Sam Altman calls consumer RAM use for AI "quite terrible"

OpenAI chief Sam Altman told the Cheeky Pint podcast (via IGN) that it is "quite terrible" for a large share of the world's limited DRAM supply to sit in personal computers used only about 5 % of the time. He argued that, in a supply-constrained environment, memory belongs in cloud data centers where it can run at near-full utilization.

• Altman: idle RAM is better deployed in the cloud at maximum load.
• The stance signals a preference for cloud over local compute.
• RAM prices have risen as AI firms buy up server and HBM stock.
• The author sees this as a potential threat to consumer interests.

What happened In the interview Altman said the intuition of people who want to run open-source models on their own hardware runs counter to global resource-allocation logic. From his perspective, every gigabyte of DRAM should go where it will be saturated 24/7 — i.e., the data center.

Technical and economic context Over the past 12–18 months DDR4 and DDR5 module prices have climbed sharply as hyperscalers (OpenAI, Microsoft, Google, Meta) expanded HBM and server DRAM orders, crowding out the consumer channel. The "5 % utilization" figure is a rule-of-thumb illustration of the efficiency gap Altman highlights.

What it means for users and developers If major players consolidate memory in the cloud, local AI inference (llama.cpp, LM Studio, Ollama) could become costlier or less accessible, and memory prices may stay elevated long-term. For developers and enthusiasts in Central Asia and the post-Soviet space, this adds a new budget variable when planning upgrades or self-hosted models.