OpenAI CEO argues centralizing RAM use improves global efficiency
Juste les faits

OpenAI CEO argues centralizing RAM use improves global efficiency

Résumé

OpenAI chief Sam Altman says keeping large-memory workloads in data centers maximizes utilization and eases pressure on the limited DRAM supply.

OpenAI chief executive Sam Altman explained that locating high-memory artificial-intelligence workloads in data centers, rather than on individual desktops, is more efficient for the global DRAM market. He noted that many users expect open-source models to run locally, but that approach would spread a scarce resource across many machines that are idle most of the time.

“From a global allocation perspective, it’s inefficient to have a large share of the world’s limited DRAM sitting in personal computers and being used only a small fraction of the time,” Altman said. He added that maximizing utilization in the cloud is essential when the supply of memory chips is constrained.

Altman’s comments come amid reports that OpenAI’s recent acquisitions have increased demand for RAM, prompting the company to emphasize cloud-based solutions for large-scale inference tasks.

Continuez à lire — gratuitement

Créez un compte gratuit pour suivre l'actualité. Aucune carte requise.