Sam Altman has questioned how personal computers use scarce memory. The OpenAI chief executive argued that cloud data centres could use it more consistently than home machines running artificial intelligence (AI) occasionally. Speaking to Stripe co-founder John Collison on the Cheeky Pint podcast, he challenged the assumption that locally run AI models would reduce the need for large computing facilities.“A lot of people want local inference. And somehow, they assume that the problem is solved now. Because we are going to have open-source models. We’re going to be able to run them on our computers. And so we’re not going to need all the stuff in the data center. But 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 using it only 5% of the time. You want maximum utilisation 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,” Altman said in the podcast.The remarks drew online objections from users who interpreted them as favouring cloud providers over people who own their hardware. Social media users criticised Altman and expressed concern that consumers could become increasingly dependent on rented computing services.
Why Sam Altman sees cloud servers as more efficient
Collison asked why high hardware prices had not attracted new suppliers. “I think the invisible hand is failing right now,” Altman replied. The phrase refers to the expectation that market incentives, including higher prices, encourage businesses to increase supply.The discussion then turned to local inference, which means running an AI model on a person’s own computer instead of a remote server. DRAM, or dynamic random-access memory, supplies the working memory those computers need to run applications.Altman argued that shared cloud servers could keep scarce memory occupied longer by serving different customers throughout the day. His five-per-cent figure was presented without supporting measurements in the quoted passage.
Why people still want AI on their own computers
Altman was discussing local AI workloads, rather than gaming computers specifically. However, a personal computer’s memory also supports browsing, spreadsheets, games and other apps. Time spent running AI alone does not capture its overall usefulness.People may also choose local models for reasons beyond hardware efficiency. Running a model on their own device can let them keep information local and avoid a recurring cloud subscription, depending on the software used.
