Companies
How Fractal, India’s first AI unicorn, is prepping for the next race
Summary
- Fractal Analytics Inc. became India’s first AI unicorn in 2022. It took a long time to get there—over 20 years. But, Srikanth Velamakanni, the company’s co-founder, has an audacious goal. Math, neuroscience and the power of compounding could help him. Keep reading.
Bengaluru: Near-death experiences are not uncommon in startups. For Fractal Analytics Inc., a provider of artificial intelligence (AI) products and solutions, that existential crisis came seven years after it started—in 2007. Two of its five co-founders suddenly left to start another company. And a third of Fractal’s employees quit to join them.
Left with 40 people, the split crippled the analytics company’s ability to execute. From then on, began a slow and gruelling journey to resurrect itself. It took Fractal over 20 years to hit a landmark today’s startups aspire to achieve in about five years. In January 2022, TPG Capital Asia pumped in $350 million into Fractal, making it India’s first AI unicorn, or a company with a valuation of over a billion dollars.
The company today has over $300 million revenue, 4,600 employees and operations spread across 17 global locations. Data intelligence platform Tracxn ranks Fractal eighth among 1,435 competitors in the data analytics space. The long list includes giants such as Accenture and PwC, as well as well-funded startups such as CleverTap, Mu Sigma and data.ai.
When Fractal started, in Mumbai in 2000, very few knew about machine learning and AI. The Tracxn listing tells you how competitive the space has become—hordes of companies are rushing to make AI-based products and deliver analytics services as global corporations struggle to make sense of the massive amount of data they already possess and generate regularly. So, why are investors backing Fractal?
This is the story of Srikanth Velamakanni and Pranay Agrawal, Fractal’s co-founders, who have stuck with the company through thick and thin. They have built the company through 10 acquisitions and is now collaborating with the hyper scalers (Amazon Web Services, Microsoft, Google, Databricks, Nvidia).
“Fractal, along with Zoho and Freshworks, is one of the few Indian technology product and solutions companies that are successful at the global level," said Jayanth Kolla, founder and partner at Convergence Catalyst, a deeptech consultancy firm. “Fractal had the vision of putting analytics at the core of its existence, placing it a generation or two ahead of its time," he added.
Zoho, founded by Sridhar Vembu, and Freshworks, founded by Girish Mathrubootham, provide software as a service (SaaS) solutions for businesses.
Before we tell you how Fractal plans to stay ahead in the AI race, let’s rewind to the beginnings.
Decade of stress
Velamakanni had a trouble-free childhood. Born in a small city in Andhra Pradesh, he grew up in an industrial town in Assam where his father worked as a senior executive for Oil India Ltd. The father, who was a “very good engineer", had only one demand: Be good at math. “Else, I had no major stress as a child," Velamakanni recalled.
After completing his bachelor of engineering from the Indian Institute of Technology (IIT-Delhi), followed by an MBA from the Indian Institute of Management (IIM, Ahmedabad), he took up a job with ANZ Investment Bank in 1998. He followed this with a stint at ICICI Bank. But on 15 August 1999, he told one of his friends, “I’m going to quit by the turn of the millennium, and will start a company."
Velamakanni kept his word. Along with Pranay Agrawal, Nirmal Palaparthi, Pradeep Suryanarayan and Ramkrishna Reddy, he started Fractal. All the co-founders pooled in about ₹2 lakh each.
“I got my ₹2 lakh from my mother. A few other friends sent us $10,000 cheques, and Sasha Mirchandani (managing director and founder of Kae Capital and co-founder of Mumbai Angels) invested about $400,000, which helped us start the engine," he recalled.
The next decade proved to be stressful. He co-founded Fractal in 2000, at the peak of the dotcom bubble—many companies started closing down. Then came the split with the co-founders. The company also worked with very little external capital. After Mirchandani’s investment, Fractal didn’t raise any more money till 2013 when TA Associates invested $45 million.
