The Hidden AI CPU Opportunity

Summary

While the market is buzzing with AI infrastructure spending, major investors like Warren Buffett are stockpiling cash, signaling a strategic move rather than simple greed. This shift highlights a crucial overlooked area: CPUs in AI data centers. Traditionally, AI relies heavily on GPUs for processing, with CPUs handling traffic. However, with emerging agentic AI, CPUs are becoming critical for orchestration, tool calls, and managing extensive context. New data suggests a much higher CPU-to-GPU ratio is needed than previously thought, with some estimates indicating a 1:3 ratio by chip count. This miscalculation has led some, including the speaker, to underestimate companies like AMD and Intel, whose data center CPU revenues are growing rapidly. AMD's significant deal with Meta, which includes a substantial warrant for Meta, illustrates the high stakes and potential dilution. Intel, conversely, is securing major customers like Tesla and Google, signaling a revival for its chip manufacturing capabilities. Adding to the competition, ARM has launched its AGI CPU, claiming significantly better performance per watt and cost savings compared to existing solutions. These developments, coupled with a potential breakthrough in AI model efficiency from startups like Sub Quadratic, suggest that the demand for AI infrastructure, especially CPUs, is poised for exponential growth, presenting a major investment opportunity.