Palo Alto Networks (PANW) CEO Nikesh Arora pushed back against the idea that AI is already cheap enough for widespread adoption, speaking on CNBC’s “Squawk on the Street” Thursday. Arora said token costs need to come down as much as 90% before AI can see truly large-scale adoption. His comments followed OpenAI CEO Sam Altman telling CNBC that OpenAI’s newest model is 54% more token-efficient for agentic coding tasks. Arora responded: “I think 54% is a good start. I think we probably need another turn at it,” adding that token efficiency needs to improve by another 20% over the next 12 months, then keep climbing toward that 90% reduction target within two years. Arora stated, “We need to see the pricing for AI come down.” Palantir CEO Alex Karp made similar comments last week, criticizing the token pricing model and suggesting open-weight models could be a better path. Pricing pressure is pushing some businesses toward cheaper open-weight alternatives, including Chinese models. Meanwhile, AI infrastructure spending continues to climb; SpaceX raised $25 billion through a bond sale last month, and Amazon raised another $25 billion in debt this week for AI investments. Arora said, “the demand continues to be infinite, and as long as you have an infinite demand curve that you’re facing, I think all these things will rationalize over time.” PANW stock price as of Jul