Executives at large companies are becoming increasingly mindful of escalating AI-related expenses, with roughly 60% of enterprises now implementing some form of guardrails to manage or throttle their spending, according to recent conversations with IT leaders compiled by UBS analysts.
In a report released this week, UBS analysts Karl Keirstead, Timothy Arcuri, Taylor McGinnis, and their colleagues highlighted a noticeable shift in corporate attitudes toward AI investments. Token spending, the unit by which many AI services are priced, has emerged as a significant concern, particularly for larger organizations where chief financial officers and chief technology officers are seeing their AI budgets balloon without always delivering clear returns.
Uber’s operations chief, Andrew Macdonald, captured the growing sentiment in May when he noted it was becoming harder to justify rising costs given the relatively modest return on investment observed so far.
The analysts, drawing from more than a dozen discussions with enterprise IT executives over recent weeks, described a “modest emerging headwind” in AI spending patterns. Follow-up conversations reinforced this view, though the degree of impact varies considerably across organizations.
“Token spend optimization has become a key issue in most organizations, resulting in a big spending speed bump for some organizations, but a smaller speed bump for others that are either too early in their AI deployments or are far deeper but are unwilling to throttle users because they see the offsetting ROI or have an organizational priority in place to drive innovation and hence AI use,” the analysts wrote.
AI model providers such as OpenAI and Anthropic are likely to feel the effects most acutely in the near term, the report suggested. Open-source and Chinese models, such as DeepSeek, could emerge as notable beneficiaries, especially for enterprises seeking cost-effective solutions for non-coding tasks.
Despite the emerging restraint, the analysts emphasized they are not sounding any alarms, viewing the trend as “a healthy problem” that reflects more disciplined management rather than a fundamental retreat from AI.
“Some measure of AI spend optimization is normal, no one is hitting the brakes on AI deployment and it is likely that we’re sitting in front of new models trained on next-gen chips that might drive token costs down further,” they wrote.
Leading AI companies have already begun highlighting improved token efficiency in their latest models. Google has its Gemini 3.5 Flash offering, while Anthropic recently launched Claude Sonnet 5, which the company said “runs autonomously at a level that just a few months ago required larger and more expensive models.”
From Experimentation to Engineering Discipline
Conversations with executives reveal a clear evolution in how companies approach AI. One organization told UBS analysts the industry is moving beyond the initial phase of broad experimentation.
“The question isn’t whether to use tokens, it’s how to use them efficiently,” the analysts quoted one executive as saying. “As a result, optimization becomes an ongoing engineering discipline rather than a reaction to a budget crisis.”
Another company described a situation where its CTO had enthusiastically embraced multiple AI tools early on, only to face budget constraints later.
“We have 5 AI tools internally and all of the LLM products. Like others, we ran into the issue where we have already used most of our token budget for the entire year,” the analysts recounted. “Now we’re only using 2 AI tools and being careful around usage.”
This shift toward greater cost consciousness does not necessarily signal a slowdown in overall AI adoption. Instead, it points to a more mature phase where organizations are becoming savvier about where and how they deploy the technology. Companies are increasingly prioritizing high-impact use cases while optimizing spending on lower-value applications.
The analysts noted that while some organizations are actively throttling usage, others, particularly those earlier in their AI journeys or those with strong innovation mandates, continue to invest aggressively. This variation suggests the overall AI spending trajectory remains upward, albeit with greater scrutiny and efficiency measures in place.
Implications for AI Providers and the Broader Ecosystem
For AI model makers, this development is expected to reshape competitive dynamics. Providers offering more token-efficient models or flexible pricing structures may gain an edge, while those relying on premium pricing for cutting-edge capabilities could face margin pressure if enterprises become more selective.
Open-source alternatives and models from Chinese developers are particularly well-positioned to capture share in cost-sensitive segments. Analysts expect this to accelerate the commoditization of certain AI capabilities and intensify competition across the industry.
At the same time, the focus on optimization is expected to drive innovation in areas such as model compression, efficient inference, and specialized AI agents designed for specific business functions. Companies that can deliver strong performance at lower token costs may find themselves better positioned as enterprises move from experimentation to scaled deployment.
However, the broader takeaway from UBS’s conversations is one of pragmatic maturation rather than disillusionment. Enterprises are not abandoning AI — they are becoming more sophisticated in how they implement and manage it. This evolution could ultimately strengthen the technology’s long-term adoption by ensuring investments are more closely tied to measurable business outcomes.
