Key Takeaways
- Nikash Arora, CEO of Palo Alto Networks, predicts “average intelligence is going to be free” and continuously improving, while “exceptional intelligence will be paid for.”
- The cost for basic AI tasks, like a call center query, is dropping to a point where paying "$6 a million tokens" makes no sense for such average needs.
- This creates an "insatiable demand" for compute, driving massive capital expenditure cycles, similar to the 3G/4G/5G telecom buildouts.
- The core inputs for this demand—“land, permits, energy, compute”—are “the thing that is going to get priced for the next 3 to 5 years,” according to Arora.
- A major risk is the "timing problem": whether revenue from AI applications will arrive fast enough to fund the enormous capex required for compute infrastructure.
The AI Economy's Stark Divide: Free vs. Priced Intelligence
Nikash Arora, the CEO of Palo Alto Networks, cut straight to the chase on 20VC: the AI economy is bifurcating. He shared a core soundbite: “Average intelligence is going to be free in the long term and the average intelligence will keep getting better. Nice. Exceptional intelligence will be paid for.” This isn't a vague future state; it's already shaping product roadmaps and pricing models. Think about simple tasks. Arora points out, “I don't think you need to pay $6 a million tokens to answer a call saying how can I help you? I'm so sorry your network connection is not working.” As models improve, baseline AI will become a commodity, almost a utility, pushing companies to define what "exceptional" truly means for their users—and what customers will actually pay a premium for. This insight forces founders to scrutinize their AI offerings: are you building for the free tier, or delivering genuine, differentiated genius?
The New Gold Rush: Compute, Land, and Power
This relentless drive towards free average intelligence and paid exceptional intelligence fuels an insatiable demand for compute power. Arora likens it to the massive capex cycles that built out the 3G, 4G, and 5G telecom networks. He's clear about what becomes scarce, and therefore expensive, in this new era: “Anybody who can produce any energy source it doesn't matter where you are is right now trading in multiple because land permits energy compute this is the this is the thing that is going to get priced for the next 3 to 5 years.” Forget pure software margins; the infrastructure layer is where the pricing action is. Founders who understand this shift will stop viewing compute as a line item to minimize. Instead, they'll see it as a strategic asset, with direct ties to real-world physical resources—land for data centers, energy to power them, and the permits to build. This isn't just about GPUs; it's about owning the real estate, the power grid, and the regulatory approvals.
The Looming Timing Problem: Revenue vs. Capex
While the compute gold rush sounds promising, Arora injects a dose of cold reality: “The biggest only problem we have right now is a timing problem. Would the revenues show up fast enough to keep funding the capex cycle or is there going to be a dislocation in in capex versus outcomes?” Companies are pouring billions into building out compute infrastructure—data centers, power, cooling, and chips. The question isn't if the demand for AI compute is real; it's whether the applications and businesses built on top of that compute will generate revenue quickly enough to justify and sustain those colossal investments. A dislocation between this capex spending and revenue realization could create significant market volatility. This risk forces founders to think critically about their business models. Are your AI solutions generating immediate, measurable value for customers, or are they relying on a "build it and they will come" strategy that might fall victim to a capex-revenue mismatch? The long-term vision for AI is solid, but the next few years will test the financial timing of this capital-intensive revolution.
What to Do With This
Audit your product strategy: precisely define what "exceptional" AI means for your customers and ensure your premium offerings go beyond what's becoming commoditized. Next, immediately map your supply chain for compute. Identify your current and future needs for GPUs, energy, and data center space, then stress-test that supply against potential pricing spikes or scarcity over the next 3-5 years.