Analysis: Spending tightening impacts growth expectations of OpenAI and Anthropic, marking the beginning of a cost-efficiency era in the AI industry.

Odaily reports that as enterprises begin to reassess their AI investment returns, the industry is shifting from a "tokenmaxxing" high-consumption model to a focus on efficiency, with AI large model providers facing new growth constraints. Several companies have already started reducing or optimizing the cost of model calls. For example, Lindy CEO at an AI startup said that they have switched 100% of their traffic from Anthropic's Claude model to the more cost-effective DeepSeek, expecting to save millions of dollars in expenses within a few months.

This shift reflects that enterprise AI budgets are tightening, and the previous "unlimited use of model resources" tokenmaxxing approach is gradually being replaced by cost control and ROI-focused strategies. Some companies have even set tiered budgets for AI tool usage; for instance, Uber has set a monthly limit for internal AI expenditures.

Analysts point out that as enterprises move from "expanding usage" to "refined calling," the rapid growth models previously relied upon by OpenAI and Anthropic now face challenges. Industry data still shows strong growth: Anthropic’s annualized revenue is approximately at the $4.7 billion level, while OpenAI’s run rate is nearly $2.5 billion, but the market is beginning to focus on the sustainability of this growth.

Meanwhile, methods for model calls are changing, with technologies like "model routing" emerging—using low-cost models to handle simple tasks instead of high-end models, in order to optimize overall computing power costs. Industry competition is also intensifying, as Microsoft, Amazon, and Google are all accelerating the launch of low-cost AI models and enterprise-grade tools, further compressing price ranges. In the context of more rational corporate AI spending, large model companies may face both "growth slowdown expectations" and "IPO window pressure." (CNBC)