Everyone says AI search is the next frontier. The metrics they point to are user growth, download counts, and total queries. But the real signal—the one that reveals the underlying mechanics—is what happens when the free ride ends. Perplexity's 60% revenue jump in India after the Airtel promo expired is not a victory lap. It's a data point that demands a deeper audit.
Let me state this plainly: I have audited smart contracts during the 2017 ICO boom. I have watched delta-neutral strategies collapse when the underlying volatility surface shifted. I have learned that the most dangerous narratives are the ones that feel intuitively correct. The narrative here is that Perplexity has cracked the Indian market. The reality is that they have found a crack in the wall—but the wall is still standing.
Context: The Airtel Experiment
Airtel, India's second-largest telecom operator, has a history of bundling third-party services—Netflix, Prime Video, Zee5—into its prepaid and postpaid plans. The model is simple: subsidize the subscription to drive stickiness and data usage. In early 2025, Airtel offered Perplexity Pro to its users for a limited period. The promo ended. Revenue in India grew 60% in the subsequent quarter.
The immediate reaction from the tech press was predictable: "Perplexity proves AI subscriptions work in emerging markets." But I smell a different odor. It smells like a liquidity trap disguised as a volume breakout.
Core: The Mechanical Arbitrage of Retention
Let's dissect the 60% figure. The base was small. Perplexity's app downloads in India are negligible compared to ChatGPT or DeepSeek. A 60% jump on a low base is not a home run—it's a single. The real story is the conversion rate from free promo to paid. That conversion indicates that the product offers something that free alternatives do not.
What is that something? It is not the underlying model. Perplexity routes queries through GPT-4, Claude, and its own Sonar series. The differentiation is the retrieval-augmented generation (RAG) layer: real-time search, multi-step reasoning, and citation tracing. In a market where users are starved for reliable, structured information—especially in local languages—this is a genuine value-add.
But here is the critical part: each query costs more than a standard chatbot interaction. The retrieval, re-ranking, generation, and citation stack burns through compute. At Indian pricing—roughly one-third of US rates—the unit economics are thin. The hidden assumption is that these users are querying frequently enough to generate high lifetime value.
Code is law, but bugs are justice. The bug here is that the cost structure is opaque. If the average Indian user runs 20 queries per day, the margin per user is negative. Revenue growth becomes a mirage. The only way to maintain profitability is to either increase prices (which kills retention) or reduce model quality (which kills the product).
Contrarian: The Retail vs. Smart Money Divergence
Retail traders see 60% growth and buy the narrative. Smart money looks at the channel concentration. Airtel is the sole distribution partner. Download counts are low, meaning organic growth is nearly zero. The revenue spike is entirely dependent on a single telecom partnership. If Airtel renegotiates the revenue split—or if Jio launches a competing AI bundle—the entire Indian revenue stream becomes toxic.
Greeks don't lie, but they can mislead if you ignore the underlying volatility surface. The implied volatility of Perplexity's future revenue in India is high because the base is small and the distribution channel is fragile. The 60% figure is a point estimate with a wide confidence interval. The next quarter's numbers will tell the real story. If revenue flattens or dips, the spike was a one-time conversion of promo users. If it continues to grow, then the unit economics might be sustainable.
There is also a structural risk that the market is ignoring: Google AI Overviews. If Google embeds AI search directly into its core search results—which it is already doing—the independent AI search app becomes redundant. The same Airtel user who is now paying for Perplexity will see no reason to pay when Google provides a similar experience for free.
Takeaway: The Takeaway Is Not a Conclusion
I am not saying Perplexity is failing. I am saying that the 60% growth signal is being misinterpreted as a trend when it is a data point. The only way to validate the thesis is to watch the next two quarters. If retention holds and more telecom partnerships emerge, then the model is real. If not, the revenue spike will be remembered as a footnote in a failed expansion.
NFT floor is a feeling, not a number. The same applies to revenue growth in emerging markets. The number is real, but the feeling behind it—the belief that the market is ready to pay for AI search—is still unproven. I will be watching the options chain on Perplexity's token (if they ever issue one) and the volume of short interest in the broader AI sector. The market never discounts the obvious. It discounts the hidden mechanics.
Here is the trade: I am short the narrative, long the data. Wait for the next quarterly report. Then decide.
