From a consistently high Net Promoter Score and the incremental business our clients keep rewarding us with, our NPS was 77 in the December quarter. Our net revenue retention, or NRR, was 114% in Q3 and 115% for the 9-month period, clearly demonstrating how our clients are expanding their engagements with us on the back of impactful outcomes we continue to deliver to them.

Q3 2026 revenue growth of 21% was driven primarily by 14% existing client expansion and strong new client additions contributing to 8 percentage points on revenue growth this quarter. Moving on to profitability, Aswath will get into more details, but I'd like to highlight just two points.

Firstly, our gross margin in Q3 was 47%. As we accelerate revenue growth, we want to continue expanding our gross margins, which are already best-in-class.

Secondly, we crossed the INR 1 billion mark in quarterly profit after tax. I'd now like to walk you through some business highlights.

Fractal won the Microsoft Partner of the Year 2025 Award in the Retail and Consumer Goods category. This award recognizes partners that have delivered significant impact with AI on Microsoft Cloud and AI stack over the past year.

We're excited to announce that we have secured a preferred supplier status with two of the Magnificent Seven clients. Lastly, on the R&D front, let me share some updates.

We launched Vaidya 2.0 at the IndiaAI Impact Summit. Vaidya is our verticalized foundation model for healthcare under the IndiaAI Mission, aimed at enabling population-scale transformation.

Vaidya 2.0 is the world's first model to achieve a 50+ score on the OpenAI HealthBench Hard, a high bar, a real-world benchmark. Healthcare is now our fastest-growing vertical with 78% year-over-year growth, and we are extending Vaidya's capabilities across healthcare and life sciences and pharma clients.

We also launched Pi-Evolve, an agentic engine for autonomous machine learning and scientific discovery. Pi-Evolve ranks among the world's top-performing agents on OpenAI's MLE-bench, becoming the first evaluated agent to cross 60% score.

Pi-Evolve will be available to Fractalites and clients very soon, helping us accelerate machine learning solutions for our clients. With this, I conclude my remarks and hand it over to Aswath for providing greater color into financials for FY26 third quarter and 9 months for FY26.

Thank you, Srikanth, and good morning, everyone. We delivered a strong quarter of profitable growth.

Our current quarter revenue from operations grew by 21% year-over-year and 7% quarter-over-quarter to 854 crores. On a constant currency basis, growth was 14% year-over-year and 5% quarter-over-quarter.

For the first 9 months of fiscal '26, growth was 20% year-over-year and 15% on a constant currency basis. This growth was entirely organic.

Three components drove our growth: first, growth from existing clients as reflected in net revenue retention; second, growth from new clients; and third, client churn. All three components are calculated on a trailing 12-month basis.

Net revenue retention, or NRR, reflects the growth with existing clients, similar to same-store sales growth in retail industry. In our revenue growth of 21% for Q3 '26, 14 points was from existing clients as reflected in NRR of 114%, 8 points of growth came from new clients added in the last 12 months, and we had half a point of churn.

We experienced particularly strong growth by way of new client additions primarily in our healthcare and life sciences and consumer packaged goods, or FMCG, industry verticals. For the first 9 months of this fiscal, NRR was at 115%, indicating 15 points of growth from existing clients, new clients contributed to 6 points of growth, and we had 1% churn.

Adding 10-20-30 clients, also referred to as must-win clients, is a key growth strategy for Fractal. As of December 2025, we are working on 127 must-win clients, up from 113 in March 2025.

The revenue share from these clients has increased to 83% in Q3 fiscal '26 compared to 81% for fiscal '25. We have scaled our greater than $1 million clients from 53 in March of 2025 to 58 clients as of Q3 fiscal '26, as measured by revenues on a trailing 12-month basis.

The clients contributing more than 20 million in revenues has increased from 5 to 6 during the same comparable periods. Consistently delivering value to our clients and scaling our business with them remains our North Star.

Diving into revenue growth, first by industry segments, growth was primarily driven by 78% year-over-year increase in healthcare and life sciences industry, followed by 26% growth in banking and financial services industry. Growth in healthcare and life sciences industry was a result of strategic investments for building capabilities to serve our clients.

Decline in tech and media telecom industry was primarily driven by a telecom client in Australia and a technology client in the United States. Our CPG and retail clients faced tariff-related headwinds in the first quarter of this fiscal year, which impacted their spending.

Excluding CPG and retail, our growth would have been 25% for the year-over-year. Secondly, we measure our revenue by geography based on clients' billing location.

In Q3 of fiscal '26, Americas and Europe both grew by 26% year-over-year, whereas APAC declined by 6% owing to the same telecom client mentioned earlier. For the first 9 months of fiscal '26, Europe has grown by 37% as we scaled our business with existing clients.

