Updated
Updated · The Verge · Oct 30
Altman Says RAM Belongs in Data Centers, Not PCs, as OpenAI Buying Squeezes Global Supply
Updated
Updated · The Verge · Oct 30

Altman Says RAM Belongs in Data Centers, Not PCs, as OpenAI Buying Squeezes Global Supply

3 articles · Updated · The Verge · Oct 30

Summary

  • Sam Altman said scarce DRAM is better deployed in data centers than personal computers, arguing cloud systems use memory far more efficiently in a supply-constrained market.
  • On the Cheeky Pint podcast, he said local AI inference wrongly assumes open-source models will reduce data-center demand, even though PCs may use their RAM only about 5% of the time.
  • Samsung and SK Hynix have already locked in long-term memory deals with cloud operators and projects such as OpenAI's Stargate, reserving much of global RAM output for the next few years.
  • That spending spree is helping push up prices for laptops, phones and other consumer devices, underscoring how AI infrastructure demand is reshaping the broader tech supply chain.

Insights

Is Sam Altman right that scarce RAM is wasted in PCs—and why is AI’s real bottleneck becoming memory, not compute?
Could memory pooling technologies solve the AI RAM crunch before higher laptop and phone prices become the new normal?