Palo Alto Networks CEO Demands 90% AI Token Price Cut

What happened

Palo Alto Networks CEO Nikesh Arora publicly stated that AI token prices need to drop dramatically for enterprises to adopt the technology at scale. Speaking on CNBC, Arora called for token costs to fall to roughly 20% of current levels within the next 12 months, and down 90% within 24 months. He acknowledged OpenAI's recent 54% improvement in token efficiency for agentic coding as a positive step, but emphasised that much steeper reductions are required. Arora's remarks came the same day OpenAI released its GPT-5.6 model family, which CEO Sam Altman promoted for its improved efficiency. Meta also launched its Muse Spark 1.1 agentic coding model, priced at $1.25 per million input tokens and $4.25 per million output tokens, positioning it competitively alongside budget offerings from Anthropic and OpenAI.

Why it matters

Rising token costs have become a critical pain point for enterprises attempting to deploy AI tools, and Arora's comments reflect mounting pressure on AI providers from major corporate buyers. Palo Alto Networks, valued at approximately $248 billion, integrates AI into its cybersecurity products and buys inference services at industrial scale, giving Arora direct exposure to pricing challenges. His call for a 90% price cut within two years signals that current economics may be unsustainable for widespread adoption, forcing businesses to limit AI deployments or seek cheaper alternatives. Other executives, including Palantir CEO Alex Karp, have echoed similar concerns about token pricing, warning that high costs are driving enterprises toward open-weight models and cheaper Chinese alternatives. For Palo Alto Networks specifically, reducing AI costs would improve the economics of its own product integrations and potentially lower operational expenses tied to AI-powered security features.

Bigger picture

Arora's comments land amid intensifying competition and pricing pressure across the AI industry. Companies like Coinbase have already cut internal AI spending by nearly half by switching to cheaper open-weight models, while Microsoft has reportedly considered hosting DeepSeek variants for internal agentic tools. Meta CEO Mark Zuckerberg broke a three-year social media silence to promote Muse Spark's competitive pricing, underscoring how cost has become a key battleground. OpenAI has delayed its IPO until 2027 and has reportedly weighed drastic price cuts to retain enterprise customers against Anthropic, while JPMorgan analysts have warned that falling prices could undermine revenue models even as demand grows. Meanwhile, infrastructure spending continues unabated. Amazon raised $25 billion in debt partly for AI buildout, SpaceX completed a $25 billion bond sale, and SK Hynix executed a $26.5 billion US listing to fund AI chip production. The contrast is stark: companies selling AI infrastructure are raising record capital at premium valuations, while companies selling AI services face mounting pressure to slash prices.

What to watch

Investors should monitor whether OpenAI, Anthropic, and other frontier labs announce further pricing cuts or efficiency improvements in the coming quarters. Watch for signs that enterprises are accelerating adoption if token costs decline meaningfully, or conversely, continuing migration to cheaper open-weight alternatives if pricing remains elevated. Palo Alto Networks' own quarterly results may offer clues about how AI integration costs are affecting its margins and product economics. Broader industry signals include whether hyperscalers like Amazon, Microsoft, and Google adjust their AI infrastructure spending plans if token pricing collapses faster than anticipated, and whether decentralised inference networks lose competitive positioning if incumbent providers achieve the 90% cost reductions Arora outlined. Finally, track whether other enterprise software CEOs publicly join Arora and Karp in pressuring AI vendors on pricing, as widespread executive pushback could accelerate deflationary dynamics across the sector.