InfoBay AI (formerly EduGorilla): Its FY26 Revenue Beat the Valuation Model's FY29 Forecast

Two numbers, both defensible, eleven months apart.

In October 2025, an IBBI-registered valuer signed off on a fair value for this company of ₹62,035.54 a share. Unlisted dealers now quote it between ₹3,75,000 and ₹4,00,001. That is roughly six times the valuation, and the obvious reaction is that somebody has lost the plot.

Then you open the audited FY26 accounts, and the gap starts making sense.

The valuer's discounted cash flow model projected revenue of ₹56.47 crore for the year ending March 2026. The company actually booked ₹107.78 crore. It did not just beat the first year of the forecast by 91%. It beat the forecast for FY29, a year that is still three years away.

That is the headline. The finding that changed my mind about the business sits further down, in the expense notes, where the margin story comes apart under its own arithmetic.

InfoBay AI at a glance

Detail	Position as at 16 September 2026
Legal name	InfoBay AI Limited, formerly EduGorilla Community Private Limited
Business	Training data, annotation and evaluation services for large language models
Incorporated	23 December 2016
ISIN / CIN	INE0SE901014 / U74999UP2016PLC088614
Unlisted share price	₹3,75,000 to ₹4,00,001, varying by dealer
Shares outstanding	23,352 (18,018 equity, 5,334 preference), face value ₹10
Market capitalisation	₹876 to ₹934 crore
FY26 revenue	₹107.78 crore, up 123%
FY26 profit after tax	₹48.62 crore, up 199%
P/E on FY26 earnings	18.0x to 19.2x
Net worth / borrowings	₹88.05 crore / nil
Promoter holding	45.00%
IPO status	Listing process under way; DRHP not filed
Registrar	Maashitla Securities Private Limited

What this company is now

InfoBay AI Limited was incorporated on 23 December 2016 as EduGorilla Community Private Limited, selling online mock tests and study material for competitive exams in vernacular languages. It became InfoBay AI Private Limited in December 2025 and InfoBay AI Limited in January 2026. That second change, from private to public limited, is the step companies take when a listing is being prepared.

The pivot is essentially finished. The FY26 segment note puts it plainly: the AI business contributed about 98% of revenue, the education business under 2%. Management did not present education as a reportable segment because it no longer meets the quantitative threshold.

What replaced it is a training-data business. The company sells curated, expert-verified datasets, annotation infrastructure and evaluation systems to firms building and fine-tuning large language models, including multilingual speech corpora, medical imaging datasets, reasoning-trace datasets for supervised fine-tuning, and factuality auditing of finished models.

Scale claims vary depending on where you read them, which is itself worth knowing. The company's own site cites 3.6 million hours of audio and 2.5 million patient records. Platform write-ups cite 2.1 million hours across 35 languages and 1.6 million patient medical records. Nobody audits a corpus. Treat all of it as a description of ambition rather than a verified asset, and note that a dataset built on millions of patient records carries privacy and consent obligations that a prospectus would have to address in detail.

One correction while we are here. The company's own investor material dates the journey to 2019. The certificate of incorporation says December 2016, and the CIN carries the same year. Small thing, but if a pitch deck and a statutory filing disagree, the filing wins.

The three-year arc

These come from the audited statements for the year ended 31 March 2026, signed 19 August 2026 by Arora Prem & Associates.

₹ crore	FY24	FY25	FY26
Revenue from operations	11.41	48.41	107.78
EBITDA from operations	2.00	25.18	63.39
Operating margin	17.5%	52.0%	58.8%
Profit before tax	(0.56)	22.77	65.17
Profit after tax	(0.19)	16.27	48.62
Net worth	3.90	36.98	88.05

Revenue is 9.4 times what it was two years ago, a compound rate above 200%. The company was loss-making in FY24, turned its first profit in FY25, and tripled that profit in FY26. Net worth went from ₹3.90 crore to ₹88.05 crore in the same window.

