Startup Deep Dive : Fractal Analytics — profitable, listed, and still worth less than private investors paid in 2025

Fractal Analytics turned a net profit of Rs 221 crore in the year to March 2025, priced India’s first pure-play AI initial public offering in February 2026, and is still worth close to Rs 13,225 crore (about $1.38 billion, at $1 ≈ Rs 96.0 as of 18 September 2026) on the stock market today. Seven months before that listing, private investors had paid a price that valued the same company at $2.4 billion.

That gap between what Fractal was worth in a private secondary sale in July 2025 and what the public market is willing to pay for it now is the real story here, not the “first AI company to list in India” headline. For a quarter of a century, Fractal has made its living doing the least glamorous part of artificial intelligence: sitting inside the finance, marketing and risk teams of companies like the ones that make your breakfast cereal and process your credit card, building the statistical models that quietly decide who gets a discount, whose insurance claim looks fraudulent, and which factory line is about to fail. What happens when a genuinely profitable, three-decade-old AI-services business meets an Indian stock market that has grown wary of AI valuations is the plot of this piece.

Quick facts

What they do

Fractal sells applied artificial intelligence and decision science as a service, mostly to large multinational enterprises in consumer packaged goods and retail, financial services, technology and media, and healthcare and life sciences. In practice, that means Fractal’s data scientists and engineers work embedded inside a client’s own marketing, risk or supply-chain teams, building and running the statistical models that make thousands of small, repeated business decisions: which shopper should see which price, which insurance claim needs a human to look twice, which piece of factory equipment is about to fail. Alongside this consulting-heavy core, Fractal has built a small portfolio of its own branded AI products, sold on subscription, including Crux Intelligence for business intelligence, Cuddle.ai as a “decision co-pilot,” Eugenie.ai for anomaly detection and Senseforth.ai for conversational AI. It also incubated Qure.ai, the healthcare-imaging AI company that reads X-rays and CT scans for tuberculosis and lung cancer — but Qure.ai was spun out as an independent, separately funded company and is not part of Fractal’s own revenue or the entity that listed in February 2026.

The origin

Fractal was incorporated in Mumbai in 2000 by five classmates from IIM Ahmedabad — Srikanth Velamakanni, Pranay Agrawal, Nirmal Palaparthi, Pradeep Suryanarayan and Ramakrishna Reddy — who walked away from investment-banking and consulting careers on a bet that mathematical modelling would eventually decide corporate strategy rather than instinct or seniority. The original plan, as later accounts of the company’s founding describe it, was closer to a consumer dot-com idea. Within months, the dot-com bubble had burst, capital for that kind of business dried up overnight, and the founders had to find a paying use for the one thing they actually had: the ability to model data. The insight that survived the crash was narrower and more durable than the original idea — that large enterprises were making expensive decisions on gut feel because nobody had built them the statistical tools to do otherwise, and that this gap could be sold, one modelling engagement at a time.

The struggle years

The dot-com crash did not just kill the original business plan; it nearly killed the company itself. With enterprise technology budgets frozen through 2001, Fractal survived its first years on a handful of contingent, results-only assignments rather than retainer contracts, and one early account of the company’s history puts its founding capital at a modest pool put in personally by the five founders, with a further round of seed money — reported at around $600,000 — coming from Onida’s Gulu Mirchandani to keep the lights on. The company’s first real validation came in 2001, when ICICI Bank hired Fractal to build a statistical credit-risk scorecard, effectively India’s earliest attempt at a data-driven consumer credit model, giving the young firm proof that banks would pay for this kind of work.

The next crisis was internal. By 2007, the five co-founders had fallen out over direction and control; Ramakrishna Reddy and Pradeep Suryanarayan exited the business, leaving Srikanth Velamakanni and Pranay Agrawal as the two founders who would run the company for the next two decades. Then came a client shock: General Motors’ 2009 bankruptcy filing, in the middle of the global financial crisis, wiped out a client relationship that had been worth a meaningful share of Fractal’s revenue at the time, forcing the firm to rebuild its automotive and credit-risk practice around recovery models for other lenders instead of relying on any single large account — a lesson in client concentration the company would still be grappling with, in a much bigger way, when it filed to go public more than fifteen years later.

