What’s causing the memory shortage?

The memory industry is notorious for feasts and famines, but the current shortage is particularly bad. Why is that?

The memory industry has faced shortage issues in the past, but some of those were at least partially due to the changing of a memory format. When the industry went from an old style of memory like DDR4 to DDR5, for example, the change was rather abrupt and the execution was nowhere near as smooth as the migration to a new generation of CPUs or GPUs.

Right now, the industry is suffering the worst memory shortage in history, and that’s with three core suppliers: Micron Technology, SK Hynix, and Samsung. TrendForce, a Taipei-based market researcher that specializes in the memory market, recently said it expects average DRAM memory prices to rise between 50% and 55% this quarter compared to the fourth quarter of 2025. Samsung recently issued a similar warning.

So what caused this? Two letters: AI. The rush to build AI-oriented data centers has resulted in virtually all of the memory supply being consumed by data centers. AI requires massive amounts of memory to process its gigantic data sets. A traditional server would usually come with 32 GB to 64 GB of memory, while AI servers have 128 GB or more.

“This sort of unprecedented buildout caught the industry a little bit by surprise,” said Tom Mainelli, group vice president of device and consumer research at IDC. “Memory is typically a sort of boom-and-bust component industry, and right now it’s booming for the memory guys, and I don’t think they’re in a huge rush to try to solve the issue, even if they could.”

They are not in a rush to solve the issue because they’re making money hand over fist. Why increase the supply and dilute the price? And even if they wanted to, Mainelli said it would take 12 to 18 months, if not longer, to increase capacity.

There are other factors at play here, too, of course. The industry is in a transition period between DDR4 and DDR5, as DDR5 comes online and DDR4 fades away. These transitions to a new memory format are never quick or easy, and it usually take years to make a full shift.

There has also been increased demand from both client and server sides. With Microsoft ending support for Windows 10, a whole lot of laptops are being replaced with Windows 11 systems, and new laptops come with DDR5 memory — the same memory used in an AI server.

In addition to the big three memory vendors, there are some smaller Chinese vendors as well, but Mainelli thinks they are too small “to move the needle.”

Another element affecting the supply of memory is the explosion in the use of high bandwidth memory, or HBM. In recent years, with the advent of GPU accelerators, HBM demand has skyrocketed. This memory sits on the GPU card right next to the GPU itself, as opposed to standard memory sticks that sit on the motherboard              

What most people don’t realize is that it takes more memory wafers to build HBM than it takes to build a byte of DDR, says Jim Handy, president of Objective Analysis. So wafers going to HBM for GPUs are wafers not being used to make regular memory. “Because of the fact that each HBM bite takes about three times as much silicon as DDR, then what that means is that it takes three times as many wafers,” Handy said.

So if HBM is eating up 10% of the memory business, then that means HBM takes away 30% of the wafers, so it ends up having a pretty massive impact.

Handy said that for the longest time, memory manufacturers got away with doing a die shrink to increase capacity, so with smaller process nodes, they were able to get more memory out of a single wafer. For that reason, they didn’t expand their wafer capacity — they just got away with increased memory per wafer.

Now, the memory makers have to actually expand capacity. “It’s like, oh, wow, we have to put up new fabs. We have to buy the equipment for those fabs and all that stuff. And there are lead times to getting the equipment, equipment you usually have to order a year or two ahead of time,” Handy said.

It’s hard for the memory manufacturers to predict their needs in the future, because the needs change every quarter, he said. “It’s hard to keep track of something that’s moving that fast.”

Both Handy and Mainelli expect the shortages to last at least into next year.

The analysts agree on several points. For starters, tariffs are not playing a part in the run up in prices, even though two of the three major manufacturers — SK Hynix and Samsung — are Korean companies.

Something else that they agree on is that OEMs, at least for now, are absorbing the increasing price and not passing it on to customers. However, that’s subject to change if the prices keep going up.

“To date, we’ve not heard various vendors talking about increasing prices, but we’ve not seen those price increases hit yet, because most of the systems that are shipped into the channel and that are selling right now were shipped before the dramatic price increases hit,” said Mainelli.

“What’s likely to happen, from a market perspective, is we’ll see the market grow less in ’26 than we had anticipated, but ASPs are likely to stay or increase. So, revenues overall may not look too bad, but from a unit volume, that’s likely going to be impacted as prices go up,” he said.

Finally, the two analysts agree that if the often-rumored AI bubble burst actually happens and construction comes to a stop, then demand will stop and that will free up supply.

“If you decide that you’re going to spend before you have the demand [for AI], then you bet that there’s going to be a lot of AI demand, so you end up increasing your capex as a percent of revenue. And that’s what these guys are doing. If investors complain because it is going to impact what their return is to investors, then eventually they’ll take their foot off the gas, and then that will cause prices to collapse,” said Handy.

“We’ll be watching very closely to look at all the hyperscalers and others that are building and leveraging all this RAM, connecting it to all these GPUs in the data center, to see if there’s any indication they might slow down. If they were to slow down, then the memory situation could clear up relatively quickly. But right now, in mid-January, there’s not really any indication of that happening,” said Mainelli.