Palo Alto Networks CEO Nikesh Arora says the cost of using artificial intelligence—particularly the price of tokens—needs to fall substantially before businesses widely adopt AI systems. Speaking to CNBC, Arora argues that sky-high token costs can limit practical deployment, slowing the pace at which companies can integrate AI into products and operations. He estimates that token prices may need to drop by as much as 90% to unlock large-scale adoption. The Next Web reports that Arora made the comments in response to claims about OpenAI’s newer GPT model, noting that even incremental improvements in model capability may not translate into broad enterprise use if running costs remain too high. Overall, both outlets frame the remarks as a cost-driven adoption barrier and present Arora’s “90%” figure as a rough threshold for more scalable deployment. The statements are focused on token pricing and the economics of running AI workloads rather than changes to specific model performance or features.