Pod Insights Finshots Daily The Fractal Analytics IPO About this episode In today’s episode on 9th Feb 2026, we break down the Fractal Analytics IPO, which opens for subscription today and closes on February 11th.
Book your FREE consultation with Ditto https://ditto.sh/vx51ji
https://ditto.sh/vx51ji
Transcript 0:00 Speaker 00 Hello folks, you're tuned into Finshots Daily. In today's episode, we break down the Fractal Analytics IPO, which opens for subscription today and closes on February 11th, 2026.
0:14 Speaker 00 Before we begin, here's a quick word from Team Ditto. Life has a way of surprising us and not always in a good way. Sometimes it's a sudden illness or an unexpected hospital visit that can shake up everything. In India, families still pay about 39% of medical expenses directly from their own pockets and just one hospital's take can wipe out years of savings. The easiest way to protect yourself is by getting a good health insurance plan. It's way cheaper than footing one huge bill. And if you're unaware where to start, book a free call with Ditto. No spam, just honest, jargon-free guidance. Trusted by over 8 lakh people for their health and term insurance needs. The link is in the description. Now, back to the story.
0:57 Speaker 00 For years, India's IT boom story followed a similar script. Write lines of code, build and manage systems for customers sitting abroad, all quietly in the shadows. The key decisions behind running the businesses are left somewhere else. Now, that separation looks like it's beginning to blur. This week, Fractal Analytics, a company behind the systems that make those decisions, is headed to the market with a 2,833 crore rupees IPO. When ChatGPT launched, everyone was astonished and equally worried enough to start their own AI models as soon as possible. That didn't just mean your new age startups, but legacy tech firms as well. That's why we saw chatbots, image generation and viral demos that kept getting better.
1:40 Speaker 00 And also why the construct of AI in our minds looks a lot like an everyday use LLM. That's large language model. But here's the catch. The AI model today that makes money rarely looks like that. Instead, it sits inside enterprises and big businesses. And unlike most AI companies that took off in late 2022, Fractal Analytics started way back in the year 2000 in Mumbai. But it didn't start right off the bat as an AI firm. It was helping enterprises use analytics to guide choices across marketing, supply chains and finance. You see, Fractal was offering services in the data analytics and statistical modeling industry, still very nascent and rare at the time. Not many firms had the bandwidth to offer it to consistent, high-paying customers like they do today.
2:25 Speaker 00 Their analytics experience then led to machine learning in the 2010s and eventually into the kind of enterprise AI systems it now sells. Today, Fractal Analytics calls itself an enterprise AI company. If that sounds like tech jargon, we got you. Let's look at it from the perspective of what problem they're trying to solve. Large organizations that operate in many businesses and countries don't struggle without having less data. If anything, they struggle with what to do with it and how to manage it properly. Which customers are important, which product needs attention, what risks to flag. Now that's Fractal's job, build the systems that answer those questions at scale. Most of this data-heavy work sits inside fractal.ai, its primary product that accounts for nearly all of its total revenue.
3:09 Speaker 00 Through this segment, Fractal sells a mix of AI-driven services and proprietary systems to a bunch of large enterprises. Now, this is the product side of what it offers. If you look at it as a service, Fractal helps clients create gen AI systems that fit their specific business needs. These aren't short projects, but long-term agreements with customers in FMCG, that's fast-moving consumer goods, healthcare, banking and more. The custom-built AI systems are deeply rooted into their customers' internal operations. With these services that Fractal provides to its customers, it's also working on its own AI products through something they call Congentech. Instead of starting from scratch, clients can use pre-built agents, workflows and connectors
3:52 Speaker 00 to integrate AI into existing systems. In effect, Fractal is trying to move from selling intelligence project by project to packaging parts of it as a platform. One way to think about Congentech is an enterprise AI suit. Just like Adobe packages different creative tools into a single platform, Congentech brings together multiple AI agents' workflows and decision applications under one umbrella, designed to be deployed together rather than as standalone tools. But unlike consumer software suits, Congentech is built for large enterprises and is rolled out alongside fractal services. The companies that buy fractal services aren't experimenting with AI. They're businesses with billions of dollars in revenues, millions of customers. They don't buy or use.
