MarketMoodz
Arora Says AI Token Prices Must Fall 90% to Scale
Palo Alto Networks CEO Nikesh Arora told CNBC that token costs need to drop as much as 90% to enable large-scale enterprise AI adoption, arguing current pricing is straining corporate budgets. He cited OpenAI's frontier model as a step toward efficiency — but said far steeper token-cost reductions are required for broad deployment.
Key Takeaways
- Nikesh Arora told CNBC token prices must fall up to 90% to make large-scale AI affordable for enterprises.
- Arora said OpenAI's frontier model is about 54% more token-efficient for agentic coding, but more gains are needed.
- He projects token efficiency to improve to roughly 20% within 12 months and 90% by the following year.
- Rising token costs are squeezing AI budgets, pushing some firms toward cheaper open-weight models and alternative pricing.
- Industry figures like Palantir's Alex Karp have publicly criticized token-based pricing as a barrier to enterprise adoption.
People Involved
- Nikesh Arora CEO, Palo Alto Networks
- Sam Altman CEO, OpenAI
- Alex Karp CEO, Palantir
Entities Involved
- Palo Alto Networks (PANW) Cybersecurity company and enterprise AI vendor; Arora is CEO
- OpenAI AI developer; its frontier model was cited for token-efficiency gains
- Palantir (PLTR) Enterprise software company; its CEO has criticized token-based pricing
MarketMoodz Analysis
Token pricing sits at the center of enterprise AI economics: every additional token equals incremental cost on large-scale deployments. If token prices don't fall materially, CIOs and CFOs will throttle projects or redirect spend from implementation to unit costs. Arora's 90% figure is a directional call to action — it signals that vendors and model developers must prioritize efficiency, alternative pricing models, or risk slower enterprise uptake and greater migration to cheaper open-weight models.
Historical parallels help frame the stakes. Cloud compute and storage prices fell dramatically as scale and competition increased, enabling new use cases and budget reallocation; a similar trajectory for token costs would unlock deployment at scale. That said, Arora's projections and the specific efficiency numbers (for example, the 54% OpenAI frontier figure and the timing of 20% then 90% improvements) come from his CNBC interview and have not been independently verified here, so treat the timeline as aspirational rather than guaranteed.
What investors should watch: vendor pricing announcements and new billing models, OpenAI's and other providers' published token-efficiency benchmarks, enterprise AI adoption metrics reported by software vendors, and any pushback from large customers on per-token billing. Also monitor vendor margins and contract renegotiations — a sustained slide in token prices would pressure AI-native revenue models but broaden total addressable market by shifting budgets toward integration and use rather than raw token spend.
Source: Original Article