Businesses are confronting significantly increased AI costs, driven by a recent shift from flat-fee to usage-based pricing by major AI providers and the growing popularity of AI agents, which consumer far more resources than basic chatbots. This has led to companies, such as Uber, exhausting their annual AI budgets early in the year, and nearly half of 2,145 global business leaders surveyed by KPMG reported scaling back AI agent use because costs outweighed benefits. Sam Altman, OpenAI chief, noted that AI cost overruns have become a "huge issue" since early 2026.
In response to these budget challenges, companies are implementing various strategies to rein in spending. This includes adopting open-source AI models, which can cut costs by up to 70% and offer greater control over data security, especially for sectors like banking and telecommunications. Enterprise software company Atlassian has imposed limits on the number of "tokens" employees can use, requiring managerial approval for additional usage. Other firms, such as Notion, are developing routing tools to select the most cost-effective AI model for specific tasks.
Despite these efforts to control costs, overall AI spending continues to rise. Goldman Sachs predicted a 24-fold increase in global token consumption by 2030, largely due to AI agents. Companies are also concerned about vendor lock-in, where reliance on a single AI provider can lead to difficulties if prices increase or performance declines. This has prompted some, like those turning to Chinese AI models, to diversify their AI sources. Mantas Lukauskas of Hostinger highlights the importance of choosing the right model for the task, noting that some are better for coding, while others excel at creative writing or image generation. Companies are emphasizing linking AI use to work outcomes to ensure value for money.