The balance sheet is cleaner than the income statement, which is unusual for a company growing this fast. No borrowings at all at year end, against ₹0.44 crore a year earlier. Cash and fixed deposits of ₹58.95 crore. Operating cash flow of ₹35.76 crore against reported profit of ₹48.62 crore, so about 74% of earnings converted to cash, with most of the shortfall in receivables and advance tax. The company parked ₹32.37 crore into deposits during the year, which is what a business does when it is generating more cash than it has decided how to spend.

Where the margin expansion actually came from

Every write-up on this company quotes the margin expansion. None of them work out where it came from.

The bull case for this business, as it is usually written, runs like this: the marginal cost of licensing an existing dataset to a second buyer is near zero, so margins expand automatically as the corpus gets re-sold. Operating margin did expand, from 52.0% to 58.8%, so the story appears to check out.

It does not. Work the bridge line by line and the expansion came from somewhere else entirely.

Cost line, as % of revenue	FY25	FY26	Effect on margin
Content cost	14.1%	25.9%	−11.9 pts
Commission	8.4%	0.0%	+8.4 pts
Materials consumed	7.6%	1.2%	+6.4 pts
Employee cost	6.9%	5.8%	+1.2 pts
All other expenses	11.0%	8.3%	+2.7 pts
Operating margin	52.0%	58.8%	+6.8 pts

Commission fell from ₹4.05 crore to ₹79,000. Materials consumed fell from ₹3.66 crore to ₹1.27 crore. Those two lines are the old edtech business: the cost of distributing test-prep subscriptions and printing books. Between them they handed back 14.8 percentage points of margin simply by ceasing to exist.

Meanwhile content cost, which is the cost of goods for the business that actually earns the money now, went from ₹6.81 crore to ₹27.95 crore. It grew 310% against revenue growth of 123%, and cost 11.9 percentage points of margin.

So the picture inverts. Margins did not expand because data licensing scales for free. They expanded because a legacy cost base was retired faster than the new one grew. That distinction matters enormously for FY27, because the legacy tailwind is now spent. Commission is already zero. Materials are down to ₹1.27 crore. There is nothing left to remove.

From here, the margin line depends almost entirely on whether content cost settles around a quarter of revenue or keeps climbing. At 58.8% there is room to absorb pressure, and a business earning that margin has earned the benefit of the doubt. But the automatic part of the story is over, and the next set of accounts will show which it was.

Where the revenue comes from

Of ₹107.78 crore of FY26 revenue, ₹103.01 crore was export services. That is 95.6%. Add export goods and foreign exchange gains and 97.3% of what this company earned came in foreign currency. Domestic services contributed ₹3.44 crore and actually shrank from ₹4.52 crore.

That is not an Indian AI company with some overseas clients. It is an export business registered in Uttar Pradesh, and it should be underwritten as one. Rupee strength hurts it. A pullback in overseas model-training budgets hurts it more.

The disclosure gap that matters most is customer concentration. The accounts do not name customers or give a revenue split by client. For a business where a handful of AI labs could plausibly account for the majority of ₹103 crore of export billings, that is the single most important number not in the file. If a DRHP arrives, it is the first table I will turn to.

There is a second-order risk worth naming, because the industry is arguing about it openly. As frontier models get better at generating synthetic training data, some demand for human-curated corpora may not need replacing. That would compress pricing across this category regardless of how well any one supplier executes. The counter-argument, which the company leans on, is that synthetic data cannot safely replace verified domain data in medicine, law or STEM reasoning. Both positions are reasonable. Neither is settled.

Before you trust any number you find online

One published figure on this company is wrong by a factor of ten, and it happens to be the figure most people check first.

The audited share capital note is unambiguous. As at 31 March 2026 the company had 18,018 equity shares and 5,334 compulsorily convertible preference shares of ₹10 each, a total of 23,352 shares, giving paid-up capital of ₹2.34 lakh. The earnings per share figures confirm it independently: a profit of ₹48.62 crore and basic EPS of ₹29,519.32 only work on a weighted average of about 16,471 equity shares.