The turning point

The moment that turned Fractal from a well-regarded but sub-scale analytics shop into an institutionally credible, IPO-track company was a pair of private equity investments seven years apart. In January 2019, UK-based Apax Partners bought a significant minority stake for $200 million, Fractal’s first large, marquee financial-sponsor cheque. Three years later, in January 2022, TPG Capital Asia invested $360 million — a mix of new capital and secondary shares bought from Apax and other holders — in a deal that pushed Fractal’s valuation above $1 billion for the first time and made it one of India’s AI “unicorns.” That single fundraise mattered less for the cash than for what it signalled: two global private equity firms were now willing to underwrite Fractal’s numbers, its governance, and its growth story at a billion-dollar-plus price, which is precisely the credibility a company needs before investment bankers will agree to take it to a public listing. Everything that followed — the further growth to Rs 2,765 crore of FY25 revenue, the August 2025 draft prospectus, and the February 2026 listing — flows in a direct line from that 2022 valuation crossing.

The money behind it

Fractal has raised roughly $685 million in disclosed private funding over its life, a relatively lean amount for a company that reached a multi-billion-dollar valuation, reflecting the fact that it built up a real, profitable-in-parts services business rather than burning capital on customer acquisition. The two backers that mattered most were Apax Partners, which put in $200 million in January 2019 and became the company’s first anchor institutional shareholder, and TPG Capital Asia, whose $360 million cheque in January 2022 both crossed Fractal into unicorn territory and bought out part of Apax’s stake, giving both firms aligned exits to plan around. A further private secondary sale in July 2025, ahead of the IPO, valued the company at $2.4 billion — a figure confirmed independently by TechCrunch’s IPO coverage and by IndexBox’s market analysis of the listing. That $2.4 billion mark turned out to be the high-water point: bankers advised a conservative pricing given a broader sell-off in Indian software and AI-adjacent stocks, and Fractal ultimately listed at a market capitalisation of about Rs 14,810 crore (roughly $1.6 billion) on 16 February 2026 — already a third below the last private price — before drifting further to around Rs 13,225 crore ($1.38 billion) by 18 September 2026. As of June 2026, foreign institutional investors held about 40.9% of the company, promoters about 17.0%, domestic institutions 12.3% and the remaining roughly 29.9% sat with the public that bought in through and after the IPO.

How it makes money

Strip away the AI branding and Fractal’s income statement reads like a staffing-heavy consulting firm, not a software company. In FY25, “analytical services” — teams of data scientists and engineers billed to clients on a time-and-scope basis for multi-year decision-science engagements — brought in Rs 2,701 crore, or roughly 97.7% of total revenue. The much smaller “subscription income” line, from its own AI products sold as software, was just Rs 64 crore, or about 2.3% of revenue, even though that line grew a striking 167% year on year. This is the part outsiders most often get wrong about Fractal: because it talks about AI products, incubated Qure.ai, and now trades as India’s first listed “AI company,” it gets mentally filed next to pure-play AI software firms. The numbers say otherwise — it earns its keep the way a large services firm does, by billing people’s time, and the product layer is a fast-growing but still marginal add-on rather than the main engine.

That services-first model shows up starkly in the cost structure. Employee benefit costs alone were Rs 2,005 crore in FY25, about 72.5% of revenue — well above the 50-60% range typical of large IT-services firms — leaving an EBITDA margin of just 14.13% and a cost of roughly 93 paise for every rupee of revenue earned. Fractal also reinvests a further single-digit percentage of revenue into research and development for its product bets. The upshot is a business that can swing from profit to loss quickly if revenue growth slows even a little while headcount and salary costs keep rising, which is close to exactly what happened in FY24, discussed next.

