OpenAI CEO Sam Altman argues against running AI models on personal computers
In a podcast interview, Sam Altman said that using cloud services for AI inference is more efficient than individuals running models on their own PCs, citing limited memory supplies.
Shares of US chipmaker Intel fell sharply on Friday, dropping nearly 30 percent, the steepest decline in 52 years, as analysts linked the slump to a slowdown in the personal-computer market and rising costs of memory and storage driven by the AI boom.
OpenAI chief executive Sam Altman expressed his view that running AI models locally on personal computers is impractical. Speaking on the Cheeky Pint podcast, Altman said the push for “local inference” assumes the problem is solved by open-source models, but it would allocate scarce dynamic RAM to individual devices rather than to data-center servers used by hyperscalers.
“A lot of people want local inference … and somehow, they assume, that the problem is solved now … But from a global allocation perspective, actually, it’s quite terrible to have everybody have a bunch of the world’s limited supply of [dynamic RAM] sitting in their own computer,” Altman said.
He added that maximizing utilization in the cloud is a better response to the constrained hardware environment.
“You want maximum utilization in the cloud if you’re trying to deal with this very constrained environment … And somehow everyone’s, many people’s intuition is the opposite of that,” he said.
Altman’s comments have sparked criticism from some PC-gaming and tech-enthusiast communities online, who argue that individuals should retain the right to own and use their own hardware.
The debate comes as the AI sector faces ongoing challenges in commercializing products while managing rising expenses and a tightening supply of AI-focused chips, with manufacturers reporting a circuit-board shortage that could extend through 2030.