Enterprises Slash AI Spending as Token Costs Spiral in 2024

According to Barron's, enterprises are reassessing their AI investments after facing unexpected Token cost overruns in 2024, with pricing transparency and budget control becoming major challenges across Wall Street.

Reasoning models and AI agents are the primary cost drivers. Reasoning models perform lengthy internal computations before generating outputs, consuming far more Tokens than the final text produced. AI coding agents are even more costly, requiring up to 1,000 times more Tokens than human programmers to complete equivalent tasks. Many companies are now implementing dashboards to monitor employee AI usage and shifting toward more cost-efficient models, including lower-cost alternatives from China or waiting for price cuts from major providers. Adding to the complexity, different model providers count Tokens differently—Anthropic's counting method shows 30-40% higher usage than competitors—making it difficult for analysts to track AI adoption trends.