OpenAI CEO Sam Altman said in an interview that much of the DRAM in personal computers sits idle most of the time, and that centrally deploying it in AI data centers could greatly improve resource utilization—loading too much memory into PCs is a waste of resources. This view has sparked concerns about a future in which intelligent services can only be rented from the cloud. He also offered several judgments: large-scale AI-driven code scanning will trigger a wave of vulnerabilities far beyond expectations; existing sandboxes cannot fully guarantee the safety of highly intelligent models, and a diverse checks-and-balances mechanism composed of a team of multiple agents is a feasible direction for safety. He disclosed that new internal models have successively solved long-standing research problems in multiple fields, and argued that adding a trillion AIs slightly smarter than humans to society can better ensure social robustness than a single AI with a trillion IQ. He pointed out that memory manufacturers are extremely conservative about expanding production because of their memories of losses from past expansion, and that the market adjustment mechanism has failed; OpenAI is advancing self-developed chips, building its own data centers, and even using AI-controlled robots to build self-replicating data centers, with the goal of greatly reducing AI computing costs. He also mentioned that the moment when personal agents become widespread has not yet arrived, and that OpenAI once considered abandoning token-based billing in favor of per-task pricing, but dropped the idea because of strong customer opposition.