Why Are We Having an SSD Shortage?

Middle Ages stained glass window image of a skeleton strangling a manToday’s HDD shortage, driven by the current AI spending surge, has created a new opportunity for SSDs.  Suddenly, QLC SSDs, which had previously been used almost exclusively in consumer PCs, are now being purchased in high volume by hyperscale datacenters for two reasons: To add a new layer to the Memory/Storage Hierarchy, and to make up for HDDs that have recently become unavailable.  Let’s look at those two situations to get a better understanding.

But first, an aside: For any who don’t know a QLC from a TLC or MLC SSD, those terms indicate how many bits are stored in a single flash memory cell.  The cell size doesn’t change, but in MLC it’s used to store two bits, in TLC it holds three, and QLC stores four.  That translates into commensurate savings.  On the downside, every time a bit is added, the SSD slows down and becomes much more sensitive to wear.

Ignoring the wear, the combination of a slower device with a lower price gives you the recipe for a good fit in the storage hierarchy: Each layer needs to be cheaper than the next-faster layer, and faster than the next-cheaper layer.  Thanks to this, QLC SSDs fit between TLC SSDs and HDDs.

Facebook’s Finding

In 2024’s OCP Summit Facebook/Meta described an experiment in which researchers added a layer of QLC SSDs between their TLC SSD layer and their hard drives.  They found that they required fewer resources in their datacenters, and this more than offset the cost of adding the QLC layer.  This was particularly true in AI applications, which is the application driving most of today’s datacenter spending.  It didn’t take too long before all of the other hyperscalers followed suit, allowing them to reduce their costs, while SSD makers were profitably enjoying the benefits a newly-constrained market caused by a new demand surge.

The AI-Driven HDD Shortage

The second ingredient of today’s SSD demand surge comes from HDDs.  It seems that the HDD market has become very constrained by the same skyrocketing AI capital spending that is fueling the new QLC SSD demand described above and that has also driven DRAM into a serious shortage, thanks to the use of HBM DRAM in GPUs and TPUs which are widely used in AI.  Hyperscalers who had hoped to increase their HDD purchases were turned away.  Where were they to go to fill this need?

Well it ends up that the hyperscalers were desperate enough to spend more money than they originally wanted to spend to purchase QLC SSDs to make up for the HDD shortage.  This added significantly to SSD demand, and has teamed up with the new QLC layer to create phenomenal growth in SSD demand.  Today SSD makers are finding it difficult to keep up with demand.  This is quite a change for NAND flash chip makers who had suffered through two long years of weak demand and low prices.  Now they’re doing everything they can to keep up.

In the mean time, the shortage is causing purchasers to compete for whatever SSDs they can get, and that always drives a price increase.

AI CapEx Growth Keeps Increasing

Like many other analysts, Objective Analysis has been having trouble keeping up with AI’s growth.  By the time we publish a forecast we find that it’s outdated, and that our numbers must be increased.  The same is true of the companies making anything used by AI, not only HDDs and SSDs, but also GPUs (Nvidia), foundry services (TSMC), and other parts of the supply chain – these suppliers plan capacity to meet demand, and then their customers hand them a new plan which is invariably higher.  It’s a good problem to have, but it’s problematic nonetheless.

The chart below graphs hyperscaler CapEx as a percent of revenue.  When this number increases above 15% it means that the hyperscalers are spending in anticipation of demand that has not yet materialized.  It’s easy to see that this has been the case since early 2024.

Line chart showing quarterly capital spending as a percentage of revenues for major hyperscalers. from 1Q21 to 3Q25. The line varies between 12% to 15% until 2024, when it starts a steep rise to 24% in 2Q25, then drops a little to 23% in 3Q25.

This large amount of spending without commensurate increases in revenue gives us cause to worry, though.  It’s quite possible that these same hyperscalers will pause this spending and wait for demand to catch up with their AI capacity, and when that happens an oversupply and price collapse are inevitable.

For now, though, SSD makers can enjoy the strong demand they see and the improved profits they are enjoying.

Understanding the Future

At Objective Analysis we focus on understanding the future and bringing likely scenarios to our clients to help them develop plans to benefit from tomorrow’s markets.  If you would like to see your company benefit from our research and analysis, please contact us to discuss ways that we can work together to provide your company with its strongest competitive advantage.

2 thoughts on “Why Are We Having an SSD Shortage?”

  1. A lot of us question when this super cycle will relax or even collapse. No end in sight right now but appreciate your insight into this surge in demand. ” It’s quite possible that these same hyperscalers will pause this spending and wait for demand to catch up with their AI capacity, and when that happens an oversupply and price collapse are inevitable”. For those us close to the supply chain it is worrisome indeed.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Privacy Overview

This website uses cookies so that we can provide you with the best user experience possible. Cookie information is stored in your browser and performs functions such as recognising you when you return to our website and helping our team to understand which sections of the website you find most interesting and useful.