UBS Reveals: Companies Cut Back on AI Spending

⚡

Key Takeaways

1UBS found that 60% of companies are cutting their AI spending while still continuing their deployment.

2Rising costs of AI tokens are worrying leaders, prompting a push for spending optimization.

3Open-source and Chinese models could benefit from these budget cuts.

💡Why it matters — This trend indicates a strategic adjustment by companies in response to AI costs, influencing future technology choices.

Companies Moderate Their AI Investments

Analysts at UBS have recently highlighted a notable trend in the field of artificial intelligence. During discussions with various companies, they observed that these firms are beginning to moderate their spending on AI. While this may seem alarming, it is crucial to note that no company has completely halted its AI projects.

This observation underscores a shift in companies' approach to AI. The days of companies investing recklessly in AI tokens appear to be over. Business leaders are now more mindful of the expenses associated with this technology.

Revealing Figures

According to a report written by analysts Karl Keirstead, Timothy Arcuri, and Taylor McGinnis, approximately 60% of companies have decided to slow down their AI spending. This figure is based on a dozen conversations with IT executives conducted over the past few weeks. These companies are implementing safeguards to better control their expenditures.

The issue of costs related to AI tokens has become a major concern, particularly for large enterprises. Chief financial and technical officers are noticing an increase in AI bills, making it difficult to justify these costs in light of often limited returns on investment. Andrew Macdonald, Chief Operating Officer at Uber, had already highlighted this difficulty last May.

A Modest Headwind

Discussions that began in early June have allowed analysts to detect what they call a "modest headwind." While the impact of this slowdown varies from company to company, the trend is confirmed by recent conversations. Some companies continue to invest in AI as they see compensatory returns on investment or have organizational priorities that encourage innovation.

Optimizing spending on tokens has become a central issue for most organizations. Some companies are significantly slowing their expenditures, while others, either at the beginning of deployment or already advanced, continue to invest to drive innovation.

AI Model Manufacturers Under Pressure

AI model manufacturers, such as OpenAI and Anthropic, may be the most affected by these short-term budget cuts. Analysts specifically mentioned that open-source and Chinese models, like DeepSeek, could benefit from this situation, especially for companies seeking solutions for non-programmatic tasks.

Despite these adjustments, UBS analysts do not view this trend as alarming. They describe it as a "healthy problem," emphasizing that optimizing AI spending is a normal practice. They add that it is likely that new models, trained on next-generation chips, could further reduce token costs.

Towards a New Generation of Models

Major AI companies continue to tout the efficiency of their models regarding tokens. Google, for example, has developed the Gemini 3.5 Flash model, while Anthropic recently launched Claude Sonnet 5, a more autonomous and cost-effective model.

One company interviewed by UBS indicated that the industry is moving away from the experimentation phase of AI. The focus is now on the efficient use of tokens, transforming optimization into a continuous engineering discipline.

Reduction of Tools Used

Another testimony gathered by UBS reveals that a company, initially very engaged in AI, had to reduce the number of tools used. With five internal AI tools, it has already consumed most of its token budget for the year. Now, it is limiting itself to two tools while closely monitoring their usage.