Token Costs Must Drop 90% for Enterprise AI, Says Palo Alto CEO

Enterprise AI adoption hinges on dramatic price cuts

AI token pricing remains the primary barrier to large-scale enterprise deployment, according to Palo Alto Networks CEO Nikesh Arora, who laid out an aggressive timeline for cost reductions during remarks Thursday. He stated that prices must fall to roughly 20% of current levels within a year, and to just 10% within two years, before companies can justify moving AI from pilot programs to production systems.

Arora's comments followed OpenAI's announcement that its latest model achieves 54% better token efficiency on agentic coding tasks. While acknowledging the improvement, Arora characterized it as insufficient. "I think 54% is a good start," he said. "I think we probably need another turn at it."

Despite the pricing concerns, Arora expressed confidence that market forces would eventually correct the imbalance. "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," he said, suggesting that model efficiency improvements would gradually ease corporate spending pressures.

Why it matters

The gap between falling unit prices and rising total costs reveals a fundamental challenge in enterprise AI economics. While per-token rates have dropped 98%, enterprise AI spending has tripled because agentic applications chain multiple model calls together, multiplying consumption faster than efficiency gains can offset. This dynamic threatens to keep AI confined to limited use cases rather than enabling the transformative enterprise-wide deployments vendors have promised.

Growing executive pushback on model pricing

Arora joins a expanding group of corporate leaders questioning current AI pricing structures. Palantir Technologies CEO Alex Karp criticized the per-token approach used by Anthropic and OpenAI last week, advocating for open-weight models as a more sustainable enterprise path. "Something has gone completely wrong," Karp told CNBC, though he avoided naming specific vendors.

The pricing pressure is already reshaping corporate AI strategies. Companies including Uber and Microsoft have imposed caps or restrictions on employee access to expensive AI coding tools after budgets exceeded projections. Other firms are shifting toward cheaper open-weight alternatives, including Chinese models that are narrowing the performance gap with American labs, according to CNBC.

Infrastructure spending continues unabated

The pricing friction has not slowed capital deployment in AI infrastructure. SpaceX raised $25 billion through debt markets last month, and Amazon followed with a $25 billion bond offering this week, with both transactions linked to AI infrastructure demands, CNBC reported.

The tension between rising costs and infinite demand suggests the enterprise AI market remains in a transitional phase, with pricing models still seeking equilibrium between provider economics and customer budgets.

These details were first reported by AI Watch and CNBC.