What Is DRAM?

TL;DR: DRAM has been all over the news lately, so what is it? DRAM (dynamic random-access memory) is the working memory inside your computer, phone and game console. It’s scarce right now because the chips that power AI use a stacked version of it that is far more demanding for factories to make. So every gigabyte made for AI uses up the capacity that could have made about three for your PC or phone. Sam Altman’s remark that it’s “quite terrible” for everyone to hold so much of it in their own PC has been a recent headline- let’s break down what he is saying.

What DRAM is

Think of DRAM as a whiteboard in a meeting room, and your storage as a filing cabinet.

The filing cabinet (an SSD or flash storage) holds everything you own. It keeps it when the power is off, but it’s slower to pull from. The whiteboard (DRAM) only holds what the team is working on right now, and anyone can read or rewrite it instantly. When the meeting ends, the board is erased.

Why does the board get erased? DRAM stores each bit of data as a tiny electrical charge that leaks away within a fraction of a second. The chip has to keep refreshing it, like someone constantly re-tracing fading marker lines, which is why it needs power and why everything disappears when the power goes out. That’s the “dynamic” in dynamic random-access memory.

That’s why opening 40 browser tabs slows a computer down. You’ve run out of whiteboard, not filing cabinet storage.

“RAM” is the general name for that working memory. DRAM, short for dynamic random-access memory, is the type of chip most of it is made from, including DDR5, the current mainstream generation of PC memory. It comes as small circuit boards called sticks that you slot into a PC. In everyday conversation people use the two words almost interchangeably. DRAM is also inside phones, game consoles and the servers in data centers, which matters for the rest of this post.

Why there’s a shortage

The same factories that make the memory in your laptop are being pulled toward making memory for AI chips.

Samsung, SK hynix and Micron make most of the world’s DRAM, according to Tom’s Hardware and Wikipedia. AI chips use a stacked version called HBM (high-bandwidth memory), which layers several DRAM chips on top of each other.

A Micron executive says HBM uses about 3x the factory capacity per gigabyte that standard DRAM does. So every gigabyte made for an AI chip uses capacity that could have made about three for a PC.

Then there’s demand. In October 2025, OpenAI announced letters of intent (early-stage agreements, not final contracts) with Samsung and SK hynix for its Stargate project. Reports put it at up to 900,000 wafers a month (a wafer is the silicon disc that chips are cut from), which outlets estimate at roughly 35-40% of global DRAM wafer capacity.

The supply side is tight too. Micron has exited consumer memory (it sold memory sticks and SSDs to everyday shoppers under the brand name Crucial), and SK hynix told investors its memory capacity was essentially sold out for 2026.

Because AI memory (HBM) takes about 3x the factory capacity per gigabyte and makers are shifting capacity toward it, less remains for the memory in your devices.

What it did to prices

Item	Price change	Source
Corsair 32GB DDR5 kit	$134.99 in Sept 2025 to over $420 by early Dec 2025	Tom’s Hardware
Samsung 32GB DDR5 module	$149 in Sept 2025 to $239	Network World
DRAM prices overall, 2025	Up about 172% (reported)	Wikipedia

How long will it last?

Most forecasts I found say consumer relief around 2028. One consulting analysis cited on Wikipedia says the shortage lasts until at least 2030. These forecasts don’t agree with each other, and that’s the point: memory is hard to forecast. New factories take years to build (Samsung’s new Pyeongtaek line, for example, won’t start mass production until 2028), and nobody knows how fast AI demand will grow.

What Sam Altman said

Sam Altman, the CEO of OpenAI (the company behind ChatGPT), recently said it’s “quite terrible” for everyone to hold a share of the world’s limited DRAM in their own computer. It came up in a conversation about supply chains.

For context, Altman co-founded OpenAI in 2015 and has been its CEO since 2019. OpenAI makes ChatGPT. Before that, he was president of the startup accelerator Y Combinator from 2014 to 2019.

The remark is from the Cheeky Pint interview with John Collison, published October 29, 2026. The quotes below come from the transcript itself. Collison had been asking why memory makers don’t just flood the market when prices are high. Altman said the supply chain is “super non-obvious” and used local AI as his example: people assume open-source models running on their own machines mean we won’t need data centers.

His point was about allocation. In his words, it’s “quite terrible” to have everybody hold a bunch of the world’s limited supply of DRAM in their own computer, being used “5% of the time.” He added that “you want maximum utilization in the cloud” in a supply-constrained environment.

Why people are upset

The reaction was strongly negative, and part of the criticism is fair.

The main complaint is about timing. People are paying far more for memory right now, and OpenAI’s own reported supply deals are part of the demand behind that. PC Gamer’s writer said it “still feels a bit rich” for Altman to complain about a scarcity OpenAI helped create. The second complaint is about ownership. Your PC’s memory is yours, and “global allocation” treats it like a shared resource.

Altman’s side has a real argument too. Shared infrastructure usually runs at higher utilization than personal devices, which is the basic logic of cloud computing. He was describing what’s efficient when supply is tight, and he didn’t say people shouldn’t own computers.

The weak spot in his argument is what the utilization argument leaves out. Running things locally can matter for privacy, speed and simply owning your tools.

What this means for you

If you’re buying a PC, phone or console, expect memory to cost more for a while.

Memory is now a bigger share of what a device costs to build. HP said memory went from roughly 15-18% to 35% of its PC build costs in a single quarter. This is why prices are rising across devices and not only on memory sticks.

Even the more hopeful forecasts put relief around 2028, so a quick price drop isn’t something to count on.

Use AI to learn something new like this

Confession: I was interested in this blog post topic because I wanted to learn what the fuss was all about over DRAM myself.

A news cycle like this one leaves a lot of people nodding along to a term nobody ever explained to them. AI is good at fixing that, because you can ask it to explain the term in plain English, use an analogy, and go as deep as you want. The part that needs your attention is checking what it tells you.

Next time you hit a term you don’t know, have AI explain it, then check the claims that matter.

Here’s a prompt you can copy and reuse. Swap in your own term:

I just ran into a term I don't understand: [TERM]. Explain it to me like a smart friend who isn't an expert in this field.

1. Start with one everyday analogy, then tell me where the analogy breaks down.
2. Define it in two sentences, then name the closest term people confuse it with and how they differ.
3. Tell me why it matters right now. For each claim, label it as established fact, a reported claim, or a forecast.
4. List what you're unsure about and what I should check in a primary source.
5. Ask me three questions to test whether I understood it.

When we're done, turn this method into a reusable skill I can run on any new term, with the steps above as the instructions.

FAQ

Sources

Checked November 1, 2026.

• Cheeky Pint transcript: Why Sam Altman is against a single superintelligence
• Tom’s Hardware: Altman on PC RAM
• PC Gamer: Altman on DRAM in your computer
• Yahoo Finance (Moneywise): Altman and the 40% supply claim
• Tom’s Hardware: HBM is eating your RAM
• Network World: Samsung warns of memory shortages
• Wikipedia: 2024-present global memory supply shortage
• Wikipedia: Sam Altman