If you’ve tried to buy new electronics recently, from gaming consoles to iPhones, you’ll have dealt with prices that are hundreds higher than they were even a year ago. That’s because the price of RAM keeps going up as AI companies race to build out datacenters to feed the insatiable compute demands of frontier LLM models. Lots of people think being unable to buy RAM for consumer products at a reasonable price is a bad thing. OpenAI’s Sam Altman disagrees.
“A lot of people want local inference,” the executive whose company is expected to go public sometime next year recently told the Cheeky Pint podcast. “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.”
There’s a debate in the AI world right now about whether the future will be people running “light weight” open-source models on local devices, like phones and laptops, or whether all meaningful AI models will be run out of the cloud, i.e. massive data centers controlled exclusively by tech giants. Altman is coming down firmly on the side of the datacenters, and thus the companies currently buying up all available memory and sending prices skyrocketing for years to come in the process.
“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 being used five percent of the time,” Altman continued. “You want maximum utilization in the cloud if you’re trying to deal with this very supply constrained environment. And somehow everyone’s—many people’s intuition is the opposite of that.”
These “all your RAM are belong to us” proclamations were delivered in-between sips of a slowly warming IPA (all the tech CEOs are now forced to discuss world historical moments while awkwardly smashing beers on podcasts), but will have far-reaching consequences if they actually come to pass. It’s not hard to imagine a future where everything is rented, not just movies and Google spreadsheets but also the AI models these companies hope will micromanage and ultimately monetize every spare moment of your life.
Being able to play Grand Theft Auto 6 locally once it gets ported to PC is a waste in the mind of Altman. Stream it from the cloud in-between datacenters burning tokens to scrape the web and replace it with AI-generated overviews instead. Better yet, have ChatGPT agents bid on your behalf against other agents for how much the spare compute is worth it to you to take a nightly drive around Vice City. Please clap.
