ChatGPT has reportedly bought over 40 percent of all memory capacity in 2026, and now CEO Sam Altman says the DRAM sitting in your home PC — used, in his words, 5% of the time — would serve the world better inside a data center. The comment, made this week on the Cheeky Pint YouTube channel alongside Stripe CEO John Collison, has reignited a fight over who gets the world’s scarce RAM, and MakeUseOf reported that Altman walked the remark back days later by renaming himself “RAM Altman” on X.
Altman’s argument is about allocation efficiency, not theft. In a supply-constrained market, he said, maximum utilization belongs in the cloud.
“From a global allocation perspective, actually, it’s quite terrible to have everybody have a bunch of the world’s limited supply of DRAM sitting in their own computer, and being used 5% of the time. Like, you want maximum utilization in the cloud if you’re trying to deal with this very supply constrained environment.”
That framing lands badly for a simple reason: the company Altman runs is a core part of why the constraint exists.
The 40 Percent Problem Sitting in the Supply Chain
AI training and inference clusters are devouring memory at a scale consumer hardware never has. OpenMarkets puts ChatGPT’s share of all 2026 capacity above 40 percent, and separate figures put the company’s intake at roughly 900,000 memory wafers per month — millions of terabytes monthly, just for ChatGPT alone. Hyperscale AI data centers broadly have locked up forward capacity with manufacturers, which is why new entrants face a wall of existing agreements even before tariffs enter the picture.
Consumers feel the same squeeze at the retail shelf. PC Part Picker’s memory charts show every DDR5 category rising more than 400 percent over 18 months. Read that again: 400+ percent in a year and a half. Anyone trying to upgrade a laptop or replace a dead stick is paying for the data center boom.
| Figure | Value | Source |
|---|---|---|
| ChatGPT share of 2026 memory capacity | Over 40% | OpenMarkets |
| Reported ChatGPT wafer intake | ~900,000 wafers/month | Industry figures cited in reporting |
| DDR5 retail price increase | 400%+ in 18 months | PC Part Picker charts |
| US import tariff on CMXT memory | 50% | Trade reporting |
Why the “Cloud Is More Efficient” Argument Doesn’t Land
The logic underneath Altman’s remark is that if all memory flowed to AI data centers and compute was leased back to users, utilization would rise and costs would fall. As engineering, that’s defensible.
Sam Altman Questions PC RAM Efficiency Amid AI Memory Supply Pressure OpenAI CEO Sam Altman has raised a provocative question about scarce computer memory: Is DRAM better used in personal computers or shared across cloud data centers? Speaking on the Cheeky Pint podcast… pic.twitter.com/vkCWdNyqWa
Elon Musk piled on publicly, and Altman responded by changing his X display name to RAM Altman and posting a series of corrections. The retraction reframed the original comment as illustrative — a call for the world to build more memory, not a claim that private ownership is wasteful.
I actually do want the world to have tons of memory and people to put as much as they want in their computers, and it is f****** annoying that this was taken out of context.
But the broader point he made in the same conversation holds up better than the memory-specific one: compute is scarce and expensive now, and, like most major technology shifts, that changes as capacity comes online. New memory makers are expected to flood the market. That is good news for consumers — and also good news for Altman, since cheaper DRAM helps the data centers buying it by the wafer.
What This Means for Anyone Buying a PC Right Now
For the person holding a laptop with 8GB of soldered memory, the debate is academic. Prices are what they are until new fabrication capacity actually ships, and even then, existing supply agreements mean hyperscalers get first call. ChangXin Memory Technologies (CMXT) is the name to watch — a tariff exemption on its current 50 percent US import duty could let it undercut incumbents — but Altman and his peers almost certainly already hold contracts on that future output. The uncomfortable arithmetic is that the same company arguing your home RAM is underused is positioned to buy the replacement supply first.
The Open Question Nobody Has Answered
Altman says RAM prices aren’t coming down soon no matter what he hopes, and the market agrees with him. What remains unanswered is who decides the allocation when supply does finally expand: the households whose idle sticks started this conversation, or the data centers that have already reserved the next wave. Until new fabs ship real volume — not roadmaps — that question stays open, and your upgrade budget stays hostage to it.
