Tech Companies Ignite Trend of Efficient AI Usage
Well-known tech blogger Gergely Orosz stated that the hottest topic in internal knowledge-sharing meetings at several tech companies has become "how to use AI efficiently."
He recalled that in the past, topics like "accelerating CI/CD" or "optimizing build speed" were never discussed as widely across multiple companies as they are now.
He believes that the main reason behind this is the cost pressure resulting from rising token prices starting to become apparent.
Source: Public Information
ABAB AI Insight
Gergely Orosz has long observed the engineering culture in tech companies, and this observation continues the recent trend of many enterprises shifting from the enthusiastic adoption of AI to cost control. When early AI tools were free or low-cost, companies focused more on exploring functionalities, but as large-scale usage increased, the rapid rise in token consumption costs has become a real pain point.
In terms of capital pathways, companies are reallocating resources from blindly expanding AI usage to internal training, prompt engineering optimization, and model routing choices (such as OpenRouter). Meanwhile, some companies are beginning to evaluate building small models or hybrid deployment solutions, motivated by the goal of controlling overall AI spending by improving single-token efficiency to avoid uncontrolled costs affecting profits.
Similar to previous discussions by companies like Uber and Microsoft regarding the cost pressure of AI coding tools, this wave of "AI cost-saving" trends marks a rapid transition phase for AI applications from expansion to refined cost management.
Essentially, this reflects capital concentration: rising token prices force organizations to focus on efficiency, with the mechanism shifting AI usage from "the more, the better" to "maximizing output per unit cost," accelerating the concentration of capital and talent towards teams and tools that can effectively manage AI costs, and promoting the evolution of AI infrastructure from extensive consumption to high cost-performance refinement.
ABAB News · Cognitive Law
Frenzied usage when free, learning to save money only when charged.
True experts never just focus on what AI can do, but rather on whether each token is worth it.
When the whole company is discussing "how to spend less on AI," the bubble is about to burst.