One platform lists the share count as 233,550. That is the paid-up capital in rupees, ₹2,33,550, read as a number of shares. It is out by a factor of ten, and because market capitalisation is price multiplied by share count, the error lands straight in the headline.

Share count used	Market cap	P/E on FY26 profit	Price to sales
233,550 (one platform)	₹8,758 crore	180.1x	81.3x
23,352 (audited accounts)	₹876 to ₹934 crore	18.0x to 19.2x	8.1x to 8.7x

An investor working from the first row concludes this trades at 180 times earnings and walks away. An investor working from the accounts sees 18. One decimal place.

For what it is worth, that platform is the outlier rather than the consensus. Three others put market capitalisation between ₹914 crore and ₹930 crore, which is the right neighbourhood. One of those three gets there using the FY25 share count of 23,138 rather than the current 23,352, which is why its book value per share prints as ₹38,054 instead of the ₹37,706 the audited net worth supports. Close enough to be useful, wrong enough to be worth knowing.

The prices are no steadier. On 16 September, four platforms quoted the same scrip at ₹3,75,000, ₹3,95,000, ₹3,98,000 and ₹4,00,001. That is a 6.7% spread on a single day for a security with one ISIN.

The price histories disagree more sharply than the prices. One platform's full history runs five sessions, from 11 to 16 September, during which the quote moved from ₹3,00,000 to ₹3,75,000, and its lifetime high and 52-week high are the same number because the scrip has existed there for under a week. Another shows a rise from about ₹3,10,000 in July to ₹3,95,000 now. A third reports a maximum-period return of minus 3.83%, with a high of ₹4,31,246.92 quoted to the paisa, and separately prints its 52-week high and low as ₹0.00.

Lot sizes disagree too: one, five, and five thousand. A lot of 5,000 shares at these prices would be ₹188 crore and roughly a fifth of the company, so that one is plainly a field nobody maintains.

None of this makes the company uninvestable. It means the data layer around it is thin and new, and anyone sizing a position off a screenshot is building on sand.

One funding claim that is not in the cash flow

Published write-ups describe a recent $2.1 million round and put aggregate historical funding at $6.3 million to $8.4 million. The audited accounts do not support either figure as stated.

FY26 financing activities were negative ₹9.49 lakh, being interest paid. The line for issue of equity shares is blank. No cash came in from shareholders during the year. The equity share count rose from 14,608 to 18,018, but every one of those 3,410 shares came from converting instruments that already existed: 743 preference shares at 1:1, 2,453 at 1:1.065, and 350 debentures of ₹10,000 each. Conversions, not cash.

So any round described as recent must have closed after 31 March 2026, which means it sits outside every figure in this article and changes the cap table again. That is worth knowing before you anchor on a share count.

On the aggregate, total paid-in capital on the balance sheet is share capital of ₹2.34 lakh plus securities premium of ₹24.32 crore, so about ₹24.3 crore has ever been contributed to this company by shareholders. Even at historical exchange rates that is closer to $3 million than to $6.3 million. Published funding totals routinely include secondary transactions, where one shareholder buys from another and the company receives nothing. Useful for gauging investor interest. Useless for gauging the balance sheet.

Who owns it

Promoter holding across equity and preference shares is 45.00%, up from 43.22%: Pushpa Manglik at 27.87% and Rohit Manglik at 17.13%. Those figures match the audited note exactly, and the published pie charts get them right.

The rest of the cap table is where published sources drift. The accounts disclose PEB Educational Services Private Limited holding 11.31% of equity and 22.67% of the preference class, having held nothing a year earlier, and Auxano Entrepreneur Trust holding 13.25% of the preference class, down from 16.59%. Against total capital those work out to about 13.9% and 3.0% respectively. Published cap tables show different names and different percentages, apparently by mixing class-level percentages with whole-capital percentages. Where they conflict, use the accounts.

