OpenAI CEO Sam Altman argued in a podcast that running AI models locally is a memory-inefficient approach that diverts scarce DRAM supplies from data centers amid hardware constraints expected to last through 2030.
According to Altman, conserving memory on personal computers, where it is used only a small percentage of the time, is not effective on a global scale. On the other hand, OpenAI partners such as Nvidia and Microsoft are promoting the benefits of local operation on new systems with high memory capacity. Local AI offers better privacy and allows consumers and businesses to bypass online storage and processing services without cost, which threatens large providers.
The lack of memory and the market reaction
The discussion highlighted the severe memory shortage facing the market, a situation that is estimated to continue until 2030. Altman said the market is failing to properly meet demand, as manufacturers are wary of building new factories, trying to avoid overproduction and the risk of bankruptcy.
Because of this narrowness, new computers and specialized systems optimized for local AI remain very expensive and are unlikely to be adopted en masse until prices come down.
G
HACKS.GR / GOOGLE SEARCH
Hacks.gr on Google
Set Hacks.gr as a preferred source.
Preferred source
↗
end of article
Top comments (0)