Meanwhile, Fractal battled to establish its own identity.
Not a BPO
None of the co-founders were thinking of building an analytics company when they started.
“We only wanted to help customers make better decisions when buying devices, appliances, etc.," said Velamakanni.
Soon, the co-founders realized that the company wouldn’t make much money with this approach. They pivoted into an analytics firm to help enterprises make better decisions using data, math, human psychology, and so on. While there were firms like Fair, Isaac and Company that were into building risk models for companies, the concept of analytics didn’t really exist back then, said Velamakanni.
An initial business came from ICICI Bank where Fractal was tasked with building a scorecard to automate lending. Then, the company worked with Hindustan Unilever, helping it “understand consumer buying behaviour through the lens of data".
But Fractal faced a big challenge. It was being branded either as an information technology (IT) services company or a business process outsourcing (BPO) company, which, back then, was mostly a call centre. “To differentiate, we took a big decision—we would never write any software for our clients," recalled Velamakanni. “At that time, it was very important to identify ourselves as an analytics company," he added.
Fractal originally focused on India. Few years later, Velamakanni decided to move to the US while the operations in India were being overseen by Ramkrishna Reddy. But in 2006, investors recommended appointing a chief executive officer (CEO) from among the co-founders, little knowing that this development would also fracture their company.
“Pranay and I were in favour of making R.K. (Reddy) the CEO," Velamakanni said.
However, investors picked Velamakanni to head Fractal. Six months later, Suryanarayan and Reddy, who were unhappy with Velamakanni’s leadership, broke ranks to start their own company.
“They left and took my team. I was very angry and upset," recounted Velmakanni. It took three years for matters to settle down. “I was in my early 30s then. I reflected and also apologized to all of them," he said.
Power of 10
Serendipitously, AI came to Fractal’s rescue since those were the early days when people started to see a future in machine learning techniques—deep learning in speech processing, text processing, natural language processing, and image processing.
These were technologies that Fractal had expertise in. Around 2015, Velamakanni decided to invest 12.5% of its annual revenue on research.
The idea appealed to Khazanah Nasional Bhd, the strategic investment fund of the Government of Malaysia. In May 2016, the sovereign fund invested $100 million in Fractal to enhance its AI- and deep learning-based software stack. The funding also helped the company invest in new businesses and acquire startups.
“Till date, we acquired (and sold some) 10 startups, including Mobius Innovations (which Fractal bought from one of its co-founders). All of them were small. None of them has contributed more than $4-5 million of revenue," said Velamakanni. “But they’ve all added significantly in terms of our capabilities."
Some of the company’s AI-powered platforms today include Asper.ai (for revenue growth), Crux Intelligence (analytics), Eugenie.ai (emissions intelligence platform), Flyfish.ai (generative AI for sales), Qure.ai (algorithms for medical imaging), Senseforth.ai (conversational AI platform), and Final Mile (behavioural sciences firm).
Senseforth.ai had built Eva, HDFC Bank’s AI-based chatbot. For Coca-Cola Company, Fractal helped unify disparate data spread across many data sources such as with the bottlers. A financial institution, meanwhile, leveraged Fractal’s machine-learning algorithms to estimate customer and merchant locations using transaction data.
GenAI playbook
On Holi, 25 March, Fractal wished users on X, formerly Twitter, in a rather unusual way: “Wishing you all a kalAIdoscopic Holi filled with boundless happiness and unforgettable memories," the post read.
The post was accompanied with a video that showed a series of pictures—Einstein and robots playing Holi; astronauts in space with a burst of colours in the background; a tiger and a bird in Holi colours. The video added another note: “All images created by Fractalites and their family & friends using kalaido.ai".
A month earlier, on 29 February, Fractal had introduced kalaido.ai, a text-to-image platform—it can generate high quality images from text prompts in English and 17 Indian languages including Hindi, Sanskrit and Tamil. It is trained on a public dataset of 70 million images.
