to provide price caps on compute and storage. Organizations IT and procurement leaders should work to identify vendors and relationships that reduce costs.

Companies should recognize that IT procurement discussions are becoming more central to company competitiveness. Companies that want to get ahead of the shortage should recognize that procurement is now not just a “buy” decision but a strategic and operational one.

Accordingly, chief information officers should work with C-suite counterparts in finance and operations to understand the new normal that defines the supply chain and navigate through it. Given that these conditions will persist for the next several years, organizations should work across the C-suite to master procurement planning.

Companies should recognize that IT procurement discussions are becoming more central to company competitiveness.

How SHI can help: Partnerships to navigate the shortage

To be sure, the IT memory shortage is not a short-term supply problem. It’s a persistent, structural issue that requires a new approach. Companies need to make strategic decisions about budgeting, refresh timing, workload priorities, lifecycle management, and AI infrastructure investments, all while working with capacity-constrained environments and forecasting for the future.

It’s not an easy task, with many organizations seeking guidance on how to navigate today’s supply chain. SHI guides organizations on which decisions to make when. Rather than simply sourcing hardware, SHI focuses on determining where to accelerate purchases, where to extend asset life, and where operational improvements can reduce demand for scarce resources.

The organizations that navigate the memory shortage most successfully will be those that combine technology changes with strategic planning and strong partnerships to create resilient architecture and strategic IT procurement.

NEXT STEPS:

Read about the storage squeeze.

Read the full Omdia/SHI research report here.

Learn more about modern data center solutions at SHI.