The numbers

The FY23 profit needs a footnote the headline number does not carry on its own: of the Rs 194 crore reported, Rs 541 crore came from a one-time exceptional gain and a deferred-tax adjustment rather than the underlying business; strip that out and Fractal actually ran an operating loss of about Rs 181.5 crore that year. FY24 then delivered a real, un-adjusted net loss of Rs 54.7 crore even as revenue kept growing, before FY25 brought the company back to a genuine operating profit of Rs 221 crore on 26% revenue growth. Read across the four years, growth has been consistent — revenue has more than doubled since FY22 — but profitability has been anything but smooth, swinging on accounting one-offs and cost discipline rather than tracking revenue in a straight line.

Where the money comes from

Fractal’s FY25 revenue of Rs 2,765 crore broke down geographically as roughly Rs 1,802 crore (65.2%) from the United States, Rs 484 crore (17.5%) from Europe, Rs 247 crore (8.9%) from the rest of the world, and just Rs 232 crore (8.4%) from India. The surprise, for a company that is now listed on Indian exchanges and headquartered partly in Mumbai, is how small its home market actually is to its business: Fractal is, in substance, a US-facing analytics exporter with an Indian delivery base and now an Indian stock-market listing, which is why its fortunes track the health of American corporate technology budgets far more closely than they track the Indian economy. On top of the geographic concentration, Fractal is also client-concentrated: it describes around 122 large global accounts internally as “Must Win Clients,” and as of the six months to September 2025, its top 10 clients alone accounted for 54.2% of revenue, with repeat business running above 90% of revenue and net revenue retention of 121.3% in FY25 — a sign that existing large clients are spending more each year, even as the client base stays narrow.

The risks

The first and most concrete risk is client concentration: with roughly half of revenue sitting inside ten accounts, a single large client pausing a renewal, insourcing a function, or cutting its analytics budget in a downturn would show up directly in Fractal’s results, not diluted across a broad customer base the way it would at a larger IT-services peer. The second is geographic concentration in the United States, which supplied about 65% of FY25 revenue; because enterprise analytics spend is often one of the more discretionary lines in a technology budget, it tends to be an early target for cuts when American corporates turn cautious, exposing Fractal to US business cycles more than to conditions in India. The third is the structurally thin, staffing-heavy margin itself: with employee costs already at about 72.5% of revenue and EBITDA margin at 14.13% in a good year, there is little room to absorb wage inflation, attrition, or a client price renegotiation without profit swinging into a loss again, which is essentially what an unadjusted look at FY24 already showed happen once.

The takeaway

Fractal’s twenty-five year run says something less obvious than “AI companies are hot”: being useful, for a long time, in an unglamorous corner of a client’s business is a survivable strategy even when the market around you is not paying attention, but it does not automatically convert into the valuation multiple a hot label might imply once you finally step into public markets. Fractal built two decades of trust inside Fortune 500 finance and marketing departments long before “AI” was a word that moved stock prices, survived a dot-com bust, a founder split and the loss of a marquee client along the way, and reached genuine profitability on its own terms. None of that stopped its public listing from being priced down and then trading down further, because public investors were pricing this quarter’s sentiment about AI-services valuations, not two decades of quiet delivery. The lesson for any founder eyeing a similar path is that durability and a premium price are two different things investors can reward at very different times, and building the first does not guarantee the second arrives on schedule.

Frequently asked questions

Is Fractal Analytics the same company as Qure.ai?

No. Qure.ai, the healthcare-imaging AI company known for reading X-rays and CT scans, was incubated inside Fractal and later spun out as an independent, separately funded company. It is not part of Fractal Analytics Limited, the entity that reported FY25 revenue of Rs 2,765 crore and listed on Indian exchanges in February 2026.

Is Fractal Analytics profitable?

