PostHole
Compose
Login
The global RAM and SSD shortage crisis, explained
A global shortage is responsible for every electronics and computer manufacturer in the world — including Apple — paying twice as much for RAM and flash storage as it did in 2025, and 10 times more than it paid in 2020. Here's why there is little hope of that improving anytime soon.
Memory is in short supply globally — Image credit: SK Hynix
Apple has historically been able to closely control the cost of its components. Buying in huge numbers, from multiple suppliers has historically given an economy of scale that made Apple a sought-after customer for everything from display makers to storage vendors.
But that dynamic has changed. A global shortage of key components like memory and storage has seen the price of both skyrocket. Apple is far from the only company impacted.
Continue Reading on AppleInsider | Discuss on our Forums
News
The global RAM and SSD shortage crisis, explained
[Oliver Haslam's profile picture]
Fri Feb 27 2026, 12:31 PM EST
4 minute read
Memory is in short supply globally -- Image credit: SK Hynix
A global shortage is responsible for every electronics and computer manufacturer in the world — including Apple — paying twice as much for RAM and flash storage as it did in 2025, and 10 times more than it paid in 2020. Here's why there is little hope of that improving anytime soon.
Apple has historically been able to closely control the cost of its components. Buying in huge numbers, from multiple suppliers has historically given an economy of scale that made Apple a sought-after customer for everything from display makers to storage vendors.
But that dynamic has changed. A global shortage of key components like memory and storage has seen the price of both skyrocket. Apple is far from the only company impacted.
But the whys and wherefores of the shortage are about more than a simple case of shorter supply and increased demand. To understand why Apple is now paying more for key components, we have to first understand what those components are.
DRAM and NAND
You've probably seen the headlines already. Stories of DRAM and NAND shortages are everywhere right now, and both components are vital to everything from cars to phones.
DRAM, or Dynamic Random Access Memory, is exactly what it sounds like. It's the temporary memory that computers and similar devices use to temporarily store data.
NAND, the other component in short supply, is a type of flash storage that is commonly used in all kinds of things. Portable SSDs, memory cards, and every single Apple device shipping to consumers all use NAND storage.
The iPhone 18 Pro will cost Apple more to manufacture as NAND storage costs rise
And there's a similar issue with hard drives too, but that doesn't impact Apple's consumer goods so much. WD's CEO said during earnings that they were effectively sold out through 2026, and into 2027.
With that said, it's easy to see why a shortage of either is posing a problem. You can blame AI for that shortage.
No memory for you
You don't need to be the biggest AI user to know that it's everywhere right now. And some companies are betting big on it being even more ubiquitous than it already is.
Those companies, like OpenAI, Google, and others, need to build new AI data centers to cope with the demand that they expect (or hope) will come about. And those data centers need memory.
But that's an understatement. In reality, AI data centers need more memory than most.
Specifically, AI data centers use HBM, or High Bandwidth Memory. As the name suggests, it's faster than normal memory, like DRAM, so that it can keep up with the data demands of AI models.
These same data centers also need huge amounts of NAND storage. Not only do AI models require petabytes of training data, but they also need that data to be retrieved as quickly as possible — ruling out the hard disks that normally fill data centers.
Unfortunately, the companies that make the HBM and NAND that AI data centers need can't keep up with demand. But they have a solution.
The three major memory manufacturers — Samsung, SK Hynix, and Micron — have shifted their production capacity away from consumer DRAM and towards data center-grade HBM. Importantly, HBM is harder to make and has smaller yields than DRAM, so more silicon wafers are required to produce it.
The result is that the amount of DRAM being produced has shrunk, and will continue to do so.
It's a similar situation with NAND, too. Manufacturers are moving their manufacturing capacity to where the money is. Right now, that money comes from huge contracts signed with AI companies.
Scaling up production isn't an immediate option
When companies run out of production runway, building more factories is usually an option. But the current situation is unique in that regard.
A key issue is time. It would take years for these companies to bring a new factory or production line online, even if it wanted to. A single stick of DRAM can take up to four months to produce from beginning to end, adding further lead time.
Producing DRAM is a multi-step process, each taking considerable time. Creating the silicon wafer from which DRAM is made can take months along. Slicing that wafer into individual chips for testing before sale is equally time consuming.
The time needed to spin up additional capacity means supply normally lags demand by 12 to 18 months. And with analysts suggesting that 2026 into 2027 will experience the peak of the shortage, it's already too late.
[Examples of older DRAM modules]
Examples of older DRAM modules — image credit: Samsung
Another reason is that memory companies are still reeling from a 2023 oversupply of consumer DRAM that saw prices plummet. Those same companies have chosen not to scale up production of DRAM at the risk of a repeat performance in the future. All available capacity will be used to produce HBM instead.
Whether that suggests memory makers are concerned that an AI bubble is about to burst is a matter for debate. But there's logic to that conclusion — scaled-up manufacturing now could leave unused factories and stores of surplus memory if the AI bubble bursts.
Instead, for years, companies have chosen to keep supply short rather than flood the market with more memory. That approach has the added benefit for manufacturers of also keeping prices high for both mainstream and AI customers.
Hope for the future
Despite all the doom and gloom, there is still some reason for hope. Just don't expect any good news until 2027 or, more likely, 2028.
"A significant shortage of memory products across the board is expected to continue for the time being," Kim Jaejune, a Samsung memory chip business executive, reportedly told analysts on a post-earnings call in January.
In a leaked internal report, SK Hynix expects DRAM supply to remain constrained through 2028. But both companies do have plans for the future.
Samsung, Micron, and SK Hynix were already in the process of building new facilities. They won't be fully operational until the middle of 2027 at the earliest.