Coinbase CEO Brian Armstrong has identified energy and compute infrastructure as the primary constraints on artificial intelligence growth, rather than model quality. Armstrong's insights came in response to investor Tommy Shaughnessy's observations on the rising costs of enterprise AI due to metered API pricing. Armstrong predicts that within 12 to 18 months, 80% of AI workloads will shift to models priced significantly lower than current top-tier options, while the remaining 20% will continue to use cutting-edge models for critical applications. Armstrong emphasized that as model costs decrease, the bottleneck will shift to the power and silicon needed to run models at scale. He noted that Coinbase is already implementing strategies to manage AI costs by routing tasks to lower-cost models. Armstrong's stance against AI overregulation reflects his belief that technological advancement should not be hindered by policy constraints. He highlighted that the real challenge will be whether energy and computing infrastructure can meet the seemingly limitless demand for AI-generated intelligence.