Is AI The Sole Cause Of Pressure On The World’s Memory Chip Market? Not Exactly

Although the AI industry is undoubtedly a massive driver of demand, this is only a partial cause of the complex market dynamics.

The global markets for memory chips, especially DRAM (Dynamic Random-Access Memory) and NAND flash, are currently experiencing a phase of significant price increases.

Officially, this development is primarily justified by the explosive demand from operators of AI applications, who are buying these essential components in enormous quantities.

Although the AI industry is undoubtedly a massive driver of demand, this is only a partial cause of the complex market dynamics.

The Shift From Commodity Products To HBM

The actual impact of AI demand is illustrated by current reports:

OpenAI is reportedly claiming a massive capacity of an estimated 900,000 wafers for its ambitious "Stargate" project, a potential next-generation data center. This represents almost 40 percent of the total global DRAM production capacity. So far, there is no formal, contractually secured purchase guarantee. However – or perhaps because of this – it brings additional uncertainty into the market.

Micron has sold almost all of its production capacities for HBM (High Bandwidth Memory), the high-performance memory primarily used in AI accelerators, up to and including 2026.

The demand in the NAND flash segment is also immense. Samsung's upcoming V9 NAND product has already been sold almost completely to strategic customers before the official start of production.

The immediate impact of this shortage and price increase affects a wide range of hardware manufacturers. In particular, manufacturers of standard computers, conventional data center servers, laptops and the entire consumer electronics industry are confronted with rising procurement costs. Ultimately, these additional costs will be passed on to end consumers in the form of higher terminal equipment prices.

An Oligopoly Dominates The Storage Market

The market for DRAM and NAND flash is characterized by a high concentration. It is effectively dominated by an oligopoly of only three main players: Micron Technology, SK Hynix and Samsung Electronics. The strategic decisions and production priorities of these three giants largely determine global prices and availability.

A decisive structural factor is the strategic realignment of the manufacturers. HBM, which is essential for the lucrative AI infrastructure, offers significantly higher profit margins than traditional memory products such as DDR memory (so-called commodity products). As a direct consequence, the three dominant manufacturers are massively shifting their production capacities and resources towards HBM. This restructuring is at the expense of the production of standard DDR memory, which further exacerbates the supply situation for conventional computers and servers. Micron is even planning to give up its entire consumer division, Crucial. What sounds like good news for enterprise could quickly turn out to be a drawback: Crucial products are also installed in PCs and laptops of corporate customers.

Hesitant Capacity Expansion After The Semiconductor Crisis

The current situation is also a late consequence of the global semiconductor crisis, triggered by the COVID pandemic. During this crisis, production capacities were reduced in many places. Since then, the necessary reconstruction and expansion of capacities have been slow and hesitant. In addition, investment decisions are further slowed down by the fundamental uncertainty regarding the sustainability of the current AI demand – the fear of a potential AI bubble.

Technological Change And Production Bottlenecks

The shortage is further exacerbated by the natural product lifecycle of the storage types. The production of DDR4 memory will be phased out in the major factories of the three main suppliers to free up capacity for newer, higher-margin technologies. While smaller chipmakers are stepping into the breach to fill the gap, they still need time to optimize their production processes and reach full capacity. This dynamic is not new; a similar development was already evident in the transition from DDR3 to DDR4.

High Investment Barriers And Long Lead Times

The construction of new production facilities (fabs) is an extremely capital-intensive and time-consuming undertaking. It typically takes several years from the laying of the foundation stone to volume production. In view of the current geopolitical and economic uncertainties, this high investment risk poses a significant additional challenge that massively slows down the rapid response to increased global demand.

There is also a backlog of hard disk drives (HDDs) that is estimated to be around two years. This significant backlog is a direct result of the massive increase in demand for storage capacity, triggered in particular by the boom in artificial intelligence (AI) and the associated data-intensive applications.

The development and operation of large language models (LLMs), neural networks, and other AI technologies require huge amounts of training and operational data. Hyperscalers and cloud providers that provide the infrastructure for these AI applications are forced to expand their storage capacities exponentially. Although solid-state drives (SSDs) are often preferred in terms of speed and performance, HDDs remain the primary choice for storing large amounts of cold or nearline data due to their significantly cheaper price per terabyte, which is essential for AI training sets.

This sudden and massive increase in demand has put a lot of pressure on the production capacities of the leading HDD manufacturers, such as Seagate, Western Digital and Toshiba. Supply chains for critical components, including platters and read/write heads, are strained, preventing manufacturers from ramping up production fast enough.

