Palo Alto’s CEO: AI Token Prices Must Fall Up to 90%
Nikesh Arora, CEO of Palo Alto Networks, welcomes OpenAI’s recent efficiency gains with its GPT-5.6 model, which is 54% more token-efficient for agentic coding. However, he argues that current token prices need to drop by as much as 90% for large-scale enterprise adoption.
Arora’s position stems from a perceived paradox: while per-token prices have collapsed, total enterprise AI bills have been rising due to the increasing usage of agentic AI models. He believes that with an "infinite" demand for AI, costs will "rationalize over time."
His logic suggests that either the market will grow to accommodate higher spending or prices will be forced down as technology becomes more efficient. Despite the high initial costs, Arora remains bullish on the future of AI, stating that the demand is insatiable.
The current situation presents a challenge for enterprises. While per-token costs have decreased dramatically, usage patterns in agentic AI can still lead to exorbitant bills. Some companies are already capping AI usage due to these rising costs. Arora’s call for lower prices reflects a widespread frustration among businesses seeking to integrate AI into their operations cost-effectively.
A price war is currently underway, with startups and existing providers offering significant discounts on inference services. This competition aims to address the high costs of AI deployment. However, achieving a 90% reduction in token prices remains uncertain, as efficiency gains might be counteracted by increasing usage patterns.