AI Efficiency Shift: Buyers Ditch ‘Tokenmaxxing’
2 min readFor two years, the rule in enterprise AI was simple: spend more, get more. That era is ending. A new CNBC report describes an AI efficiency shift, with businesses moving away from maximizing usage, a habit some insiders call “tokenmaxxing,” and toward squeezing the most value from every dollar. The change lands squarely on OpenAI and Anthropic, the two companies that benefited most from the old approach.
From Spend-at-All-Costs to Efficiency
Both OpenAI and Anthropic filed confidentially for IPOs in early June, with valuations approaching $1 trillion. Their rapid growth was powered by a market willing to pay almost any price for the most capable models. Customers ran ever larger prompts, longer context windows, and heavier agentic workloads without watching the meter too closely.
Now the meter matters. As AI moves from experiments into production systems, finance teams are scrutinizing the bill. The question has shifted from which model is smartest to which model is good enough at the lowest cost.
What Changed
The clearest example comes from Lindy, an AI startup whose CEO switched the company entirely off Anthropic’s Claude models and moved all of its traffic to DeepSeek, a Chinese provider known for cheaper alternatives. The result was a steep cut in costs. It is a small case on its own, but it signals a broader willingness to trade brand-name frontier models for cheaper options that clear the bar.
Bigger players are leaning in too. Microsoft, which has invested more than $13 billion in OpenAI and as much as $5 billion in Anthropic, rolled out a suite of low-cost models this month. Amazon and Google are also expanding lineups aimed at business users who care about price and predictability.
Why the AI Efficiency Shift Matters
If buyers keep optimizing for cost, the economics that justify trillion-dollar valuations could tighten. Frontier labs may face pressure to defend premium pricing, while cheaper and open-weight models absorb routine workloads. Watch for OpenAI and Anthropic to push efficiency features, smaller model tiers, and pricing changes designed to keep customers from drifting to rivals.
The smartest model no longer wins by default. In the next phase of the AI market, the winner may be whichever provider delivers acceptable quality at a price the finance team will approve.