To conclude, our revenue per billable headcount increased to $85,000, representing an increase of 6% in rupee terms and 2% in dollar terms, as measured on a trailing 12-month basis. Now moving over to profitability, I will start with gross margins.

We define gross margin as revenue from operations less direct costs, which includes both employee benefit expenses and other direct expenses. Our Q3 '26 gross margin expanded by 17 bps year-over-year to 47.2%.

This comprises of 115 bps benefit arising from change in mix of engagement type moving towards output-based contracts, price increases, and productivity improvements. The benefit was partially offset by the net impact of annual merit increases and weaker rupee.

Gross margin for the first 9 months of fiscal '26 expanded by 110 bps year-over-year to 46.3%. Increase in mix of output-based contracts along with strong growth in Fractal Alpha contributed to this gross margin increase.

Now moving to adjusted EBITDA. In Q3 '26, adjusted EBITDA was at 17.8%, representing an increase of 43 bps year-over-year.

This was driven by SG&A coming down by 30 bps to 25.3% of revenue along with gross margin expansion. Adjusted EBITDA for the first 9 months of fiscal '26 was at 16% versus 16.4% in the same period of last year.

While our gross margin expanded by 110 bps, growth investments in relationship management for key clients and opening of new offices led to 180 bps uptick in SG&A as percentage of revenue versus last year. Excluding the impact of R&D spend, which we expense in our P&L, our adjusted EBITDA would have been 22% for Q3 '26 and 20% for the first 9 months of fiscal year '26.

Our current margin takes into account necessary investments to benefit from the massive AI opportunities which lie ahead of us. Moving to Fractal Alpha, I would like to highlight our journey of rapid growth and profitability improvement.

Fractal Alpha has grown by 51% year-over-year in the first 9 months of fiscal '26, with Asper growing at 43% and Analytics Vidhya at 69%. Gross margin for Fractal Alpha has expanded by 276 basis points year-over-year.

The loss in the segment for the same period has come down by 51%, while investments into R&D and sales and marketing have continued. The losses in Fractal Alpha have been coming down since fiscal '23.

Our segment loss was at 54 crores in fiscal '23, which reduced to 44 crores in fiscal '24, further reducing to 26 crores in fiscal '25, and it currently stands at 10 crores for the first 9 months of fiscal '26. Now I would like to move over to profit after tax.

Profit after tax for Q3 '26 was at 100 crores or 11.7%, versus 92 crores or 13% for the corresponding period last year. Profit after tax of 100 crores in Q3 '26 increased by 10% year-over-year despite increased losses from associate company and lower other income on account of forex losses.

In Q3 '26, we created a deferred tax asset of 50 crores to reflect the tax benefit from carryforward losses in the USA. Qure.ai is an associate company where Fractal owns 31.5%, and as per Ind AS, we account for the share of our profit or losses in our consolidated P&L proportionately.

Qure.ai has been facing headwinds with cuts to USAID, which has led to higher losses versus previous year. Our share of losses from Qure.ai stood at 19 crores or 2.2% of revenue for Q3 '26, versus 3 crores or .4% of revenue for the same period last year.

In addition to this, because of higher forex losses, other income went down from 24 crores or 3.4% of the revenue in Q3 '25 to 2 crores or .2% of the revenue in Q3 '26. Excluding the increased loss from Qure.ai and reduction in other income, our profit after tax would be 138 crores versus 92 crores for the same period last year.

In Q3 '26, ESOP charges, including options-linked cash bonus and retention bonus, declined to 2.8% of the revenue versus 4.9% of the revenue for the same period last year. ESOP charges, including options-linked cash bonus, has come down from 9.9% of revenue in fiscal '23 to 2.3% of revenue in the first 9 months of current fiscal.

Now moving over to cash, we generated 129 crores of cash from operations, which was 16% higher than the same period last year. This was primarily driven by 14 days reduction in days of sales outstanding from 92 days in the last fiscal to 78 days in the current fiscal.

We include billed and unbilled trade receivables to calculate DSO. For the first 9 months of fiscal '26, cash from operations came in at 108 crores versus 120 crores at the same time last year.

IPO-related expenses to be recovered and GST refund timing impacted the cash from operations adversely by 57 crores. Adjusting for this, cash from operations for the 9 months would be 165 crores, which represents 38% year-on-year growth.

As of December 31st, 2025, we had cash and cash equivalents of 816 crores.

In summary, overall, we had a great quarter where we delivered 21% revenue growth, 47% gross margin, 18% adjusted EBITDA, and 100 crores of profit after tax. With that, now we can move over to your questions.

Back to you, moderator.

