What did sam altman say about ram

On 28 October 2026, Sam Altman addressed the growing bottleneck of Random Access Memory (RAM) in the development of next-generation artificial intelligence models. Speaking at a technology summit, Altman emphasized that the current architecture of AI hardware is increasingly limited by memory bandwidth and capacity rather than just raw computational power. He noted that as models grow in complexity, the ability to move data efficiently between memory and processors has become the primary constraint for scaling intelligence. Altman suggested that the industry must move toward more integrated memory solutions, potentially moving away from traditional von Neumann architectures to reduce latency. He highlighted that future AI systems will require a fundamental rethink of how memory is tiered and accessed to support real-time reasoning at scale. By prioritizing high-bandwidth memory (HBM) and exploring novel memory-centric computing designs, Altman believes developers can overcome the "memory wall" that currently hinders the deployment of more autonomous and capable agents. His comments underscore a strategic shift for OpenAI, focusing on hardware-software co-design to ensure that future infrastructure can sustain the massive data throughput required for advanced cognitive tasks.