Nikesh Arora said businesses need far cheaper AI usage pricing over the next two years, joining a broader push from executives including Palantir's Alex Karp as infrastructure spending climbs.
Palo Alto Networks CEO Nikesh Arora said AI token costs need to fall sharply to unlock broad enterprise adoption, arguing current pricing is making the tools too expensive for many businesses to deploy at scale. Speaking on CNBC on July 9 and building on remarks he made on the 20VC podcast on June 22, Arora said OpenAI's latest model being 54% more token-efficient for agentic coding was progress but not enough. He outlined a path of about 20% additional efficiency improvement over the next 12 months and a full 90% reduction in prices within 24 months. The comments add to a growing debate over AI economics as companies weigh rising inference costs and consumer AI losses against heavy infrastructure investment. Palantir CEO Alex Karp has also criticized token-based pricing and backed more open alternatives, saying enterprises are hesitant under the current cost structure. The discussion matters beyond software companies because major spending commitments, including Amazon's recent $25 billion bond sale tied partly to its AI buildout, will need to be supported by business models that can eventually generate returns.