How Rahul Agarwalla of SenseAI Ventures Predicted India's AI Unicorn Boom and Why He Believes 100 More Are Coming
In a conference room in Gurgaon, Rahul Agarwalla pulls up a spreadsheet that would make most venture capitalists envious. The numbers tell a story of prescience that borders on prophetic: while the world was still debating whether artificial intelligence was overhyped, his AI-first fund SenseAI Ventures had already been quietly building a portfolio that would vindicate his boldest predictions.
The year was 2017. ChatGPT was still six years away. Most investors were focused on fintech and e-commerce. But Agarwalla was making an audacious claim: India would produce 100 AI unicorns within a decade.
People used to look at me like there would be all kinds of skepticism around it. It's too small a world. Today, everything is here.
Fast forward to 2025, and the data speaks volumes. India now boasts 118 unicorns, with AI emerging as the dominant force. Global AI funding hit a record $131.5 billion in 2024, up 52% year-over-year, with AI companies representing 35.7% of all global deal value. Six new Indian unicorns emerged in 2024 alone.
Check out the video of the conversation here, or read on for insights.
The Pattern Recognition Pioneer
Agarwalla's journey to becoming India's AI oracle began with a gutsy decision in 1996. Fresh out of business school, he chose a risky internet startup over a secure corporate job when most people were still trying to understand what this global network meant for business.
I went and did an internet startup. There were no VCs actually to approach in 1997. None.
His first venture, Matrix Information, became India's pioneering digital content aggregator, serving the dot-com boom by syndicating content from 200 providers to portals desperate for pages to drive traffic.
But it was his second venture that would prove prophetic. In 2002, Agarwalla tackled an impossibly specific problem: Japanese text has no spaces. For computers processing language, this created a nightmare. Where does one word end and another begin in a sentence with 30 kanji characters?
So a single kanji can be a word. Of course, combinations are also words. So if you have a sentence which has 30 kanji on it, it could logically be 30 words. Of course, it's not going to be 30. But is 7, 8, 12, 15? It was very hard for computers to figure out then.
The solution: machine learning algorithms that broke Japanese sentences into constituent words through a process called tokenization. Today, anyone paying for OpenAI's services understands tokens as AI's currency. What Agarwalla solved in 2003 became the foundation for every large language model powering today's AI revolution.
This breaking up that Japanese sentence into constituent words is called tokenization. And now you see, now you know why you use the word tokens all the time and why a token is not exactly equivalent to a word.
The Four-Layer Framework Revolution
By 2017, Agarwalla had crystallized his AI understanding into a framework guiding SenseAI's strategy. He envisioned four layers: infrastructure (data centers, hardware), foundational models (OpenAI, Anthropic), tooling (security, observability), and applications (user-facing solutions).
So we actually focus on the top two layers, which is applications and a majority of our investment will go into applications because we think the applications for the next era are yet to be built.
This wasn't just theory. Agarwalla identified a crucial insight about value creation: infrastructure requires massive capital but provides limited returns, while applications create disproportionate wealth.
Most of the wealth in the internet era was created by applications. You go back, you take electricity. The factories that got built, which ran on electricity, actually produce much more value than the electricity companies.
The data vindicated this thesis. Infrastructure investments provide steady yields but limited multiple expansion, while AI application companies achieved unprecedented metrics: $100 million revenue with just 20 employees, as demonstrated by companies like Lovable.
The Variety Principle: AI's Secret Success Metric
While others focused on technical capabilities, Agarwalla developed a fundamental filter: variety. This insight came from studying a traffic optimization system in Las Vegas that used AI to count cars and adjust signal timing.
It basically puts a camera on top of each traffic signal, which counts the number of cars. And based on the count of cars, it changes the light from red, yellow and green. So it has one input variable.
He recognized this as a low-variety problem masquerading as an AI opportunity. A basic algorithm could achieve 98% of the same results without expensive AI infrastructure.
As variety multiplies, you'll find that apart from human beings, the only other way to solve the problem is using AI. So fundamentally, those are the set of problems... Focus on things which clearly the only other solution is a human being or AI.
