Altman says data centres use scarce memory more efficiently than personal computers
Sam Altman argued that, during a severe memory shortage, it is better for the world’s limited DRAM supply to be concentrated in data centres than in personal computers. He said cloud servers can keep memory in use more of the time, challenging the view that running AI models locally would reduce demand for data-centre resources.
Altman made the comments on the Cheeky Pint podcast while discussing supply-chain constraints and local AI inference. The article says Samsung and SK Hynix have struck long-term deals with cloud operators and projects such as OpenAI’s Stargate, reserving much of their expected RAM output for the next few years. Heavy spending on memory by these buyers is contributing to higher prices for laptops, phones and other technology.
In short
- Altman says cloud data centres can use scarce RAM more efficiently.
- Major cloud and AI projects have reserved much of future memory supply.
- The competition for chips is pushing up consumer technology prices.
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
When DRAM is scarce, allocating more of it to data centres could make each unit of memory serve more users and spend less time idle than memory installed in individual PCs. Centralised infrastructure may therefore deliver more computing access from a constrained supply, while long-term purchasing can help fund the capacity needed for cloud services.
The case against
Concentrating scarce DRAM in data centres can give a few large buyers priority and push up costs for consumers and businesses that need memory in their own devices. Local AI processing may also offer benefits such as privacy, offline access and less dependence on cloud infrastructure; judging efficiency only by utilisation could overlook those advantages and the risks of relying on centralised services.
Explained simply
DRAM, often called RAM, is short-term working memory used by computers to run programmes and handle data. It is made by a small number of large manufacturers, and shortages can affect the cost and availability of many kinds of technology.
Data centres are large facilities that house powerful computers and provide online services, including AI tools. Sam Altman leads OpenAI, which develops AI systems; Samsung and SK Hynix are major memory makers, while cloud operators run data centres for businesses and the public.
The debate is about how limited memory supplies should be used as demand grows. Concentrating memory in data centres could keep it in active use across many users, while running AI on personal computers could reduce reliance on remote servers; choices about supply and investment can also affect prices for devices such as laptops and phones.
AI Business Companies Technology
Read the full article at the source →
Originally published by The Verge as “Sam Altman says it’s better for RAM to be in a data center than on your desk”.