The two-year order backlog points to a profound shift in the storage market. It signals not only a short-term bottleneck, but also a longer-term structural challenge that could potentially affect the speed of global AI growth. Companies that rely on large HDD shipments – from data centers to big data analytics companies to supercomputing facilities – must now plan for significant delays in expanding their infrastructure. This bottleneck is also expected to drive up the prices of storage capacity in the spot market, further increasing the total cost of ownership for AI-centric projects.

Moment! AI and HDD? Wasn't There Something About ‘The Faster, The Better’?

At first glance, it may seem paradoxical that AI development relies on HDD technology, which is slower compared to SSDs. However, the explanation lies in the way storage is used in the AI infrastructure:

• Storage space to store training datasets: AI models, especially large language models (LLMs) and complex neural networks, require huge amounts of training data (terabytes, often petabytes). This data does not have to be constantly retrieved at top speed; their main requirement is cost-effective and dense storage. HDDs offer the best price-per-terabyte ratio.
• Storage space for the generation (acquisition) of training datasets: The process of collecting, pre-processing and storing new, potential training data – whether through web crawling, sensor technology or simulations – also generates enormous volumes of data that must be archived primarily.

The Role of QLC-NAND in the AI Age

In parallel with the HDD demand, AI development is also driving the proliferation of QLC (Quad-Level Cell) NAND flash memory. QLC stores four bits per cell, providing the highest memory density among common NAND technologies (as opposed to TLC/Triple-Level Cell or MLC/Multi-Level Cell).

The need for large, relatively inexpensive SSDs for faster data access (e.g. for hot storage, caching or parts of the training pipelines) is driving up the quantities of QLC storage massively.

Mass production leads to increased investment in the research and development of controller firmware and associated algorithms. This is crucial because QLC's higher data density tends to result in lower durability (fewer write cycles) and more complex error correction. The continuous improvement of controller technology thus improves the quality, durability, and performance of QLC SSDs (also SATA SSDs).

The Paradox

AI-induced HDD shortage could boost QLC storage if it wasn't for the problem with semiconductors, with the cat biting its own tail right now.

But the production bottlenecks are not the biggest problem for QLC memory.

Although QLC technology was originally positioned as a more cost-effective alternative, the sharp increase in demand, especially from the AI sector, is leading to a paradoxical development: the increased demand for QLC NAND is causing QLC memory to become more expensive, contrary to initial expectations. Vendors can push through higher prices due to the critical role of storage density in AI infrastructure. This indicates a market narrowing in the high-density memory space, affecting both HDDs and QLC-NAND.

Acting With Foresight Instead Of Panic

In the face of volatile markets and potential supply bottlenecks, a prudent and strategic approach to the procurement of IT components is essential. We vehemently advise against panicked hoarding of hardware that far exceeds current demand. Such behavior distorts prices, places an unnecessary burden on working capital, and can lead to the purchase of inappropriate technology.

Forward-looking planning is essential. Reactive procurement management is a significant competitive disadvantage in today's business world. Organizations that accurately anticipate their current and future needs – based on medium- to long-term growth targets, the introduction of new projects or the natural life cycle of existing systems – gain significant advantages. This strategic foresight allows for more efficient capacity planning and early sourcing of resources. By placing orders with longer lead times or using framework agreements, companies can cushion price fluctuations and ensure the availability of critical components.

Investment in quality pays off. The short-term savings from purchasing consumer-grade or low-quality components are often offset by higher total cost of ownership (TCO). Higher-quality models have a longer service life and significantly higher reliability. If you initially invest in high-quality ECC (Error-Correcting Code) memory or enterprise-grade HDD or SSD (with a higher DWPD rate – Drive Writes Per Day), you drastically reduce the likelihood of system failure and have to replace defective modules less often. This not only minimizes the direct cost of spare parts, but also the cost of downtime, technician hours, and data loss.

Leverage existing and alternative technologies efficiently. The constant hunt for the very latest technology is not always economically sensible or technically necessary. For many areas of application, the latest, fastest generation is not necessarily required especially for applications with moderate performance requirements or well-optimized workloads, older generations can still provide excellent service.

Organizations that analyze and know their applications and user behavior in detail can often fall back on models from previous generations. This relieves the supply chain of the latest products and often offers significant cost advantages. Another sustainable and cost-efficient alternative is the use of refurbished modules from certified suppliers. These components are professionally tested and often come with warranties, providing a sensible bridge between cost savings and reliability.

This article originally appeared on MES Computing’s sister site Computing Deutschland.