OpenAI's Sam Altman Says Running AI on Your Own Computer Is a Waste
He argues it's better to rent AI from the cloud—but is that just because he sells it?
Sam Altman, the head of OpenAI (the company behind ChatGPT), recently said that running AI models on your own computer—called 'local inference'—is a 'terrible allocation of scarce resources.' In plain English, he means it's wasteful. He argues that if everyone runs AI on their own machines, the world's limited supply of memory chips (DRAM) would sit idle most of the time, used only 5% of the time. Instead, he says, it's much more efficient to run AI in big data centers (the cloud) where hardware is shared and used nearly all the time.
Why should you care? This debate is about whether you'll be able to own and run AI tools yourself, or whether you'll have to rent them from big companies like OpenAI. If Altman's view wins, AI could become more centralized and potentially more expensive, with companies controlling access. If local inference becomes popular, you might have more control, privacy, and potentially lower costs over time. The catch is that Altman is the CEO of a company that sells cloud-based AI, so he has a financial incentive to discourage local inference. As one critic joked, 'Guy selling product: Here's why owning your own product instead of renting it from me is a terrible idea.' So while his argument about efficiency has merit, it's also self-serving. The real question is whether the convenience and efficiency of cloud AI outweigh the benefits of local control and privacy.
- Sam Altman says running AI on your own computer wastes memory chips because they'd be idle most of the time.
- He argues cloud data centers use hardware more efficiently, which is better for everyone.
- But Altman sells cloud AI, so his opinion may be biased—critics say he's just protecting his business.
Why It Matters
This could affect whether you can own AI tools or must rent them, impacting cost and privacy.