Two movements are worth flagging. A holder with 6.75% of equity in FY25 shows nil in FY26, so somebody sold out entirely during the year. And the employee stock option charge rose from ₹19.30 lakh to ₹2.10 crore, an elevenfold increase. Dilution is arriving from several directions at once, and the pre-listing register is still moving.

What the price is actually paying for

Using the audited count of 23,352 shares and the range of quotes doing the rounds:

Measure	At ₹3,75,000	At ₹4,00,001
Market capitalisation	₹876 crore	₹934 crore
P/E on FY26 profit	18.0x	19.2x
Enterprise value, net of cash	₹817 crore	₹875 crore
EV / FY26 EBITDA	12.9x	13.8x
EV / FY26 revenue	7.6x	8.1x

For a company that grew revenue 123% and profit 199%, carries no debt, earns an operating margin near 59% and a return on equity between 55% and 78% depending on whether you use closing or average equity, 18 to 19 times earnings is not an obviously stretched price. That is the part the ₹62,035 comparison obscures.

One caution on secondary data. Some pages carry an FY26 EPS near 24,300. The audited accounts report basic EPS of ₹29,519.32 and diluted of ₹20,771.22. The 24,300 figure matches neither, and a P/E built on it will be wrong in either direction.

What would a target price even be based on?

People search for a target price on this company and there is not an honest one to give. No broker covers it, no analyst publishes on it, and any number you see attached to an unlisted scrip is either a dealer talking his own book or somebody extrapolating from a multiple. We do not publish targets. What we can do is show the arithmetic, so you can supply your own assumptions and see where they land.

Three anchors exist below the current quote, and all three are stale for the same reason: they pre-date the FY26 result.

• The registered valuer put fair value at ₹62,035.54 as at 30 September 2025, after a 21.65% illiquidity discount.
• Strip that discount and the same exercise gives ₹79,181.
• Debentures actually converted during FY26 at about ₹62,500 a share, which is a real transaction rather than an opinion.

Above them sits the market at ₹3,75,000 to ₹4,00,001, which is 18.0 to 19.2 times FY26 earnings. Whether that is cheap or dear depends entirely on FY27, so here is the grid. Each cell is what the share is worth on 23,352 shares at a given profit and a given multiple.

FY27 profit after tax	At 15x	At 20x	At 25x
Flat at ₹48.6 crore	₹3,12,307	₹4,16,410	₹5,20,512
₹63.2 crore (+30%)	₹4,06,025	₹5,41,367	₹6,76,709
₹77.8 crore (+60%)	₹4,99,679	₹6,66,238	₹8,32,798
₹97.2 crore (+100%)	₹6,24,615	₹8,32,819	₹10,41,024

That table is arithmetic, not a forecast. It tells you nothing about which row is likely. It does tell you the shape of the bet: at today's upper quote you are already paying more than 15 times a profit that has to grow 28% to exist, and if FY27 profit merely holds flat, 15 times puts the share below ₹3,15,000.

Run it the other way and the question sharpens. At ₹4,00,001 the market capitalisation is ₹934 crore. For that to represent 15 times forward earnings, FY27 profit has to reach ₹62.3 crore, which is 28% growth. At the lower quote of ₹3,75,000 the same test needs ₹58.4 crore, or 20% growth. After a year of 199% profit growth, 20% to 28% does not sound demanding. The margin bridge is the reason to check that instinct rather than trust it.

One caveat that voids every number in the table. If the bonus issue this company needs before listing goes ahead, per-share figures are restated across a much larger share count and none of these levels mean anything in their current form. The market capitalisation columns survive. The rupee-per-share columns do not.

This capital structure cannot list as it stands

Something has to happen to the share count before any of this reaches an exchange, and I have not seen it discussed anywhere.

The company has 23,352 shares in issue against authorised capital of ₹10 lakh, or 1,00,000 shares. At a quote near ₹4 lakh, one share costs more than a small car. A main-board offer needs at least 2,000 allottees; there are not enough shares in existence to allot meaningfully to 2,000 people, let alone leave a float that trades.