In some ways, kalaido.ai is a showcase of the power of generative AI (AI technology that can create new types of content including text, images, audio, video, and synthetic data) and its possibilities. Fractal is pumping research dollars into many such models and use cases to future proof its business.
For instance, FractalGPT, an off-the-shelf product of enterprises, is designed to boost productivity, save time, and enrich call centre operations. Meanwhile, Fractal is also focusing on developing ‘AI agents’. Unlike chatbots, AI agents can replicate personalities and emotional aspects of human interactions by responding to queries with personalized responses. One such AI agent is called Dexter. It can assist employees in their interactions with their company.
“We’re also working on quantum computing. There is an incredible amount of research going on at Fractal, which helps us stay ahead," the co-founder said. “If GPT-5 comes in and it’s better, we’ll build on top of that. We are also building our own foundation models like the Kaleido text-image model," he added.
GPT is short for Generative Pre-trained Transformer, a large language model created by OpenAI.
Forrester Research, a market research company, in a June 2023 report, named Fractal a ‘leader’ among customer analytics service providers, also stressing on the edge the company has because of its investment in research.
“Fractal correctly differentiates with its emphasis on design and engineering, which help clients bridge the insights-to-action gap," Brandon Purcell, vice president and principal analyst with Forrester, wrote in the report. “Fractal’s innovation approach reflects this ethos…dedicated funding for research and development in generative AI and neuroscience should strengthen its leadership position," he added.
Staying cute
Nonetheless, leaders can get disrupted. Fractal has done well but it still has to deal with stiff competition—from every big and small tech company that offers AI and GenAI solutions. Like we mentioned earlier, the analytics market is getting crowded.
The company will need to continuously differentiate its offerings and demonstrate unique value propositions, analysts said. For example, it will have to showcase proprietary algorithms or industry-specific expertise to stand out in competitive bids. “Convincing large and first time clients that it can provide significant value could be a challenge," Kolla of Convergence Catalyst said. “Because in a lot of cases, the value offered by analytics could be incremental… which might not look impressive from the recipient’s point of view," he added.
Second, the demand for skilled data scientists, analysts, and AI specialists is high. There’s a shortage of talent in the market, which makes recruitment and retention challenging for any firm. Fractal won’t be immune to this. The company, meanwhile, has launched AI courses for free on Coursera, an e-learning platform, in a bid to attract the best talent.
Last, but not the least, since Fractal generates a majority of its revenue from the US and EU, it will have to ensure responsible use of AI algorithms, avoid biases in decision-making, and comply with the ever evolving data protection regulations, analysts pointed out.
Compounding Magic
In 2022-23, Fractal posted a 55.5% jump in consolidated revenue to ₹2,043 crore from a year earlier, as per its latest filings sourced from Tofler. The company reported a profit of ₹194 crore during the year from a loss of ₹148 crore a year earlier.
“We are currently at a $300 million annual run rate," said Velamakanni. About 70% of Fractal’s current revenue is generated in the US; 20% from Europe; 10% from Australia.
The company is clearly piggybacking on the mushrooming of AI and generative AI-based solutions market. The global generative AI market size is projected to grow from $67 billion in 2024 to $968 billion by 2032, a CAGR of 39.6%, according to Fortune Business Insights. That’s a super-sized playground for analytics companies.
Fractal, therefore, believes it is time to exponentially expand. Velamakanni wants to eventually grow Fractal into a $15 billion revenue company but has not set a timeline yet.
How does he plan to achieve this?
Target 500 companies doing an annual business of $30 million each. “Of these, some will be making more than $100 million revenue a year, but the average has to be $30 million. The choice of clients is crucial. We think this business can grow at a compounding growth rate of 25% or so for the foreseeable future," he said.
“While $15 billion will take us time, the next goal would be to get to $1 billion of revenue. Compounding is magical. We will get there, sooner than later," Velamakanni concluded.
