In an informal beer-chat streamed on the Cheeky Pint YouTube channel, OpenAI’s CEO Sam Altman sparked controversy by criticizing people for having RAM in their desktop PCs, arguing that, in his view, this memory is used only about 5% of the time, a level he deems inefficient at a moment of globally limited supply. He spoke about this in the context of people who want to run open-source AI models on their own machines.
From Altman’s perspective on global allocation, the most efficient approach would be for cloud data centers to make the most of that memory supply, rather than spreading it across home devices that run AI models locally—a idea that, according to him, goes against the intuition of many who think exactly the opposite.
The cloud business model of Sam Altman and the concern about hardware control
Altman indicated that, from a global allocation standpoint, it is “quite terrible” that everyone has a share of the world’s limited DRAM supply on their own PC and uses it only about 5% of the time, while in a memory-scarce environment data centers must extract the maximum value from those same resources to process artificial intelligence.
Although the user community values the freedom to run open-source tools directly on their machines, Tom’s Hardware notes that OpenAI’s core business revolves around closed models that operate in data centers, so the outlet sees Altman defending the company’s model, which, although it includes some open-source models, hinges on people paying subscriptions to use its models remotely.
These statements fuel fears among technology enthusiasts about losing ownership of their equipment, since the outlet suggests they could lead to a scenario in which people neither own their hardware nor their software and must rent a cloud assistant and a computer through monthly payments.
All of this stems from the premise that the memory that runs 24/7 in the cloud for AI is worth more than memory used for gaming, spreadsheets, or web browsing, and from the belief that building data centers will not slow down, two assumptions that Tom’s Hardware questions.
Additional points of the conversation and the future of technology infrastructure
Beyond the RAM debate, the interview covered discussion of PHP vulnerabilities detected this week and the practice of isolating models in controlled environments during training, known as sandboxing.
There was also talk about OpenAI’s research in mathematics, underscoring that the chat was broad and that the RAM topic was only one part of it.
Taken together, Altman’s remarks reveal the tension between large corporations that aim to maximize the use of cloud hardware and users who defend the right to own and control their own technology.
You can watch the moment in the following video (in English), starting at 29:00.