Portfolio Proof Points
SenseAI's investments demonstrate these principles in action. Continuo.ai tackled construction management by placing 360-degree cameras on hard hats, automatically capturing site progress without manual data entry.
When he walks through the site, the data is automatically captured. He doesn't have to do anything specifically to capture data, not even walk a specific path. Data gets captured, gets translated into a 3D model by AI and progress judged, defects detected, workflows done.
IRAME discovered unexpected applications in employee moonlighting detection, while PipeShift enabled 4x efficiency improvements in AI model deployment. Multiple portfolio companies scaled from zero to nearly $1 million revenue within 6-9 months, representing 3-4x acceleration compared to traditional software.
The Wrapper Trap
Agarwalla developed a crucial distinction between sustainable AI businesses and "wrappers" destined to fail. Wrappers build attractive interfaces around existing AI models without adding meaningful intelligence.
The day OpenAI decides to build a wrapper, you're done. Your business is gone. You're basically, what you do is you go prove a market for OpenAI.
True AI companies embed domain knowledge, develop proprietary data sources, and create defensible value between foundational models and users.
The Pricing Revolution
AI enables a fundamental shift from input-based pricing (seats, storage) to outcome-based pricing. Agarwalla's portfolio company Flowworks exemplifies this, pricing meetings generated rather than emails sent.
AI applications are delivering outcomes almost from day one. Definitely within two weeks of you buying the AI product, you have outcomes, measurable outcomes in your business.
This creates unprecedented alignment between vendor success and customer value.
The 100 Unicorn Prediction
Agarwalla's forecast of 100 AI unicorns from India by 2033 rests on observed growth patterns. Portfolio companies scaling 3-4x faster than traditional software, combined with AI companies trading at 25-40x revenue multiples, makes $30-40 million revenue achievable within 7-8 years.
Today, given the kind of multiples we are seeing in the AI era, we are talking about 25x revenue at least. This means that to hit a billion dollars in valuation, you need to hit 40 million dollars in revenue.
Global context supports this timeline. The 32 new AI unicorns in 2024 represented nearly half of all new unicorn formations. India's 118 total unicorns and strong AI talent concentration position the country well for continued expansion.
How many companies can we build in the next 10 years or nine years which will do at least 25-30 million in revenue starting today... my bet will be at least 100. If not, you'll be saying, Rahul, you're underestimated.
The Contrarian Edge
What separates Agarwalla isn't just early AI timing, but contrarian bets others consider counterintuitive. He prioritizes technical founders over sales leaders, focuses on global markets rather than India-specific solutions, and emphasizes open-source AI adoption over proprietary models.
Without the tech genius, I'm not going to invest. But without some native ability to sell, of course, we won't touch a team which is only pure coders.
The Discovery Engine
SenseAI has evolved into the "discovery layer" of India's startup ecosystem, evaluating 100 AI startups monthly through their VDAT framework (Variety, Data, Architecture, Team).
We are the discovery layer in the ecosystem. There is no list to work off. Sequoia is looking at our list of people we funded.
The Prophet Vindicated
As 2025 unfolds, Agarwalla's predictions gain validation. AI investment shows no slowing, India's position has strengthened, and the transition from experimentation to deployment accelerates across enterprises.
Revenue momentum is never driven by hype. Revenue momentum is driven by products that deliver value. If AI keeps on delivering value at the pace it is delivering value, building the 100 AI unicorns, I'm telling you, is going to be an underestimate.
The data prophet's vision no longer seems audacious, it seems inevitable. The only question is whether his prediction proves conservative or ambitious. Given his track record, smart money suggests conservative.
Listen now!
Other ways to listen:
Your Feedback matters
As always, I’d love to hear your thoughts! Whether it's about this episode or ideas you’ve been playing around with, shoot me an email at ad@thepodium.in. Your feedback keeps these conversations going, and I’m always up for chatting about your startup ideas too.
Until next time,
Your Host,
Akshay Datt