4:34 Speaker 00 Use AI systems the same way that you and I would by downloading an application. Once fractal systems are in place, they tend to spread. A project that begins in one function often expands to others across geographies and teams. That expansion is visible in the company's numbers. A majority of its revenue comes from existing clients and those clients typically spend more overtime. In practice, this creates long relationships and high switching costs. As of September twenty twenty five, Fractal works with over one twenty of what it calls must-win clients, which bring in recurring revenue and multi-year agreements. These big-ticket clients include Citibank, Nestle, Costco, Marsh, Mandela's and Philips.
5:18 Speaker 00 More than half of its revenues come from its top ten clients, most of whom have been around with them for over eight years. If you were to lay out their customers on a world map, most of them would be sitting in the US. Around sixty four point nine percent, twenty one percent of them from Europe and seven point six percent from India. Since this is a B two B enterprise AI business, client churn tends to be lower. That's why Fractal's existing customers spend more each year, a sign that once AI systems are embedded, they become part of an organization's operating fabric. And the proof is in the pudding. Take a look at how Fractal's revenues have grown. They went from two thousand two hundred and forty one crore rupees to two thousand eight hundred and sixteen crore rupees from FY twenty four to FY twenty five,
6:03 Speaker 00 even turning profitable from a net loss of minus fifty four crore rupees to two twenty crore rupees over the same period. One number quietly captures this dynamic. Fractal's existing clients spend more with the company each year, pushing its net revenue retention above a hundred and ten percent. In simple terms, even without signing new customers, Fractal's revenue base tends to grow. There is one more number that clearly explains this behavior and that's NPS or the Net Promoter Score. It's the measure of how satisfied a customer is with the service that they receive. Fractal's NPS consistently sits in the high seventies year after year. That matters because in enterprise AI, satisfaction isn't about a better interface.
6:47 Speaker 00 It determines whether a project expands to another team, another geography or another year. In today's AI boom, profitability is the exception, not the rule. For Fractal, that profitability still comes with a caveat. Like most enterprise AI firms, its biggest cost is people. Highly skilled data scientists, engineers and domain experts don't come cheap and employee expenses make up a large share of operating costs. Of the 1594 crore earned until September 2025, 1125 crore went into employee benefits expenses. Fractal's strengths are also the source of its key risks.
7:27 Speaker 00 More than half of its revenues come from its top 10 clients, making the company sensitive to spending decisions by relatively small group of enterprises. Geographic exposure adds another layer of risk. With nearly two thirds of revenues originating from the United States, Fractal is closely tied to North American enterprise tech budgets. Any slowdown in IT and AI spending, changes in regulations or currency volatility could affect growth and margins sharply. Set at a price band of 857 rupees to 900 rupees, Fractal's 2833 crore IPO is a mix of a fresh issue and an OFS, that's offer for sale.
8:05 Speaker 00 1024 crore raised will go into company itself, giving it capital to invest in its AI platforms, expand into new markets and deepen relationships with large enterprise clients. Over 25% of the fresh issue will go into debt repayment of Fractal USA, its American subsidiary. This move is significant because of all Fractal subsidiaries, the USA office alone brings in 77% of revenues. But debt repayment isn't the biggest use of funds. 355 crore is reserved for R and D into both fractal.ai and fractal alpha, its two growth engines. Recognizing the importance of expansion, they've also earmarked 121 crore to set up new offices in India and 57 crore just for purchase of laptops for their employees.
8:52 Speaker 00 The rest will be used for general corporate purposes and inorganic growth. The remaining 1810 crore is an offer for sale, letting existing shareholders bear down their stakes. If you look at the valuation, the IPO prices the company at 15,400 crore, which is a whopping 79 times its FY25 earnings. And since this is India's first ever AI company to go public, it's kind of hard to judge whether that price is fair. But that also makes it hard to compare or replace. Whether the market embraces it or not is something that we'll have to wait and see.
9:30 Speaker 00 Thank you for listening to today's episode. And if you want to share your feedback or suggestions, do drop us an email to [email protected] . Until next time.
Previous A vegan cricket ball that lasts 50 overs? Next The AI crisis no one’s talking about Page updated November 10, 2026 · transcript and summary are machine-generated