For companies to deploy artificial intelligence at scale, token costs need to come down, Palo Alto Networks CEO Nikesh Arora said on Thursday (July 9).
Costs need to fall by 20% over the next 12 months and 90% by next year, Arora told CNBC’s “Squawk on the Street,” CNBC reported Thursday.
Asked about OpenAI CEO Sam Altman’s comments to CNBC that OpenAI’s latest model was 54% more efficient at coding, Arora said, “I think 54% is a good start. I think there’s probably one more thing that needs to be done.”
In June, it was reported that companies are looking to better manage their use of the technology as they see the cost of AI rising.
Companies that encouraged employees to use AI tools when costs were lower are now reducing costs in a variety of ways, including implementing usage caps, encouraging employees to use the right tools for each task, sharing cost-saving ideas such as switching to older, cheaper models, and adopting open source models.
The report also found opportunities for Chinese AI labs, which can pay less than U.S. companies due to more efficient models and China’s lower energy costs.
In May, it was reported that a “token shock” had hit some of Silicon Valley’s biggest spenders.
For example, Uber will burn through its full-year 2026 AI budget by April, with Chief Technology Officer Praveen Nepali Naga saying the company is “back to square one,” and Chief Operating Officer Andrew McDonald saying Uber will directly weigh the cost of tokens against the cost of hiring engineers.
PYMNTS reported at the time that agent coding tools increase the cost burden compared to standard chatbot interactions because a single turn conversation generates one inference call, whereas an agent session generates many more inference calls.
The PYMNTS Intelligence report, “Enterprise AI Benchmark Report: Financial Services Leads the Way in the Enterprise AI Race,” finds that companies in financial services, insurance, healthcare, and media and advertising are putting more money into AI.
In doing so, these companies are beginning to determine which projects are worthy of real capital and which still require proof, the report said.
