Sam Altman on Storage and Data Centers

Sam Altman, CEO of OpenAI, recently shared a controversial opinion regarding the use of memory on personal computers in a video interview on the YouTube channel "Cheeky Pint." In the conversation with John Collison, President of Stripe, Altman argued that it is problematic for consumers to have large storage capacities on their devices. He suggests that this storage would be better housed in large, centralized data centers. Altman explained that many users aim to perform local inference by running language models on their own devices. He pointed out that the availability of open-source models does not automatically solve the issue of computational power.

"We will be able to run these on our computers. And that’s why we won’t need all that stuff in the data center anymore," he said. The discussion about shifting storage to data centers is intensified by the current market situation. Consumers are currently paying about 500 percent more for storage than they were in early 2025. These price increases affect the costs of computers, graphics cards, mobile phones, and gaming consoles, making the idea of removing storage from end devices unattractive for many users.

A central argument from Altman is the efficiency that could be achieved by concentrating computational power on a few nodes. This could be particularly beneficial for large language models. However, critics fear that such centralization could also bring new problems, including economic, societal, and environmental challenges. Altman's statements have elicited mixed reactions within the tech community. While some view the idea of centralized computing power as forward-thinking, others see it as a threat to individual control over technology and data.

Concerns about dependence on large companies operating such data centers are also a common theme in the discussion. The notion that every user has a portion of the limited DRAM supply in their own computer is seen by Altman as problematic. He argues that this memory is only utilized five percent of the time, questioning the necessity of redistributing the storage. Critics, however, warn that such redistribution could also impair accessibility and availability of technology for end consumers. The debate over the future of storage management is further fueled by the growing demand for AI services and the associated requirements for computational power.

Companies like OpenAI, Anthropic, and Google are already operating large data centers to meet the demands of their AI models. The question remains how these developments will impact consumers and the industry as a whole. Altman concluded the conversation by noting that the discussion about storage and data centers has not only technological but also societal implications. "We need to think about how to use resources efficiently while considering the needs of users," he said. The discussion about shifting storage to data centers is expected to continue in the coming months.