What happened
CNBC reported that companies which once maximized frontier-model usage now want measurable ROI, tighter controls and lower-cost options. The 'tokenmaxxing' mindset, where employers encouraged developers to spend freely on the best models, is fading in favor of budget discipline.
The shift shows up in concrete moves. Uber implemented AI-tool spending tiers starting at a $1,500 monthly base, Lindy switched off Claude and routed all traffic to DeepSeek, and Microsoft unveiled low-cost models while steering users to the most appropriate model for each task.
Why it matters
OpenAI and Anthropic have seen valuations approach roughly $1 trillion on the back of the spending boom. If customers rationalize budgets and lean on cheaper alternatives, that growth story faces real pressure.
The change also rewards routing and tiering: matching each task to a model by cost and capability rather than defaulting to the most expensive option.
MintedBrain take
Efficiency is becoming the default expectation, not a nice-to-have. Instrument your AI spend, set tiers, and be ready to route cheaper models at tasks that do not need the frontier.
Discussion
Sign in to comment. Your account must be at least 1 day old.