It reported a net profit of Rs 221 crore in FY25 (year ended March 2025), reversing a net loss of Rs 54.7 crore in FY24. Profitability has been uneven across recent years: FY23’s headline profit of Rs 194 crore relied heavily on a one-time exceptional gain, without which the company would have posted an operating loss that year.

When did Fractal Analytics list, and how has the stock performed?

Fractal listed on the NSE and BSE on 16 February 2026 at an issue price of Rs 900 per share, but debuted below that price, closing its first day around Rs 873.70, roughly 7% lower. By 18 September 2026 the stock traded near Rs 764, giving the company a market capitalisation of about Rs 13,225 crore ($1.38 billion), well below the $2.4 billion valuation a private secondary sale had put on it in July 2025.

Who owns Fractal Analytics?

Co-founders Srikanth Velamakanni and Pranay Agrawal continue to lead the company, with Agrawal as CEO since 2016 and Velamakanni as Group Chief Executive and Vice-Chairman. Private equity firms Apax Partners and TPG Capital Asia are major institutional shareholders, having invested $200 million in January 2019 and $360 million in January 2022 respectively; as of June 2026, foreign institutional investors held about 40.9% of the company and promoters about 17.0%.

What does Fractal Analytics actually sell?

Overwhelmingly, it sells analytics and AI consulting services — teams of data scientists embedded with large enterprise clients — which made up about 97.7% of FY25 revenue. Its own AI products, such as Crux Intelligence and Cuddle.ai, are sold on subscription and are growing fast but still contributed only around 2.3% of FY25 revenue.

Sources

Figures are as of September 2026. Currency converted at $1 ≈ ₹96.0 as of 18 September 2026 (Trading Economics).

• Wikipedia, “Fractal Analytics” — accessed September 2026
• TechCrunch, “Fractal Analytics’ muted IPO debut signals persistent AI fears in India” — 16 February 2026
• IndexBox, “Fractal Analytics IPO: Stock Declines in Debut After Valuation Cut” — February 2026
• Entrackr, “Fractal crosses Rs 2,700 Cr revenue in FY25, regains profitability” — 2025
• Entrackr, “AI unicorn Fractal posts Rs 1,985 Cr revenue and Rs 194 Cr profit in FY23” — October 2023
• Inc42, “Exceptional Gain Helps SaaS Unicorn Fractal Post INR 194 Cr Profit In FY23” — 2023
• Inc42, “IPO-Bound Fractal Slips Into The Red In FY24, Reports INR 54.7 Cr Loss” — 2024
• Business Standard, “PE firm TPG invests $360 mn in AI and analytics services firm Fractal” — 5 January 2022
• TPG, “Fractal announces US$360 million investment from TPG” — January 2022
• Apax Partners, “Fractal announces US$360 million investment from TPG” — January 2022
• Samco, “Fractal Analytics IPO Listing At 3% Discount” — February 2026
• Entrepreneur India, “Fractal Analytics Scales Down IPO; Sets Price Band at INR 857–900” — 2026
• Groww, “Fractal Analytics Files DRHP with SEBI, Plans for Rs 4,900 crore IPO” — August 2025
• Screener.in, “Fractal Analytics Ltd” financial data — accessed 18 September 2026
• Analytics India Magazine, “Fractal Analytics & Its Fixation With Spinning-Off Product Companies” — background on Qure.ai spin-out
• GlobeNewswire, “Fractal Announces Launch of Crux Intelligence” — 14 July 2021
• PR Newswire, “Fractal announces fundraise and spin-out of Theremin.ai” — background on product spin-outs and later divestment, June 2024
• Bastion Research (Substack), “Fractal Analytics IPO: A Consulting Business with a Product Ambition” — 2025/2026
• ajuniorvc, “Can 25,000 Cr Fractal Take Enterprise AI from India Globally?” — 2025/2026, for early company history

Found an error? Write to us and we’ll correct it in the open, dated, on the piece.