The era of cheap technology could be over

Businesses are staring down an unprecedented shortage in components, with low memory stocks in particular pushing up prices across IT. Industry observers predict a fall in sales of everything from servers to smartphones.

The blame, at least in part, is laid at the feet of AI, with its seemingly insatiable demand for high-performance hardware.

AI projects are Hoovering up the available inventories of CPUs, GPUs, DRAM and storage, to run ever-larger, more powerful AI models. Manufacturers have responded by switching capacity from “commodity” DRAM and SSDs, to higher-end, higher margin components better suited to AI.

Some OEMs report that they have already sold all their capacity for 2026. Others have stopped making the components used in smartphones, PCs and servers altogether. Micron, for example, is set to close its Crucial consumer division. Others may well follow.

And, as component prices rise, sales of lower end PCs and notebooks look set to fall, in some cases dramatically. Analyst firm Gartner expects PC shipments to fall 10.4%, and smartphones sales to drop 8.4% during 2026.

Memories fade

Memory, or DRAM, has so far borne the brunt of price increases and supply shortages. But storage components, above all solid state storage, are not far behind.

Gartner predicts a 130% “surge” in combined DRAM and solid state drive (SSD) prices by the end of this year. And pressures are building beyond the consumer and personal device markets. Higher specification SSD and NVMe drives, used in servers and data centers are already in demand.

“AI‑driven demand continues to strain global memory supply chains, with DRAM and NAND flash, critical components for SSDs, in short supply. Leading hard‑drive manufacturers are reporting that their production capacity for 2026 is already sold out and prices are rising rapidly,” Jimmy Tam, CEO at data management company Peer Software, tells ITPro.

One reason is high-end storage is critical for AI model training. NVMe modules act as cache memory in AI vendors’ storage systems. These caches help storage arrays keep up with GPUs and in the AI model business, GPU utilization equals money.

But DRAM prices are pushing systems designers to make more use of cache instead of expensive memory. This adds to the demand for SLC and TLC NVMe modules used in tier-zero storage.

“High throughput and low latency NVMe flash demand increases as a palliative for prohibitive DRAM costs and availability,” says Max Mortillaro, chief research officer at analysts Osmium Data Group.

Then there is the ongoing demand from the “hyperscalers” for high-density, QLC NAND storage. “On the QLC front, demand for capacity is nowhere near stopping, especially for high-capacity drives, driven by hyperscalers,” he says.

If that is not enough, Osmium reports three-fold increases in the price of servers. This leaves CIOs, and the business services that depend on technology pipelines, facing unreliable supplies and higher costs.

Vendors agree.

“Strong component demand driven by tech titan AI buildouts has outstripped supply across the industry, dramatically increasing NAND, memory and CPU pricing. We expect that the industry, including Everpure, will see unpredictable component shortages, which could lead to extended lead times and potential shipment delays,” Everpure CEO Charlie Giancarlo told the financial markets earlier this year.

Other hardware vendors have made similar statements. NetApp, for example, describes the trend as an “unprecedented seismic shock for IT procurement strategies”.

Pandemic hangover

The storage capacity crisis is not, though, only a result of AI. Shortages are also the result of market cycles following the Covid pandemic.

In 2022, the markets saw what Scott Shadley, board director at the Storage Networking Industry Association, calls a “hoarding event”. This was followed by a downturn in the market, and an oversupply.

“This caused many companies to lower outputs. Then the ‘super’ cycle hit in late 2024 with the AI boom,” he says. “The need for all semiconductor-based products, DRAM, NAND, SSDs, and even GPUs and CPUs, went from average, to above average demand, to overly aggressive demand.”

Other factors, such as current geopolitical events, also threaten to disrupt supply chains, raise costs and lengthen lead times. This will force CIOs to make some hard choices.

“For infrastructure and data leaders, the question isn’t whether flash [storage] is valuable. It is,” Paul Speciale, CMO at object and hybrid cloud vendor Scality, tells ITPro.

“The question is whether your storage platform forces you into an all-flash posture, or whether it gives you architectural freedom to use each medium where it makes the most sense, while still delivering the performance and operational consistency modern workloads require.

“When markets are stable, it’s easy to treat storage media as a design preference. In a supply crunch, media becomes a constraint, and architectures that require one ‘perfect’ component will get exposed very quickly.”

Diversification, data cleansing and upcycling

Businesses, though, can offset some of the costs by looking at where they store data, what they store, and by delaying upgrades.

One way is to make more use of tiered storage, mixing NAND, hard drives and even tape. And organizations can also save costs by removing outdated or redundant data, as well as thinking harder about the data they capture and store. Enterprises might also keep hardware running longer, potentially using older storage arrays for less critical data.

“Organizations must recognize that placing data across different storage technologies is no longer theoretical, it is a practical necessity,” Valéry Guilleaume, CEO at data management firm Nodeum tells ITPro.

“This requires the ability to move, catalogue, and place data across heterogeneous storage systems. As data volumes continue to grow, organizations must remain open to integrating multiple storage technologies.”

Nelson Nahum, co-founder and CEO at Geyser Data, agrees. “High-speed memory and flash should be prioritized for active workloads such as AI training, analytics and production applications.

“Storing large volumes of inactive or archival data on these tiers can quickly drive up costs and exacerbate supply constraints,” he adds. “In short, the current supply pressures are encouraging organizations to rethink how they manage the full lifecycle of data.”

Storage shortages look set to continue to put pressure on CIO hardware budgets. But, in the medium term, this could yet benefit businesses, if the result is better data housekeeping.