The head of a major cybersecurity firm is calling for a dramatic overhaul of artificial intelligence pricing. Nikesh Arora, CEO of Palo Alto Networks, argues the current cost structure for AI tokens must fall by 90% within three to five years to enable large-scale enterprise adoption.
Arora stated during a CNBC interview that rising inference costs and losses on consumer AI products are actively discouraging the business deployments the industry needs. He pointed to efficiency improvements, like OpenAI's latest model showing 54% better token efficiency for coding tasks, as proof that such cost reductions are possible.
His call echoes concerns from other tech leaders. Palantir CEO Alex Karp has also criticized token-based pricing, linking it directly to hesitation from corporate customers.
The statement carries significant weight because Palo Alto Networks sits at the nexus of enterprise software and AI, integrating the technology directly into its security products. Arora's view is grounded in the practical, quarterly costs of running AI at scale.
The demand for lower prices creates new pressure on the entire AI infrastructure chain. Projects offering decentralized compute services, which market themselves on cost savings, could see their advantage shrink. Meanwhile, massive investments like Amazon's recent $25 billion bond sale for AI infrastructure hinge on future returns that could be jeopardized by a rapid price collapse.