UBS Group reported that while artificial intelligence (AI) spending remains robust, enterprises are increasingly prioritizing return on investment (ROI) and efficient AI expenditure. Analysts led by Karl Keirstead attended the Ai4 conference in Las Vegas, engaging with approximately 15 enterprise IT executives and industry partners. The research indicates many enterprises have moved beyond 'maximizing AI tokens' into a 'token optimization' stage, where efficiency and ROI are prioritized. A shared concern is whether this emphasis could dampen AI expenditure growth, impacting frontier labs such as Anthropic and OpenAI, though the shift has not yet translated into any decline in AI spending. One primary method for reducing AI task costs is routing tokens to open-source models, including NVIDIA’s Nemotron or Chinese AI models, but willingness to adopt Chinese models was low among the large non-tech enterprises surveyed. Keirstead noted that many Fortune 500 companies expressed a preference for 'building their own AI products,' a trend exerting downward pressure on AI product growth for SaaS and application software firms. The analysts rarely encountered on-premises AI deployments; instead, AI solutions are almost always hosted on Amazon Web Services (AWS), Microsoft Azure, and Alphabet Google Cloud.