Thank you. Ladies and gentlemen, we will now move to the Q&A segment.

Participants are requested to please use the raise hand icon located at the bottom toolbar on your screen. When called upon, you will receive a prompt to unmute.

We will wait for a moment while the question queue assembles. Ladies and gentlemen, if you wish to ask a question, you may click on the raise hand icon.

We'll take the first question from Manish Adukia of Goldman Sachs. Please go ahead.

Hi, good morning. Thank you for taking my questions, and congratulations to the entire team for the listing.

A few questions from me. Firstly, since this is your first earnings call, would you be able to give any kind of indicative, if not specific, some kind of range of guidance in terms of what your aspirations maybe from a revenue growth perspective over the next maybe 1 to 2 years, and where do you see adjusted EBITDA margins for the Fractal AI segment to maybe get to in the next couple of years, and what the building blocks for that may look like?

That's my first question, please.

Thank you, Manish. We have seen a significant AI-related expansion as enterprise adoption of AI takes off.

Today, enterprise adoption of AI is somewhat moderate because accuracy of the AI systems still don't consistently match and exceed human performance in every place. This is changing every day, and as more and more AI solutions become feasible and inexpensive in the context of large enterprises, adoption takes off.

We do expect that our revenues will continue to accelerate at a faster pace. Historically, we have grown at 30% year-over-year for the last 10 years, and even in the last 5 years, our growth rate, CAGR, has been 29%.

We see an amazing opportunity in terms of continuing the kind of revenue growth that we have seen historically. That is number one.

In terms of margins, as a public company, we are clear that the expectations from the investors is to expand our net income and our EBITDA margins. As we accelerate our revenue growth, we want to make sure that our gross margins, which we see as one of the most important indicators of the quality of the business, we want that to expand, and then that translating into both EBITDA expansion as well as PAT percentage expansion as well.

That's what we expect, especially also because some of our ESOP charges and others that were earlier part of our P&L, as a percentage of revenue, we expect them to decline. Overall, we expect EBITDA margins and PAT margins to expand while gross margins also will expand.

This is the kind of objective that we're working with: accelerate revenue growth while expanding gross margin and continue to expand EBITDA as well as profit margins.

Thank you for that response. My other maybe related question to that was your gross margin, obviously, like you mentioned, best-in-class at around 47-odd percent, which is significantly higher than at least the services companies that are listed in India across market cap.

When we look at, let's say, your adjusted EBITDA margin today, which is maybe lower than some of those companies, is there any reason why over a longer period of time, given your gross margin profile is so strong, you're moving more towards output-based contracts, plus there's a significant focus on your products business, that your EBIT or EBITDA margin also should not be higher than a typical services company, given your gross margin already is significantly higher than those companies? Would love to hear your thoughts on that.

No, thank you. I'll add something, and then maybe I request Aswath to continue to add to my answer.

Number one is that, yes, we expect our gross margins are our industry best, but we want to continue expanding our gross margins, number one. Number two is from gross margins to EBITDA margins, we have SG&A as well as our R&D expenses.

Like Aswath also pointed out, we take roughly 4% of revenue, 4.1% of revenue, into our expenses today, which drops the overall EBITDA margins by 4 points. When you're comparing us with other firms, this 4 points have to be adjusted for because most of them don't have any AI R&D today.

That's number one. Second is that our SG&A, we expect to continue to make improvements on SG&A as a fraction of sales, because we see productivity improvements in sales as well as G&A.

We have a whole host of AI-led programs within Fractal to improve sales productivity. We launched a tool internally called PitchDoc, which helps our salespeople sell better and prepare for meetings much more effectively, much faster.

It drops the time from days to minutes, and that we expect will help us in improving our sales productivity overall. Secondly, on G&A, for example, in hiring, one of our big costs is the cost of hiring and the time it takes to hire.

We have implemented a solution called Ikigai within Fractal, which is a way in which we can completely build the hiring process, recruitment process, and matching process AI-driven. Right from a recruiter agent who can create the job description, to a candidate agent who can have this conversation with the candidate, to a 360-degree assessment which is completely AI-driven, to AI matching engine, we have automated the process of hiring within Fractal, which makes the entire SG&A come down as a fraction of sales.

We expect SG&A improvements as well as improvements on overall gross margin, leading to higher adjusted EBITDA margins at Fractal. Aswath, you want to add?

Just to reiterate some of the points as said, Manish, the key drivers will continue to be how we improve our gross margins. I talked about the mix change more towards output-based, which come with a higher profitability indicators.

Then also SG&A, we have already reduced our SG&A, as laid out in the presentation also, by almost 7 points, and that continues to be the case. We have operating leverage as the revenue scales, we will be able to bring our SG&A down.

