When Airtel’s free Perplexity Pro trial ended in India, most analysts braced for a drop-off. Instead, revenue jumped 60%. That’s not a growth signal—it’s a retention signal, and it’s worth dissecting with the same cold precision I apply to on-chain data.
I’ve spent years tracking liquidity flows and wallet behavior. In crypto, the first thing I look for after a hype event is whether the users who came for free yield actually stayed after the rewards dried up. The Perplexity-Airtel story is the same playbook, just with AI subscriptions instead of DeFi tokens. Let me walk you through the data.
Context: The Airtel Distribution Model
Perplexity AI, an answer-engine startup, partnered with India’s telecom giant Airtel in early 2025 to bundle a free Pro subscription with prepaid and postpaid plans. The deal gave Airtel’s 350 million+ subscribers access to Perplexity’s real-time search, GPT-4/Claude model routing, and cited sourcing—all for a limited period. After the trial, users had to pay locally priced tiers (roughly INR 200–300 per month, about one-third of US pricing).
Standard industry wisdom says free trials in price-sensitive markets generate high churn. But here, the opposite happened. Perplexity’s India revenue grew 60% post-trial, while app downloads remained low. That combination screams one thing: the channel is delivering high-quality users who convert, even if the total addressable base is still small.
Follow the gas, not the hype. The gas here is the subscription revenue after the free trial expires. Most analysts would cheer the 60% headline, but I need to see the on-chain evidence—or in this case, the unit economics.
Core: The Evidence Chain
Let’s break down the numbers as if they were wallet addresses. First, the revenue growth is from a small base. Perplexity’s India user count is likely in the hundreds of thousands, not millions. A 60% jump on a tiny number is easy to achieve. The real test is whether this growth can compound without another Airtel promotion.
Second, the low downloads suggest that many users access Perplexity via mobile web or desktop, not the app. This is a known pattern in India: app stores are dominated by free tools, and paid productivity apps often struggle to get installs. But web-based usage can still generate strong retention. I’ve seen this in DeFi—where users interact via dApps on mobile browsers, not native apps. The app download metric is a red herring.
Third, the retention signal is strong. When a free trial ends and revenue goes up, it means users are voluntarily paying. That’s rare in any market, especially one where free alternatives like Google Gemini and ChatGPT exist. The value proposition must be clear: real-time, cited answers that are better than traditional search. I’ve audited enough tokenomics to know that sticky product-market fit is the hardest thing to replicate. Perplexity seems to have found it in India’s information-seeking demographic.
But here’s where I put on my data detective hat. The real question is: what drives the retention? Is it the model routing (access to GPT-4/Claude) or the RAG pipeline (retrieval-augmented generation with citations)? If it’s the former, Google or OpenAI could undercut Perplexity by offering similar models at lower prices. If it’s the latter, Perplexity has a moat—because building a real-time search index with accurate citations is hard infrastructure work.
Whales move in silence. Listen closely. The whales here are the Airtel users who converted. They are not buying the narrative of “AI search”; they are buying the utility of getting verified answers quickly. That utility is sticky, but it’s also expensive to provide.
Contrarian: The Fragility of the Model
Every data point has a flip side. The 60% revenue growth might be masking negative unit economics. AI search queries are expensive—each one requires retrieval, reranking, multi-step reasoning, and citation generation. The cost per query can be 10x higher than a standard chatbot interaction. If Perplexity’s India pricing is too low, every new subscriber could be a loss leader.
Check the supply. Trust the chain. In this case, the supply is the compute and API costs. Perplexity uses third-party models like GPT-4 and Claude, which charge per token. Even with their own Sonar models, the inference costs are non-trivial. If the gross margin is negative, the growth is just burning cash faster.
Another blind spot: the Airtel channel is a single point of failure. If Airtel demands a higher revenue share or switches to a competitor like Google, Perplexity’s user acquisition pipeline dries up. The low download numbers suggest a lack of organic growth—the company is relying on B2B partnerships to find users. That’s fragile.
Liquidity leaves first. Panic follows. If Google AI Overviews becomes the default search experience on Android (which dominates India), Perplexity’s independent search value proposition could evaporate overnight. Users won’t need a separate app if Google provides similar answers for free.
Takeaway: The Next Signal to Watch
I’m not dismissing the 60% growth—it’s a genuine validation that AI search can find a paying audience in a price-sensitive market. But the next quarter’s data will tell the real story. I’ll be watching three metrics: (1) churn rate after the first paid month, (2) gross margin on India subscriptions, and (3) whether Airtel renews the partnership at the same terms.
From my experience auditing ICO tokenomics during the 2017 boom, I learned that sustainable growth requires positive unit economics. Perplexity’s India experiment is a test of whether the “AI search as a service” model can work outside the US. If the retention holds and the margins improve, it’s a signal for the entire industry. If not, it’s a warning to follow the gas, not the hype.
The data is clean. The story is not. Let’s see what the next block brings.