As AI moved away from using normal GPUs, PC builders celebrated the expected return to normal pricing and availability. But now the AI boom is putting pressure on storage and memory, sending the prices surging.

Why exactly has the AI boom led to a memory and storage shortage? And how long is it expected to last?

A Need for Chips

The AI race has been heating up, with virtually every tech giant investing heavily in AI. The problem is that normal computers aren’t very good at handling the kind of workloads AI models need, so we are seeing a massive surge in new datacenters specialized for AI.

And as you might expect, this has translated into a massive demand for the hardware needed as well. That includes motherboards, processors, storage, and memory. The stickler here is the last part, as the kind of memory that these data centers require is different from the standard RAM modules being made for the consumer market, or even most embedded PCs.

What AI needs is high-bandwidth memory (HBM), which requires a very different process of production. So the companies that manufacture PC memory need to choose – are they going to prioritize standard memory modules or shift gears to producing HBM, with guaranteed large-scale orders? And they have chosen AI.

Sold Capacity

The way these large enterprise orders work is a bit different from normal commerce. Instead of the company producing modules and then finding a customer, big firms buy up the RAM manufacturer’s wafer “capacity” itself.

This means that big tech can (and has) book out a manufacturer’s entire output for months, locking the factory into producing wafers for its own usage. The manufacturers like this as it cuts down on any sales uncertainties, but this also means that the projected supply for the standard market dries up to zero.

This is exactly the situation right now, with most memory production booked out for 2026, sending the prices of what RAM remains on the market sky high, since no new supply will be coming out to meet the demand. While you can still find stock in the market, it won’t be long before RAM sticks simply start running out entirely.

Why is it Affecting SSD Prices as Well?

While SSDs aren’t what the AI datacenters are after, the same manufacturing facilities produce the wafers for both memory modules and storage modules. This means that as their capacities get bought up to produce high-bandwidth memory, it cuts down the production of storage as well.

The impact has not been felt that sharply on storage prices yet, as the trickle-down is slower. But as the SSD manufacturers stop getting the NAND modules they need to assemble SSDs, the prices will shoot up further.

What About Other Manufacturers?

A natural question to pose in light of this memory crisis is why isn’t the production just ramping up to meet the demand? If AI datacenters are buying up manufacturer capacity, where are the competing manufacturers?

That’s when you realize that there are, in fact, none. A grand total of three companies dominate the memory market: Samsung, SK Hynix, and Micron. Of these, Micron is the only American company, and it has already announced that it is exiting the consumer memory market, and will focus purely on its enterprise customers.

That leaves only Samsung and SK Hynix, the two Korean companies, to cater to the market demand. And they have been trying to scale up their production, investing in new factories, but it will be a while before that shows fruit. It takes years for a single fab to come online, and the amount of capital involved means that they aren’t exactly being set up in high numbers.

So What’s the Outlook?

For now, the situation isn’t looking too good. The AI race is still in its early stages, and you can expect the industry to corner a lion’s share of the memory market for the foreseeable future.

Now, new facilities have been announced, both in the US and elsewhere. But the earliest these plants can start fabricating wafers is 2028, leaving us with at least two years of production crunch to deal with.

If your business is looking to upgrade its embedded PCs, there is no point in waiting. Prices will only keep going up, and delivery times will get delayed as the components get scarcer. Unless it is an upgrade you can put off for the next three years, the best time is right now.