Also SG&A will also, one, just the pure operating leverage, but also some of the internal AI tools that Srikanth talked about, be it in our hiring and recruiting side of things, or even in something like sales. Across both sales and marketing and G&A, we will be able to bring our cost down as a percentage of revenue, while continue to invest in R&D.

The main answer why at times our adjusted EBITDA looks a little different, obviously there is 4.1% of R&D expense in that. Second, we have shown operating leverage or ability to reduce the SG&A as a percentage of revenue over the last 3 years.

There's no reason why we can't continue to do that. Our adjusted EBITDA 3 years ago used to be 7%, last quarter was 18%.

We expect that to keep increasing as we scale.

Thank you for clarifying that. My other question was on these two client issues that you called out in your TMT vertical, one in Australia and one in the US.

Are you able to provide us any more color on what kind of issues these might be, and are these business as usual in your business, or was there any kind of, let's say, surprise or one-off here? Any color would be helpful there.

See, our business is dependent on client success, and sometimes if clients are having a challenge in their own business, that will reflect on us as well. One of the clients that we spoke about here has been going through a restructuring inside their own organization, and therefore, that has impacted us.

It is not a typical Fractal client in terms of scale, and therefore, that has also been the reason for volatility on that client relationship. Typically, Fractal works with the 10-20-30, which is 10 billion in revenue, 20 billion in market cap, and 30 million customers.

These are very large companies which do not usually see any volatility in their business. They're very stable and very profitable and so on.

This client has been an exception to that, in that sense. Then the one in APAC, I think it's been a case of decreasing their business with us.

These do happen from time to time. I would not say this is not business as usual.

This can happen, but overall, this is reflective in the overall churn number, which is roughly 1% of revenue for Fractal in any given year. That is how we have to process that part of the exception.

Mr. Adukia, may we request you to return to the queue?

Thank you.

Sure, we'll do. Thanks a lot.

Thank you. We'll move to our next question.

That's from Gaurav Rateria of Morgan Stanley. Mr.

Rateria, you may ask your question now.

Am I audible? Just want to confirm.

Yes, sir. Please go ahead.

Thank you. Congratulations, Srikanth and team, for the listing.

I have a couple of questions. My first question is in face of changing technology landscape, how do you think the perception of Fractal has changed in front of clients?

I know that you spoke about metrics like NPS and NRR, which gives us some comfort.

Just trying to understand that the clients may be also trying to have a list of strategic partners for providing the AI-related services, which may not necessarily coincide with their incumbent vendors. How's Fractal perception changed in last 1 or 2 years in wake of the technology landscape changing?

The second question is on the engagement models. You did talk about the increase in output-based models.

What are these various outcome-based models that you can underwrite because of your capabilities, which may not be possible for other vendors to underwrite that? The last, you talked about the R&D investments.

We are currently probably at 4% plus. Where do you think in the medium term these R&D investments should stabilize at, and which recognizes our capability, but at the same time also gives us some operating leverage?

Thank you. Yeah.

First on the perception of clients, Gaurav, we have been a pure-play AI firm for a very long time. What has happened on the client front is that they need much more specialized, higher-quality vendors today than ever before.

As AI progress accelerates, really the ones who think of AI as also one of the levers are struggling. They don't have AI R&D spends, they don't have significant capabilities, and therefore, a generic competition tends to suffer in a place where significant progress is happening on a week-by-week basis.

From a client perception standpoint, Fractal's perception remains very strong and has actually accelerated, because they see the incredible amount of AI products and research outputs that we produce, and they see as best-in-class in the AI field, especially in the enterprise AI space. There's no one like Fractal today.

They see that, and that helps us in expanding our business with them, especially as they think of reimagining their workflows with AI. The biggest opportunity in AI is that every single business workflow can be reimagined with AI.

This is a thing which requires not only deep AI expertise, but significant domain expertise and the ability to navigate this enterprise architectures, enterprise landscape, and so on. That is why Fractal is a very, very credible, one of the very few credible names in this space.

That's why we see a continued expansion with the clients and interest. Even where people think of, "Oh, I have too many vendors, let me consolidate vendors," Fractal has a special place because there's no one like a pure-play AI vendor for most of the clients except Fractal, and that's what makes them very reliant on Fractal.

That's number one. Number three question was on AI R&D spends.

Let me answer that. Our AI R&D spends, we expect to continue investing in AI R&D spend.

In fact, we hope that we can continue to expand the amount of investment we make in AI R&D. This is a significant part of our overall credibility as a company in a place where AI is changing every day.

The kind of R&D investments that we're making helps us signal to clients the kind of incredible work that we're doing and the kind of capabilities that we have. We expect these investments to continue, and we expect these investments to accelerate our revenue growth and our gross margins.

