Enterprises aiming to scale artificial intelligence adoption face a significant hurdle in the form of rising token costs, which must drop substantially to support widespread use, Palo Alto Networks CEO Nikesh Arora said. He outlined that token costs should fall by roughly 20% within the next year and by as much as 90% the following year to make AI deployments economically viable on a large scale.
Arora’s remarks came during an interview on CNBC’s “Squawk on the Street,” where he also responded to OpenAI CEO Sam Altman’s claim that OpenAI’s newest AI model improved coding efficiency by 54%. Arora acknowledged this progress but emphasized the need for further improvements to reach a tipping point in enterprise adoption.
The escalating expense of AI usage has already prompted many companies to implement measures aimed at controlling costs. Organizations that initially promoted AI tools during cheaper pricing phases are now tightening usage through caps and encouraging employees to match tools to specific tasks. Other cost-saving strategies include shifting to older, less costly AI models and exploring open-source alternatives, which can lower dependency on pricier proprietary offerings.
Meanwhile, Chinese AI developers are gaining attention for providing more affordable AI solutions, benefiting from efficient models combined with lower energy costs. This competitive edge introduces new dynamics in global AI markets as enterprises weigh cost-efficiency alongside capability.
Several Silicon Valley firms have experienced what has been termed “token shock,” where AI spending quickly outpaces initial budgets. Uber, for instance, exhausted its full-year AI budget by early spring, prompting internal reassessments that contrast the financial impact of AI tokens against traditional engineering hires. This reflects broader industry challenges where advanced AI applications, including agentic coding tools—which involve complex multi-step AI calls—generate significantly greater token consumption than simpler chatbot interactions.
According to the PYMNTS Intelligence "Enterprise AI Benchmark Report," sectors like financial services, insurance, healthcare, media, and advertising are investing more heavily in AI. These industries are increasingly discerning about which AI projects warrant meaningful capital investments versus those still under evaluation for their potential returns.