Anthropic Demand Intelligence

Enterprise agent consumption vs. compute efficiency: The checks show two forces moving in opposite directions. Claude Code adoption and spend are rising rapidly, but enterprises are simultaneously learning to cut tokens, turns and cost per completed task at remarkable rates. The evidence is strong enough to reject a simple equation of 'more AI users = proportional GPU demand.' The more useful NVDA question is whether growth in concurrently executed agent work is outrunning the decline in compute required per unit of useful work.