The way we are thinking of it is that as we expand revenue growth and as we expand gross margins, some of that gross margin expansion will be plowed back into increasing the AI R&D spend. That is how we are thinking of how to build Fractal for the future.

Now, I have forgotten your second question, but maybe, Aswath, you want to answer that?

Yeah. I think the question was on the engagement type, moving more towards output-based metrics, Gaurav, that's what we said.

Yes, we have some outcome-based too, but that's quite limited at this stage. What we're seeing is that more versus input-based, where it is time-based, we see that the mix is changing more towards the output-based metrics.

That's a consistent trend that we've seen over the last 7 to 8 months. Output-based metrics, we can use our internal productivity and other tools to ensure that margins are pretty high.

We manage the whole project end-to-end, and the margin quality of output-based projects tends to be higher, it's much more beneficial for us. Just on the credibility with the clients, only one metric that I'll indicate is if you look at our scaling-related numbers, like be it the total number of MWCs clients, that's expanding, total of clients above 1 million is expanding, total clients with above 20 million is expanding, above 10 millions is expanding.

That, I think, is a good way to look at how we are a credible AI player in front of our clients.

I'll just add to the output-based, etc. We are extremely confident that output-based and outcome-based models will actually expand our gross margins.

We are actively re-initiating conversations wherever we feel like we can convert any existing relationship into more output or outcome or license-driven conversation. The side effect of that is that they are all at higher gross margins than our other business models.

Therefore, overall, this will be a gross margin accretive kind of work that we'll do.

Thank you for the detailed answers. Just to put my second question in context of outcome-based, the reason I was asking is because not every vendor would be well-placed to shift the business model or engagement model to outcome-based.

There will be very few companies who will be able to underwrite that, and probably you guys are well-placed around that. I was just trying to ask that the capabilities, do they give you readiness to move to outcome-based because you are able to underwrite those outcomes which others are not?

That was the context of the question. Thank you for the detailed answer.

Thank you, Gaurav. Absolutely.

We have very credible AI capabilities, lots of AI research, our Cogentiq platform. These help us in reimagining business process with AI and being confident about doing this in much shorter time.

Whenever we have an output-based or outcome-based conversation, we are very eager for that because it gives us an opportunity to expand our margins while delivering better outcomes to our clients faster. Absolutely ready for this kind of a transformation across the board.

Thank you. We move to our next question.

That's from Kawaljeet Saluja of Kotak Securities. Please go ahead.

Hey, hi, Srikanth. Hi, everyone.

Congratulations on your listing. Just a couple of questions.

The first one is for Srikanth. Srikanth, you did allude to the fact that you have grown by 30% in the past and in a way implying that the aspiration is to grow at an elevated rate, whereas your current growth rate is lower than what you aspire for.

What would you attribute the gap between an aspiration versus the current growth rate, and how do you intend to bridge the gap between the two? That's the first question.

Thank you, Kawaljeet. Absolutely.

Our aspiration continues to be to grow. We have, even if you look at the last 10 years, not every year we have grown at 30%.

There are some years we've grown a little lower, some years we've grown much faster. Net of that is the 30%, and this is maintained over every 3 to 4-year period in that the growth rate is 30%.

This year specifically started off with a little bit of the trade headwinds coming from in the CPG vertical, which, as is one of the largest vertical for Fractal. In that segment, we found that there was massive uncertainty especially in April of this year when the financial year just started, where the Liberation Day announcements came in and some of the announcements came in a little before that, which created uncertainties.

When businesses feel uncertain, they kind of wait on their spend, they tend to delay their spend. That is one of the reasons why our growth rate was not as spectacular as we would like it to be.

If you just exclude that, I think the numbers would be around 26% or so, the rest of the business grew. Even if you see for the 9 months, we have grown 26% in the US and 37% in Europe for the 9 months.

They are pretty decent numbers, which just shows you that the potential of this business to grow at the historical dates is there. We have to make sure that we execute well, and obviously there are things that we may not always be able to control in the way things happen, but we do see AI-related expansion as a massive opportunity for Fractal to continue its revenue trajectory.

That's very helpful and that's clear. The second question that I had is on R&D spend.

You did indicate that that's an area that you want to double down on. Now, at a high level, we are aware of the areas where you are spending on R&D, but can you flesh out in more detail where would the incremental spend be allocated on R&D, and how do we assess the efficacy of the spend?

Okay. Now first, in terms of efficacy of spends, Kawaljeet, eventually we have to expand our revenues faster, accelerate our revenue growth, as well as expand our gross margins.

