AI TRENDS | Companies Shift AI Spending to Efficiency as Model Providers Face Growth Constraints

Companies are reassessing returns on AI spending and moving away from high-consumption “tokenmaxxing” toward efficiency-first usage, creating new growth constraints for large-model providers. According to Odaily, multiple firms have begun cutting or optimizing model-calling costs.

Lindy CEO said the AI startup shifted 100% of its traffic from Anthropic’s Claude to the lower-cost DeepSeek and expects to save millions of dollars in expenses within a few months.

The report said corporate AI budgets are tightening, with cost controls and ROI-focused approaches replacing previously unlimited model usage. Some companies have introduced tiered budgets for AI tools; Uber has set a monthly cap on internal AI spending.

Analysts said that as enterprises move from expanding usage to more refined, selective calling, the high-growth model previously relied on by OpenAI and Anthropic is facing challenges. Industry data still indicates strong growth, with Anthropic’s annualized revenue at about $47 billion and OpenAI’s run rate near $25 billion, though attention is turning to the sustainability of that growth.

Model usage patterns are also changing, with techniques such as “model routing” gaining traction to assign simpler tasks to lower-cost models and reduce overall compute costs. Competition is intensifying as Microsoft, Amazon, and Google accelerate the rollout of lower-cost AI models and enterprise tools, further compressing pricing. The report added that large-model companies may face both expectations of slower growth and IPO window pressure as enterprise AI spending becomes more rational.