A stock split alone will not fix it. Face value is already ₹10 and the regulatory floor is ₹1, so the deepest possible split is ten for one. That takes the equity base to 1,80,180 shares at roughly ₹40,000 each, still nowhere near a listable price.

Which leaves a bonus issue, and the company can fund one comfortably. Free reserves stand at ₹88.03 crore, of which ₹61.42 crore sits in the profit and loss surplus. Capitalising a portion of that into bonus shares, alongside a resolution lifting authorised capital well above the current ₹10 lakh, is the mechanical prerequisite for listing.

That changes nothing about what the company is worth. It matters for expectations. Buy one share at ₹4 lakh today and you should expect to hold a much larger number of much cheaper shares by the time this lists. Treat an authorised-capital increase or a bonus announcement as evidence the process is real, not as a windfall.

Why the valuation and the market price both make sense

Back to the two numbers we started with. The ₹62,035.54 was not wrong. It was answering a different question, at a different time, under rules that force a particular answer.

It was commissioned for compliance under the Companies Act, to price a debenture conversion, and the report says explicitly it should not be used for any other purpose. It valued the company as at 30 September 2025, six months before the FY26 books closed. It rejected market comparables because there were no listed peers, so it ran a DCF on management projections. And it applied an illiquidity discount of 21.65%, because on that date these shares had no route to a public market.

Reference point	Date	Per share
Registered valuer fair value, after illiquidity discount	30 Sep 2025	₹62,035.54
Same valuation before that discount	30 Sep 2025	₹79,181
Actual debenture conversion during FY26	FY26	~₹62,500
Dealer quotes, spread across four platforms	16 Sep 2026	₹3,75,000 – ₹4,00,001

Three things changed after that report was signed. The forecast underpinning it was overtaken by a factor of nearly two. The illiquidity discount became harder to justify as a listing process advanced. And the company added ₹51 crore to net worth in a single year.

None of that makes ₹4 lakh correct. It does mean the gap is mostly the distance between a conservative compliance exercise and a fast-moving business, not evidence of mania. Anyone quoting ₹62,035 as today's fair value is quoting a number built on revenue assumptions the company has already left behind.

The listing is real, and further along than most sources suggest

Several public pages still show this company as having no IPO process, and every platform we checked reports the DRHP as not filed. That second part is accurate. The first is not.

Note 16 of the audited accounts discloses ₹45 lakh of pre-IPO expenses capitalised in other current assets, and explains that the company has initiated listing of its equity shares, appointed a merchant banker, legal advisor and registrar to the issue, and is "actively coordinating with the appointed intermediaries to complete the required formalities withing next few months".

Two caveats. The note refers to the "NSE Platform", which could mean the SME platform or the main board, and those are materially different outcomes for valuation and liquidity. And capitalising issue expenses rather than charging them assumes the issue completes. At ₹45 lakh against ₹48.62 crore of profit the accounting is immaterial, but it tells you what management expects.

On mechanics, the shares carry ISIN INE0SE901014, are held through both NSDL and CDSL, and the registrar is Maashitla Securities Private Limited in New Delhi. Platform records indicate no right of first refusal applies, which matters more than it sounds: where one exists, existing shareholders can match a sale and stall a transfer for weeks. One platform nonetheless tags its offering as physical rather than dematerialised, which for a company holding an active ISIN is the sort of inconsistency worth resolving with your broker before you wire money.

One loose end. The audited accounts signed in August 2026 give a registered office in Farrukhabad. Platform records and the company's own materials point to Lucknow. A shift of registered office to a larger city is an ordinary pre-listing step, but I cannot confirm from the documents that it has happened.