Eventually, this is the most important indicator of AI R&D's success if you take a longer term. In the medium term to shorter term, we have to achieve benchmark results on everything that we do.

The way we are thinking of the AI R&D spends is it has to either help our Fractal.ai business expand its revenue growth. Therefore, for example, the Pi-Evolve machine learning agent is something that we can use across the problems that we solve today and dramatically accelerate our productivity, and therefore expand our margins.

That is one such example. Second place we want to use AI R&D is on our products.

For example, the Cogentiq platform, the Asper platform. Here, we want to make sure that we're investing AI R&D, to build the best-in-class agentic AI platforms, and this is an incredibly big opportunity and also an opportunity where things are changing pretty fast.

We want to make sure that we have the best-in-class platform to help the biggest companies in the world reimagine their workflows with AI. That's the second place where we want to connect the AI R&D to that.

The third one is a place where we are building AI research on, let's say, Vaidya, which is a healthcare area. It has impact on our healthcare segment, but also it is something that we're thinking of as part of IndiaAI Mission to build a India-level model that the 1.4 billion people of this country can use.

If we are very successful with that, and if we're able to launch with the IndiaAI Mission, it could be game-changing, and that feedback from that 1.4 billion people using could help us achieve frontier AI and India achieving frontier AI. That's a big dream, that's the third aspect.

These are the three places where we are sort of dividing the spends: number one, expand Fractal.ai and improve the overall gross margins and revenue growth; number two, build the Cogentiq platform, again helping companies reimagine their AI workflows; and three is building this Vaidya platform for IndiaAI or India healthcare. These are the three places where we are apportioning the spends.

Aswath, if you want to add to that, please feel free. Nothing much to add, Srikanth.

That's a fantastic response, Srikanth. Just a final question on agentic AI platform, and that's fascinating as well.

Can you just Of course, what it seems is that you're working on building homegrown agents, can you just tell me the underlying models that you're using? Is it Llama or any of the open source models?

Second is that in a client environment, Srikanth, what do you think eventually what will work? Would it be the frontier models, a company's agents, which players would use to deploy in a client environment, or would clients be comfortable working with, let's say, the homegrown agents of players like you?

If it is your agents, then what is the pricing model that you have adopted? Is it an all-in pricing including token costs?

There are multiple thoughts that I have on agentic AI, maybe just some higher thoughts would be helpful.

Thank you, Kawaljeet, and very, very good questions. Number one is that the way we've thought about Cogentiq is that we're not competing at the model layer with OpenAI or Anthropic or Claude, etc.

One very important metric to think of is, if OpenAI launches their new model or Claude launches their new model, should we be excited or should we be scared? The way we built Cogentiq is that we should be excited, because any underlying model improvement dramatically improves Cogentiq's performance.

Cogentiq is an ontological layer on top of these models, and then an agentic layer to go and solve enterprise workflows, reimagine them. If the underlying base model improves, all of the stuff that we do on top of that automatically improves, and therefore we can get better results, much faster results.

The way we price this today is that we do not charge them for the underlying model usage. That is an expense that they incur directly from the underlying model providers.

They can use the Fractal Fathom models, etc., which are fine, but they can use OpenAI's GPT-5 or Claude 4.6 Opus or any other underlying model that makes sense to them. The Cogentiq platform builds the ontological layers as well as the agentic layers to help them reimagine their workflows.

The way it works is that we have built a bunch of agents which are specialists in a certain tasks. These agents have access to a whole host of machine learning tools, and these machine learning tools directly connect with enterprise data sources, including SAP and etc.

Therefore, when these agents are able to access machine learning tools and access data, they're able to solve these problems. This is something that Fractal is uniquely positioned to do, because we have deep understanding of the enterprise landscape, the problems they're solving, the data flows that they have, and what's happening with their business.

That really puts Fractal in a really good position. We charge them on a license basis, on a output basis, but not on a model consumption basis, because the inference token cost will be directly charged to them through the model provider, unless they're using our model.

That's fascinating, Srikanth. Just a final question.

In an agentic system, the biggest problem is a compounding of errors. With your agents, I think current agent accuracy is also not something in which an agentic layer can work very well.

How much is the human-in-the-loop involvement, and second, if you can just dwell upon how frequent is the agent drift problem in a client environment?

The last part, I couldn't get.

Agent, last sentence? Agent drift, yeah.

Drift, okay, got it. See, you're bringing up a very important point, and this is one of the reasons why the enterprise adoption of AI is still not taken off in a very significant way.

When you think of consumer AI, error is okay. If you have 50% hallucination rate or 40% hallucination rate, it's fine.

It works just fine. In case of enterprise AI, we have to consistently meet and exceed human accuracy in that process.

