AI is driving rising costs for companies as AI agents run tasks and consume tokens, prompting a shift from subsidized access to cost-aware usage.

OpenAI and Anthropic aim to go public, signaling a path toward profitability for major AI players amid mounting costs.

Industry watchers note a broader shift to sustainable AI usage in enterprises, with emphasis on cost control and practical productivity gains.

Note: The article excerpt continues behind a paywall, with additional details unavailable here.

Executives weighed in on the economics, including cautions about subsidized intelligence, rising coding costs, and token-driven productivity claims.

Corporates are responding by urging prudent AI use, while some question whether AI spending translates into real productivity gains.

Observers warn of a potential disconnect between AI spending and productivity, pointing to Meta, Uber, and others as examples.

Token pricing varies widely between large monolithic models and smaller models, with costs potentially dropping from around $15 per million tokens to five cents for smaller models.

The cost gap between big models and smaller models remains substantial, with smaller models offering much lower per-token expenses.

The phenomenon of tokenmaxxing describes binge usage where token costs can outstrip employee costs in a short period.

Tokenmaxxing illustrates excessive internal AI usage contributing to runaway costs before efficiencies catch up.

A trend is emerging to treat AI as a commodity, prioritizing cost-effective model selection over always using the most advanced options.