Sam Altman RAM Remark: Your PC Memory Belongs in Data Centers
OpenAI CEO Sam Altman says it is “quite terrible” for the world’s limited supply of DRAM to sit inside personal computers, where he estimates it is used about 5% of the time. Speaking on Stripe’s Cheeky Pint interview show, published this week, Altman argued that memory does more work in cloud data centers, a view that puts the Sam Altman RAM debate at the center of the ongoing memory shortage.
What Sam Altman Said About RAM and Local Inference
The comment came about 29 minutes into the conversation, when the discussion turned to how the supply chain is reacting to high prices for memory and GPUs. Altman pushed back on the idea that local inference, running open source AI models on your own computer, will reduce the need for data centers.
“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,” Altman said. “You want maximum utilization in the cloud if you’re trying to deal with this very supply constrained environment. And somehow many people’s intuition is the opposite of that.”
Why Memory Makers Are Not Flooding the Market
Earlier in the same segment, Altman explained why higher prices have not quickly brought more supply. For a long time, he said, a DRAM producer would “invest $20 billion of CapEx to make $19 billion,” riding cycles of booms and busts. Now, in his words, these companies are each on track to make around “$100 billion of free cash flow in a year,” but they remember how close they came to bankruptcy by overbuilding fabs, which keeps their spending cautious.
That caution helps explain why prices are still climbing. Memory suppliers are posting record results, from CXMT’s $8.2 billion quarterly profit to Sandisk’s 378% revenue surge, while PC and phone buyers pay more. Our explainer on why RAM is so expensive covers how AI demand is absorbing production.
What It Means for PC Users and Local AI
Altman’s argument is about utilization: a memory chip in a busy server works around the clock, while the 32 GB in a gaming PC or laptop sits mostly idle. Critics point out that this framing also matches OpenAI’s business model, which depends on paid access to closed models running in its own data centers. If capable open models run well at home, fewer people may need a cloud subscription.
The remark is likely to land badly with PC builders and developers who have been paying higher prices for RAM upgrades, and with privacy-focused users who prefer to keep AI workloads on their own hardware. It also comes as OpenAI looks for more capital to fund its infrastructure, with its latest funding round reportedly scaled back to $20 billion.
The wide-ranging interview also covered this week’s PHP security bugs, AI progress in mathematics and Altman’s view on why the world should not rely on a single superintelligence. The full conversation is available on the Cheeky Pint YouTube channel, and the RAM comments were first highlighted by Tom’s Hardware.
Your comment has been submitted,
it will be published after approval.