When we're building systems, they're not always building these systems to be 100% autonomous, because that will create the drift that you spoke about, create accuracy issues, etc. We built the system to be human plus machine automatically, so that the agent drift or the model error, which get compounded, etc., those are actually solved for from the human input and human approvals, etc., that are required, number one.

Number two is that we build this in such a way that as AI gets better, the need for human input comes down, accuracy automatically improves. In processes where overrides are possible, for example, humans can override any AI output, but then we keep track of when is the override working and when is the override actually a bad idea.

After some time, people start to figure out that the human overrides are not necessary, in fact they're detrimental, and therefore that increase the trust in the AI systems. Our goal is to build trusted AI systems which take care of these kinds of issues, and as underlying model improvements happen, all of this continues to improve in the right direction.

Thank you. Before we take our next question, we'd like to remind participants to please limit your questions to two questions per turn.

Time permitting, you may come back in the queue for a follow-up question. We'll take the next question from Pritesh Thakkar of PL Capital.

Please go ahead.

Congratulations on listing and on good set of numbers. My first question is on Fractal Alpha.

There is a sequential decline in the top line that we saw in quarter three. Is it something related to cyclicality or is there any challenge in this or how would you put it?

I can take that. In Fractal Alpha, it's made up of Asper and Analytics Vidhya.

In Analytics Vidhya, our Q2 tends to be a place where we do a big DataHack Summit, which creates revenue, and generally, it's cyclical in nature. Most of the years, it tends to happen in the second quarter.

In the Asper side of the business too, as I mentioned, in CPG and retail with all the tariff-related tumult or tantrums, clients were a little bit slower to sign up for new contracts. As some of the initial setup phase revenue came off, there's a sequential decline in Asper too, which is not really a seasonal thing.

What we saw in Analytics Vidhya is more seasonal, which we expect even next year, but Asper, we did see the new clients addition was slower in the first 9 months of the year. That led to small sequential decline in Asper.

Understood. Just wanted to understand the normal course of our performance of our business in H2.

Is it Q3 or Q4 tends to have more weightage in terms of higher growth and better margins? This is my first question.

Secondly, when do we consider compensation revision? You also highlighted in your remark also you've given out some compensation revision this quarter itself, but what is our usual compensation revision during the quarter?

Historically, we have done compensation revision as of 1st of April. When I talked about Q3, I talked about obviously previous year to this year Q3, hence there is that impact of merit increases.

It always happens in April, the reason you see the profitability being higher in second half of the year is because as we keep growing quarter-on-quarter, margins tend to keep going up with both operating leverage as we scale further. In the first quarter where there is merit increase, the margins in that particular quarter tends to come down and then picks back up again in second quarter, third quarter, and fourth quarter.

That's how we see. As I said, historically, we have always done it as of 1st of April, but we have a very set timeline for the next year right now, but this is something that we review every year.

We are not missed a year at least in a while right now.

Just wanted to understand, on Q3 or Q4 tends to be more weightage in terms of higher growth or better margins, how should we consider Q4

then? No, we don't really see any kind of budget flush kind of stuff which other people talk about in Q3 specifically.

There's no real seasonality in terms of quarter-on-quarter growth, what we expect is to keep growing quarter-on-quarter at a steady pace over a period of time. It's not that we expect, okay, Q3 to be pretty big because a lot of people tend to spend the budgets.

Those seasonal effects are very minimal for us, I can't say that there's absolutely no impact, but very, very minimal for us in terms of growth. Our expectation is that quarter-on-quarter, we keep scaling our revenues, and in one quarter where there is merit increase, obviously there will be impact on the margins.

That generally tends to be the first quarter of the year, so as the year progresses, margins keep improving.

Lastly from my side, as we move into next year, what are the lead indicators that we should focus on? If you can highlight anything in terms of TCV or executable order book that we have carried in the books that gives us visibility for next year growth.

We look at the must-win clients that we have, as of December was 127, that's continues to be growing, that's a great lead indicator, and 58 clients above million, that's also great lead indicator. Net revenue retention is another great lead indicator.

We generally enter the year with around two-third of visibility from both order book and the renewals and weighted pipeline, and we are seeing similar trends versus the previous year. The two-third of the visibility into the next year.

We are not specifically reporting order book and TCV because business is not run like that. We are more focused on how do we grow our net revenue retention and growth from new clients.

That's where the focus is, and that's what we look at as the lead indicator for our future success. In terms of visibility, we enter the year with almost two-third of the revenue visibility into next year.

Understood. Thank you so much and best of luck.

Thank you. Our next question is from Abhishek Shindadkar of InCred Capital.

Please go ahead.

Hi, thanks for the opportunity and congrats on the listing. Three questions.

