
Altman Claims DRAM Prices Rose '5x, 7x' as He Walks Back PC RAM Comments
Altman estimates DRAM prices rose '5x, 7x' as AI firms buy up memory supply, and now says he wants everyone to have 'a lot of RAM' after his datacenter remarks drew backlash.
- By
- Sophie Lindqvist
- Filed
- Channel
- Memory & Storage
- Read
- 4 min read
DRAM prices have climbed roughly five to sevenfold in recent months, according to Sam Altman's own estimate, and the OpenAI CEO says the AI industry's memory appetite is part of the reason. He offered the figure on Monday at Semafor's Silicon Valley & The World summit, with an unusual hedge attached. "I don't know exactly, but maybe they went up 5x, 7x, something like that," Altman said on stage.
The number is not a market datum. It is the CEO of the world's most capitalized AI lab guessing at a price trend that his own company's purchasing helps drive. Buyers of phones and game consoles have already felt the effect: AI firms have been buying up global memory supply for datacenters, and prices have risen across consumer products as a result. Altman did not dispute that dynamic. He confirmed it.
What did Altman actually say?
The appearance was an attempt to contain a month-old controversy. On the Cheeky Pint podcast in October, Altman said it is "quite terrible" for everyone to hold RAM in personal computers rather than in datacenters. The remark spread quickly. On X earlier this month, he posted "i want you to have so much ram" and claimed the original comments were "taken out of context."
On Monday he extended the walkback. "I want to be clear, I want people to have a lot of RAM, I think everybody should be able to buy a lot of RAM for their own computer," Altman said. "I think the home tinkering is great."
Note the sequence. The clarification arrived only after public backlash, first on X, then on a summit stage. The original claim came unprompted.
Is the 'shared memory' argument sound?
Beneath the apology sits an economic argument Altman refuses to abandon. He said OpenAI began arguing "a few years ago" that "the world needs to build a lot more memory supply chain in particular." In his telling, the world ignored the warning. The shortage arrived anyway.
Altman then restated the core of his October position, softened but intact:
"There is a true part of it, which I will stand by is that, if we're going to be in this memory supply shortage, then in some sense — whether or not we'll decide to do it, because I think there are other tradeoffs like people having their own at-home inference — the optimal thing would be, can we share this memory and have it working 100 percent of the time rather than sometimes off in someone's edge device?"
The logic is utilization. Memory sitting idle in a laptop serves no model. Memory pooled in a datacenter runs continuously. Altman framed this as "optimal use of this limited supply," while conceding it is "maybe not the reason to make the decision." Who decides, and on what terms, he did not say.
The claim also needs context he did not provide. Datacenter memory purchases earn revenue for OpenAI's workloads. Consumer RAM purchases do not. An argument for pooling scarce DRAM in AI facilities is, not coincidentally, an argument that concentrates the scarce resource where OpenAI operates. Altman offered no independent analysis to separate the efficiency case from the commercial one.
What are the supply constraints?
Altman acknowledged that OpenAI has "been constrained by different things at different times." He listed memory, rack assembly, datacenter capacity, and "some random component." The list is the clearest public admission yet that OpenAI's buildout competes for the same supply chain that serves consumer electronics. He gave no dates, no volumes, and no forecasts.
His 5x-to-7x price estimate, by his own admission approximate, still understates the problem he describes. If AI demand lifted DRAM contract prices by even the low end of his range, the cost passed through to phones, consoles, and PCs is a direct transfer from consumers to AI infrastructure buyers. Altman's remedy — build more supply — is a multi-year proposition. Memory fabs take years to plan and equip. No supplier commitments were named.
Does the Huang question land?
Altman declined to weigh in on Jensen Huang's resignation as Nvidia's CEO. He said he "doesn't know enough" to say whether Huang should be reinstated. The answer avoids a public position on the supplier behind most of OpenAI's compute.
The pattern across the appearance is consistent: Altman concedes the shortage, concedes the price spike, and concedes his industry's role in both, while framing datacenter concentration of memory as an optimization rather than a land grab. With DRAM pricing still climbing and consumer products absorbing the cost, that framing will face continued scrutiny from everyone who now pays more for their own RAM.
Source: theverge.com
More from Sophie Lindqvist
Show full bio
News editor covering money, funding and market shifts around Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.
1,065 articles
Related articles
musk-mocks-altman-s-claim-that-pcs-having-ram-is-quite-terrible
Musk Mocks Altman's Claim That PCs Having RAM Is 'Quite Terrible'
apple-s-memory-bill-climbs-33-billion-as-ram-prices-surge-into
Apple's Memory Bill Climbs $33 Billion as RAM Prices Surge Into 2027
openai-s-altman-calls-consumer-dram-quite-terrible-for-ai-era
OpenAI's Altman Calls Consumer DRAM 'Quite Terrible' for AI Era
nvidia-flags-15-percent-nvl72-hike-as-memory-spot-prices-climb-sevenfold-92e6ff36
Nvidia Flags 15 Percent NVL72 Hike As Memory Spot Prices Climb Sevenfold