What I would want answered before the price moves again

The strengths are genuine and I do not want to bury them. This is a profitable, debt-free, cash-generating business with a 59% operating margin in a category most Indian investors cannot otherwise access. The pivot from vernacular test prep to AI training data was executed in under two years and the numbers went up, not down. Revenue per employee of ₹1.61 crore on a headcount of 67 is remarkable if it holds.

That last figure is also a question. A data business generating ₹1.6 crore per employee is either unusually well-automated or leaning heavily on contracted labour that sits inside the content cost line rather than the salary line. Both are viable. They carry different risks, and the accounts do not tell you which it is. Given what we now know about the content line, I suspect the second.

Three other things I would want in a prospectus.

• Receivables. They grew 276% while revenue grew 123%, pushing collection from about 25 days to 41. The ageing shows ₹1.56 crore outstanding beyond a year, 12.7% of the book, with no provision against it. Management states it sees no material credit loss. On export billings to overseas counterparties, I would like that tested rather than asserted.
• Related-party reimbursements. Director remuneration for the founder was ₹38.48 lakh. Reimbursement of expenses to the same director was ₹1.55 crore, four times his salary and double the prior year. That may be entirely ordinary pass-through cost. It is large enough to deserve a breakdown.
• Data provenance. A corpus containing millions of patient medical records and ISBN-attributed textbooks raises consent, licensing and privacy questions across several jurisdictions at once. This is the kind of exposure that does not appear anywhere in a profit and loss account until it does.

I will also note, without making too much of it, that the valuation report contains two clerical slips: a reference to a different company's initials in the conversion paragraph, and a section dating the exercise to 31 March 2025 while the rest of the document says 30 September 2025. Neither changes the arithmetic. Both are reminders that a compliance valuation is a compliance valuation.

The bottom line

The market is not paying ₹4 lakh for the company the valuer described in September 2025. It is paying for the company the FY26 accounts describe, and on those accounts the price works out to roughly 18 to 19 times earnings for a business that tripled its profit and holds no debt.

What you are underwriting is whether FY26 was a step change or a spike. The margin bridge is where I would start, because it shows the last two years of expansion were bought with legacy costs that have now run out. Export services at 96% of revenue and undisclosed customer concentration are where the rest of the answer sits. A DRHP would settle most of it.

And check your denominator before anything else. On the numbers circulating this week, the difference between diligence and a screenshot is a factor of ten.

Browse other unlisted and pre-IPO companies we track, follow the pipeline on our IPO news page, or read more market insights.

Sources: All financial figures are from the audited financial statements of InfoBay AI Limited (CIN U74999UP2016PLC088614, PAN AAECE4977H, ISIN INE0SE901014) for the year ended 31 March 2026, audited by Arora Prem & Associates, Chartered Accountants, and signed 19 August 2026, including the balance sheet, statement of profit and loss, cash flow statement, and the notes on share capital, shareholding, revenue, other expenses, related-party transactions, segment reporting, trade receivable ageing and financial ratios. FY24 comparatives are as reported in secondary databases and reconcile to the FY25 opening balances in the audited accounts. The ₹62,035.54 fair value, the discounted cash flow projections, the 21.65% illiquidity discount and the cost-of-capital build-up are from the share valuation report of EduGorilla Community Private Limited prepared by an IBBI-registered valuer, valuation date 30 September 2025, report dated 30 October 2025, UDIN 25437466BMIBMK8552. Corpus scale figures, market-size estimates and funding claims are as published by the company and by unlisted-share dealing platforms, and are reported here as published, including where they conflict with the audited accounts. Indicative prices, lot sizes, share counts, depository availability, registrar details and price history are likewise as displayed on those platforms; such quotes are indicative over-the-counter reference prices, vary between dealers, are not exchange-regulated, and do not represent an offer to deal. Amounts reported in the statutory accounts in lakhs have been converted to crore.

This is not investment advice. Consult a SEBI-registered investment adviser before making any investment decision.

Related reading on The Finance Network: the NSE IPO and GFCL EV Products. Browse every company we price on the pre-IPO companies page, or see where our prices come from.