The first one is on, let's say, the competitive moat. Can you elaborate what is our long-term defensive moat for our vertical-specific models that we are trying to build to eventually kind of avoid any cannibalization?

That's the first. The second, you talked about the monetization, which is moving to a license fee.

At points, is there a possibility that we may get competition from the frontier models? That's the second.

The third is especially for the IndiaAI Mission work that you're doing,

the data sets and the compute requirements for us are solved from Indian providers, or how do we go about it? These are my three questions.

Thank you for taking my questions.

Thank you, Abhishek. On the third question first, for the IndiaAI Mission, we will have compute coming from the approved vendors of the IndiaAI Mission, and we get a specific preferential rate on that H100s that we end up using for that, and we also get a subsidy from the government.

They fund part of that negotiated rate. That's how the compute is set up.

There are a few providers that IndiaAI Mission has selected and has pre-agreed negotiated rates. We get those rates in the way we use the compute for the IndiaAI Mission-related work.

That is on your third question. Can you please remind me the first question?

Moat and

Moat. Competitive moat.

On the competitive moat of Fractal across domains, number one is that we have a deep understanding of the domain. We know the enterprise workflows.

We have solved these problems multiple times. For example, think of revenue growth management, understanding all the complexity of thousands, sometimes millions of SKUs across all the geographies, and understanding how pricing, promotion, and revenue growth and strategic revenue management happens inside companies.

We understand that really, really well, and we're able to bring in very solid AI models that are trained or fine-tuned on that vertical for them, which makes it very advantageous for companies to work with us. Again, secondly, we have to make this work inside an enterprise context, which means that we have to know all the data that is available inside the enterprise.

This is dark data that others don't get to see, and our proprietary data sets that the models have never seen. We see those data sets, we're able to bring those data sets into the way we solve problems.

The entire work of Fractal is inside the enterprise data systems, right? We can build models, train models for the specific context of the enterprise data.

This is a very defensible moat. Fractal also has a deep design capabilities, which we call as behavioral sciences capabilities, just to understand human behavior and how humans actually make decisions.

When we combine that with deep understanding of AI models, that creates a very unique competence that at least we haven't seen anyone else match as of right now. These are the things that make us very competitive in the space in every vertical that we choose to operate in.

We do think of it carefully in terms of when we expand in a new vertical. We'll do that when we know that we have something very credible, or we intend to build something credible in that space.

The last one is on the competition from the frontier models. The frontier models obviously are expanding everywhere at all at once, and they're very ambitious.

They also realize that they need partners and they need people who can build on their generic capabilities, which are very, very good. When we built our Cogentiq platform, we build an ontological layer on the models per se, and then help them reimagine their workflows, which is again extremely challenging for the OpenAI and the frontier models to directly do themselves.

If you think about a comparison here, Palantir is doing exactly that, and that's working really well for them. They're able to drive high gross margin.

We expect a similar way that this will be a very collaborative ecosystem. Of course, there'll be some competition too, but we see a huge opportunity for us to take underlying powerful models and build on top of that the kind of platforms that we're building, including Cogentiq.

Thank you. Thank you very much.

Ladies and gentlemen, that was our last question, and with that, we conclude the question and answer session. Before I hand over the call to Anjali Garg from Fractal's Investor Relations team for closing comments, request Srikanth to share his final remarks.

Thank you, Inba. We had a great December quarter with improvements across almost every metric: revenue growth of 21% driven by strong growth in life sciences and healthcare, as well as banking and financial services verticals.

Our strong client relationships are evident from the NRR, our net revenue retention of 114% for Q3 and 115% for the 9-month period. On the profitability front, we reported a best-in-class 47.2% gross margin, and our PAT crossed the 1 billion milestone.

Lastly, we continue to invest heavily in R&D, focused on foundational AI research as well as AI products. In addition to helping us build solid capabilities that help us solve our clients' most pressing business challenges, these investments will create significant growth opportunities in the future.

Looking ahead, we're very excited by the opportunities that the AI revolution is opening up for us. As a pure-play, AI-native company that has significantly invested in AI R&D, we're very well-placed to fully benefit from these opportunities, drive exponential growth, and create tremendous value for our shareholders.

Over to you, Anjali.

Thank you, Srikanth. Thank you, everyone, for joining us on our first earnings call as a public company.

This is a significant milestone for Fractal, and we are excited about the opportunities ahead. If you have any further questions, including any that we were unable to address during the call today, feel free to reach out to us at [email protected].

We look forward to seeing you again next quarter. Thank you once again, and wish you a good day.

Thank you. Thank you, everyone, for joining us today.

You may now click on the leave button to exit the meeting. Goodbye.