Altman Says Your PC's RAM Is a Waste. Nvidia Has Stopped Selling You the GPU to Use It.

OpenAI's CEO called it "quite terrible" that people have the world's scarce DRAM sitting in their own computers, used 5% of the time — and said the search for solutions is going "bad." The same week, the RTX 5080 disappeared from Nvidia's US store, two weeks after the RTX 5090, with the cheapest in-stock card at $2,199 against a $999 launch price.

AI synthesis3 sources2

Sam Altman has a problem with your computer. Not with the software on it — with the memory inside it.

"It's quite terrible to have everybody have a bunch of the world's limited supply of DRAM sitting in their own computer. And be used 5% of the time," the OpenAI chief executive said in an interview on the Cheeky Pint podcast. His argument is about utilisation: in a supply-constrained world, memory that idles is memory wasted. "You want maximum utilization in the cloud if you're trying to deal with this very supply-constrained environment," he said. "And somehow, many people's intuition is the opposite of that."

Asked how the search for solutions was going, Altman answered in one word: "Bad." He was no more hopeful about prices fixing the problem by themselves. "I think the invisible hand is failing right now," he said. "It has a checkered track record at best. But, I think it's not quite working here." Memory manufacturers, he added, are right to be careful about building capacity — they remember the gluts that nearly bankrupted them.

What turns a podcast remark into a story is what Nvidia did in the same week. After the GeForce RTX 5090 vanished from Nvidia's US marketplace, the RTX 5080 followed: the listing and its filters are gone, and partner cards have gone with them. Tom's Hardware found the cheapest in-stock RTX 5080 — a PNY model — at $2,199.99, or 2.2 times the $999 price the card launched at in January 2025.

The pattern matters more than the stockout. Nvidia's RTX 5070 Ti and RTX 5070 are still listed, marked out of stock, which lets buyers anticipate restocks. The 5090 and 5080 were removed outright. That has fed rumours that Nvidia has halted production of both, diverting GB202 silicon to the far more profitable RTX Pro lineup, and that a 24GB RTX 5080 may take the flagship slot. Nvidia maintains the RTX 5090 is still alive and has not commented on the 5080's status.

The contradiction is the point. Nvidia and Microsoft have spent the past year selling local AI — machines with enough memory to run models at home, including the expensive RTX Spark laptops that are unlikely to reach mainstream prices soon. OpenAI's CEO is now arguing that exactly that arrangement is a misallocation of the world's memory. It is an obviously self-serving argument, and it is not an invented one: the shortage is expected to persist for years, with projections cited by PCMag running to 2030.

It also lands on ground that has already shifted. AI demand has wrecked the PC businesses that chipmakers still depend on, and the consumer market is absorbing the bill one withdrawn SKU at a time. The question of who gets the world's DRAM is being settled in storefronts and street prices long before anyone votes on it.