Sam Altman says it’s better for RAM to be in a data center
OpenAI CEO explains why memory belongs in the cloud rather than personal PCs amid global RAM supply shortages.
Stock photo for illustration only, not from the actual event
- OpenAI CEO Sam Altman calls having RAM in personal computers terrible
- Distributing memory to individual devices leads to underutilization
- Memory makers secure long-term deals with cloud operators and Stargate
- Massive chip spending drives up the cost of consumer laptops and phones
As OpenAI’s aggressive buying spree puts pressure on the global supply of RAM, the company's chief executive has laid out the reasoning behind his massive appetite for memory chips. Sam Altman argued that keeping valuable RAM inside personal computers is far inferior to housing it inside centralized data centers.
During an appearance on the Cheeky Pint podcast, where discussions also touched on AI's influence on mathematics, Altman addressed ongoing supply chain bottlenecks surrounding the global memory shortage and criticized the common intuition regarding local computing infrastructure.
"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," Altman stated.
Stock photo for illustration only, not from the actual event
"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."
Sam Altman
Altman further emphasized that achieving maximum utilization in the cloud is essential when dealing with such a severely constrained supply environment, noting that many people's instincts run entirely counter to this reality.
Sam Altman's perspective highlights a fundamental tension between the push for Edge AI and the massive scalability demands of modern cloud architecture. Advanced AI models require immense memory bandwidth, meaning decentralized execution on personal hardware results in poor asset utilization compared to pooled cloud resources.
Whether referring to high-end gaming rigs or developers utilizing specialized hardware like the Surface RTX Spark Dev Box or the new touchscreen MacBook Pro, the economic and logistical outcome remains unchanged.
At present, major memory manufacturers Samsung and SK Hynix have secured long-term contracts with cloud service operators and infrastructure initiatives like OpenAI's Stargate, effectively reserving a massive share of worldwide RAM output for the foreseeable future.
The billions of dollars these tech giants are willing to commit for advanced chips directly explains why consumer electronics such as laptops, smartphones, and other everyday gadgets are seeing price increases.
Source: The Verge
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment