The 25-year-old Mumbai firm sells data and AI services to Fortune 500 companies. Now it wants Rs 2,834 crore from retail and institutional investors to fund its next chapter.
Fractal’s business model is different. Fractal builds an AI model that monitors inventory across the entire network, classifies slow-moving items, identifies the root cause, and sends early warnings before stock piles up. The client does not need to hire a data science team. It does not need to build infrastructure. Fractal does both.
Srikanth Velamakanni and Pranay Agrawal started the company in Mumbai in March 2000, originally under the name Fractal Communications Limited. India’s technology industry at the time was writing code for American banks and running call centres for British telecom firms. Data analytics, as a standalone corporate function, did not exist.
The two IIM Ahmedabad graduates spent the next two decades building the business one client at a time — adding consumer goods companies first, then financial services, then healthcare and technology firms. They expanded into the US and Europe. They acquired smaller companies to fill capability gaps. Final Mile brought behavioural science. Senseforth added conversational AI. Neal Analytics contributed cloud data engineering. Samya.ai, now called Asper, provided supply chain intelligence.
Fractal has also been building its own AI foundation models — Kalaido.ai for image generation, Vaidya.ai for medical diagnostics, Fathom-R1-14B as an open-source reasoning model. By September 2025, the company had 5,722 employees, 122 large enterprise clients, offices across seven countries, and revenue run-rating above Rs 3,100 crore.
The company now operates two segments. Fractal.ai is the core — AI services and products, anchored by Cogentiq, its agentic AI platform. Fractal Alpha houses independent AI ventures including Analytics Vidhya, a data science community with nearly 5 million users, and Asper, a supply chain platform. Fractal also holds a minority stake in Qure.ai, a radiology AI startup with US FDA clearances.
The Client Roster
The list of companies Fractal works with is the strongest line in its prospectus. Citibank, Costco, Franklin Templeton, Mars, Mondelez, Nationwide, Nestle, Philips. A majority of the Magnificent Seven. Ten of the world’s 20 largest CPG companies. Eight of the top 20 in technology, media, and telecom. Ten of the top 20 in healthcare and life sciences.
Fractal defines its target customers as “Must Win Clients” — enterprises with either $10 billion in revenue, $20 billion in market cap, or 30 million end-customers. It served 122 such clients as of September 2025. The top 10 have been with the company for an average of over eight years. The Net Promoter Score for the Fractal.ai segment has stayed above 73 since FY23.
The company has also developed its own IP. Twenty-eight registered patents, 38 pending. Foundation models in image generation, medical imaging, and mathematical reasoning. Cogentiq as an agentic AI platform. None of these are meaningful revenue contributors today. Subscription income — the product-linked revenue line — was Rs 47 crore in H1 FY26, up from Rs 16 crore in FY23.
Reading the Numbers
Revenue compounded at 18% from FY23 to FY25, reaching Rs 2,765 crore. H1 FY26 produced Rs 1,559 crore, up 19.9% over the year-ago period. Over 90% of revenue comes from outside India. The United States alone accounts for roughly 65%.
Profitability has been uneven. Net profit was Rs 194 crore in FY23, inflated by a Rs 524 crore exceptional gain. FY24 produced a net loss of Rs 55 crore. FY25 recovered to Rs 221 crore. H1 FY26 stands at Rs 71 crore. The adjusted numbers tell a different story — adjusted EBITDA margins went from 6.8% in FY23 to 10.6% in FY24 to 17.4% in FY25, settling at 15% in H1 FY26. Stock option costs (Rs 27 crore in H1 FY26), share of losses from associate Qure.ai (Rs 45 crore in H1 FY26), and exceptional items create persistent distance between reported and adjusted profitability.
Cash generation has been lumpy. Operating cash flow went from negative Rs 31 crore in FY23 to Rs 397 crore in FY25, then back to negative Rs 21 crore in H1 FY26. Trade receivables stood at Rs 620 crore as of September 2025 — roughly 40% of half-yearly revenue. Working capital is the swing factor, and for a company seeking a Rs 15,480 crore valuation with Rs 275 crore in borrowings, cash conversion will matter to long-term holders.
Where the Risks Sit
Client concentration is the most measurable risk. Top 10 clients generate 54% of Fractal.ai revenue. One client alone is 8.2%. Top 20 clients account for 72%. CPGR and TMT together make up nearly two-thirds of industry mix. The US is 65% of geography. Long tenures provide stability. But the arithmetic does not forgive — losing one or two major accounts would register immediately.
There is also a structural tension at the heart of the business. Fractal builds AI tools that automate human decision-making. Generative AI and low-code platforms are now automating parts of the analytics services market — the same market Fractal bills into. The company names this risk explicitly: service revenue erosion due to productisation. It is simultaneously building the products — Cogentiq, Asper, the foundation models — that could eventually eat into its own services revenue. Managing that transition is one of the hardest problems in enterprise technology.
Talent competition is constant. AI engineers are in demand at hyperscalers, startups, and captive centres. Fractal has earned Great Place to Work recognition for eight consecutive years across seven countries. Awards do not set compensation benchmarks.
Regulatory risk is building. The EU AI Act, US executive orders on AI governance, and India’s data protection framework all add compliance costs and could slow deployment in regulated industries like healthcare and financial services.
The IPO Structure
The offer opens February 9 at Rs 857 to Rs 900 per share and closes February 11. Total size is Rs 2,834 crore — Rs 1,024 crore as a fresh issue and Rs 1,810 crore as an offer for sale. At the upper band, implied market capitalisation is Rs 15,480 crore. Listing is expected February 16.
The OFS is 64% of the total issue. Quinag Bidco, the Apax Partners vehicle, is selling Rs 881 crore worth of shares it bought at Rs 173 apiece — roughly a 5x return at the upper band. TPG is offloading Rs 450 crore at a cost of Rs 642 per share. The GLM Family Trust is selling Rs 450 crore at a cost basis of nil. The Remala family is exiting Rs 30 crore at a cost of Rs 2 per share. More money is leaving the cap table than entering the company’s bank account.
Post-IPO, investor selling shareholders go from 63.5% to 47.6%. Promoters retain 16.7%. Fresh issue proceeds are being allocated to debt repayment at the US subsidiary (Rs 265 crore), R&D and sales for Fractal Alpha (Rs 355 crore), new Indian offices (Rs 121 crore), laptops (Rs 57 crore), and a combination of acquisitions and general corporate purposes.
At Rs 900 per share, the company is valued at roughly 56 times FY25 net profit, or about 32 times adjusted EBITDA. There is no listed comparable. Investors will construct their own frameworks — part Indian IT services, part global AI companies, part SaaS multiples. None will fit precisely.
| Metric | FY23 | FY24 | FY25 | H1 FY26 |
|---|---|---|---|---|
| Revenue | 1,985 | 2,196 | 2,765 | 1,559 |
| Net Profit/(Loss) | 194 | (55) | 221 | 71 |
| Adj. EBITDA | 134 | 232 | 482 | 233 |
| Adj. EBITDA Margin | 6.8% | 10.6% | 17.4% | 15.0% |
| Operating Cash Flow | (31) | 160 | 397 | (21) |
| Employees | 4,221 | 4,639 | 5,254 | 5,722 |
Issue Snapshot
| Detail | |
|---|---|
| Price Band | Rs 857–900 |
| Issue Size | Rs 2,834 Cr |
| Fresh Issue | Rs 1,024 Cr |
| OFS | Rs 1,810 Cr |
| Market Cap (upper band) | Rs 15,480 Cr |
| P/E (FY25 reported) | ~56x |
| P/E (FY25 adjusted) | ~34x |
| EV/Adj. EBITDA (FY25) | ~32x |
| Open / Close | Feb 9–11, 2026 |
| Listing | Feb 16, 2026 |
| BRLMs | Axis Capital, Kotak, Morgan Stanley, Goldman Sachs |
