Companies are tightening their AI budgets and redirecting spending toward measurable return on investment, a structural shift that threatens to slow growth rates at OpenAI and Anthropic. The pivot — from what insiders call "tokenmaxxing," the unconstrained consumption of AI tokens, to a discipline of efficiency — marks a maturation point in enterprise AI adoption with direct implications for the two companies most exposed to that spending.
The End of Tokenmaxxing
The tokenmaxxing era was defined by experimentation: companies integrating AI broadly, accepting high inference costs as a feature of moving fast. That posture is changing. Budget holders are now asking what each deployment actually returns, and whether the volume of AI usage justifies the line item. For OpenAI and Anthropic, whose revenue models are tied directly to consumption, any sustained slowdown in discretionary model usage is a direct revenue variable.
The shift is less about dissatisfaction with the technology than about the normal maturation of any enterprise software cycle. Early adopters buy broadly; the second wave buys deliberately.
What Efficiency Mode Means for OpenAI and Anthropic
Both OpenAI and Anthropic built their growth trajectories on the assumption that AI consumption would expand rapidly and continuously. A budget-tightening cycle compresses that runway. Companies focused on efficiency will consolidate AI vendor relationships, retire redundant use cases, and demand clearer unit economics before scaling new deployments.
That dynamic creates pressure on pricing, on usage volume, and on the pace at which new enterprise contracts ramp. Dampened growth rates — even if still positive — recalibrate the valuation logic both companies have carried into their funding rounds.
The Second-Order Read
When corporate buyers begin optimizing rather than adopting, they signal that AI has crossed from the experimental budget into the operational budget, where CFOs apply standard scrutiny. That is, in one sense, evidence of the technology's legitimacy. In another, it is the moment when growth-at-any-cost models meet the discipline of the P&L.
OpenAI and Anthropic will need to demonstrate that efficiency-focused customers still expand usage over time rather than simply rationalize it. The growth narrative shifts from land-and-prove to land-prove-and-expand — a harder sequence to sustain at the same velocity.