I hate to say this, but Sam is right here. And the way this is phrased is incredibly disingenuous.
He is NOT saying nobody should have RAM.
The surrounding discussion was cut out: the question was about compute shortages and why higher memory prices weren’t bringing enough new supply. His point was that everyone buying dedicated hardware with 128GB of RAM to run local models wouldn’t solve that shortage. And he’s right.
Even if local models get exceptionally good (which they aren't rn), your hardware sits idle when you aren’t running them and you have limited burst usage. When supply is limited, sharing compute makes better use of what’s available: you can access more capacity than you could reasonably own, pay for what you use, and leave that capacity available to someone else when you’re done.
If you’re reading that as “everyone should just use closed source models”, that doesn’t follow either. Shared hardware can run open models too and this is purely an allocation argument in a compute crisis.
A library is a useful comparison here: several shared copies of books can serve readers who would otherwise each need their own. That doesn’t eliminate the reasons to own books, neither does it make more books appear. It just lets more people use the copies available.
I'm pissed that I even have to defend this, but framing his statement like this is misleading and prevents meaningful discussion.
OpenAI CEO Sam Altman believes people owning RAM is wasteful, and that it belongs in data centers instead.
"From a global allocation perspective, actually, it’s quite terrible to have everyone have a bunch of the world’s limited supply of DRAM sitting in their own computer."