InfoBay AI: what investors need to know about this unlisted pivot story

InfoBay AI Limited is an unlisted company circulating in India's grey-market pre-IPO channel under a compelling pitch: a former ed-tech test-prep business that repurposed its content library into training data for frontier AI labs. Story is coherent on paper, but the numbers backing it up are a much messier picture -- and the gap between the two is the whole story here.

What does InfoBay AI do?

InfoBay AI was incorporated in December 2016 as EduGorilla, a Lucknow-based ed-tech startup offering multilingual online test series and study material for competitive exam prep, along with a white-label SaaS platform called Gibbon that let educators and institutions run their own learning portals. Built a real business on this model over several years, serving large base of exam-takers, educators, and institutional partners across India and internationally.

Starting around 2025, company repositioned itself as InfoBay.AI, describing itself as a training-data and AI-infrastructure provider: expert-verified datasets, annotation infrastructure, evaluation systems for large language models. Company's own framing is explicit about the connection -- it describes this as "asset repurposing," turning the structured STEM and reasoning content built for competitive-exam test prep into training data for AI models. Markets a proprietary "TIBET" data engine, says repository spans STEM, healthcare records, video, and audio content across dozens of vernacular languages.

That lineage matters: this isn't an AI-native startup built from scratch for this purpose. It's an existing ed-tech content business relabeled around a hotter theme. Underlying question -- how much of the original asset base genuinely translates into commercially valuable AI training data, versus how much of this is narrative repositioning.

How has the company performed?

This is where story gets genuinely hard to pin down. Two different sets of numbers circulate for the same company, and they don't reconcile.

• MCA-filed data (most reliable source, since it's what company actually files with regulators): entity did roughly ₹10.6 crore in revenue in FY23, up from about ₹5 crore the year before. Real but modest ed-tech-scale numbers, alongside a loss of over ₹76 lakh that year.
• PR-wire press release (explicitly labeled an advertorial, not independent reporting): claims company crossed ₹100 crore in revenue in FY 2025-26, quoting the founder directly. Also claims serving "frontier AI labs" globally -- a claim with no named clients attached anywhere, including on the company's own website, which describes dataset categories but doesn't name a single enterprise customer.

Separately, broker/distributor marketing materials aimed at unlisted-share buyers have circulated a much steeper revenue curve for the same stretch of years -- figures running meaningfully higher than what shows up in the MCA-sourced trackers. That gap between filing-grade numbers and sales-collateral numbers is worth taking seriously: promotional figures are what's shaping how retail buyers perceive this stock, and they don't match the more conservative, verifiable trend line.

On capital structure: Company's authorized share capital is ₹10 lakh, paid-up capital just ₹2.34 lakh -- implying roughly 24,350 shares outstanding. Against that share count, grey-market platforms have quoted the stock trading around ₹3 lakh per share as of mid-2026, implying a market capitalization near ₹700-₹750 crore. With float this thin, a handful of trades can set the "price." This is a case where the quoted price is closer to declared than discovered.

Who are the competitors of InfoBay AI?

InfoBay AI positions itself in the AI training-data and annotation space. This market has several real, well-capitalized players. Here's how it stacks against each:

• Scale AI: The category's former default vendor, now 49% owned by Meta. This ownership has cost it several frontier-lab customers -- Google, OpenAI, Microsoft, xAI have all reportedly reduced or ended contracts, citing conflict-of-interest concerns. InfoBay AI has no such ownership entanglement, but also none of Scale's decade of build-out or brand recognition. The opening Scale left is being filled by better-capitalized independents, not InfoBay.
• Surge AI: Bootstrapped, profitable, reportedly doing well over $1 billion in annual revenue. Works directly and named-ly with OpenAI, Google, Anthropic, and Microsoft. Scale gap shows starkest here -- Surge runs on roughly 130 full-time employees, InfoBay on a reported 18. InfoBay's own marketing claims a footprint ("powering half of the world's frontier labs") in the same league as Surge's actual named client roster, with no equivalent named clients to back it up.
• Mercor and Turing: Newer, fast-growing independents, absorbed business from labs leaving Scale, backed by significant institutional funding. Both raised capital at valuations far beyond InfoBay's roughly $6 million lifetime funding. This gives them balance sheet to compete for large enterprise contracts that InfoBay's capital base likely can't match.
• Deccan AI: Most directly comparable competitor -- also India-based, sources expert contributors domestically. Raised $25 million Series A led by A91 Partners, with participation from Susquehanna International Group and Prosus Ventures. Real institutional capital at a credible valuation, founder speaking on record about the business. InfoBay AI's most recent funding ($2.1 million, per press reports) is a fraction of that, and hasn't attracted comparable-tier investors.

Taken together, this comparison set is unflattering to any small, thinly capitalized new entrant. Every name above has raised significant institutional capital at credible valuations, employs a larger team, and in Surge AI's case works directly and named-ly with frontier labs. InfoBay AI, with around 18 people and roughly $6 million lifetime funding, shows a real scale mismatch against the "powering half of the world's frontier labs" language in its own press materials. No independent source could confirm any named frontier-lab client relationship for InfoBay AI specifically.

Is InfoBay AI a good investment opportunity?

Green flags, taken seriously on their own terms:

• The underlying insight isn't nonsense. STEM and competitive-exam question banks, built over years for a completely different purpose, plausibly do have secondary value as structured reasoning data for AI training. This is a real trend in the industry, and being an early mover on repurposing this kind of asset is a legitimate strategy.
• Auxano Capital, an institutional VC, exited its EduGorilla/InfoBay position at a reported 5.5x return in September 2025, a real financial outcome for a real investor, not just a claim.
• The company has converted from private to public limited status, which can sometimes signal pre-IPO preparation, and it has reported a fresh $2.1 million funding round aimed at R&D and team expansion.

Red flags that are hard to look past:

• The financial picture doesn't hold together across sources. Filing-grade revenue (₹11 to ₹50 crore, depending on year) is a full order of magnitude below what's circulating in broker marketing materials and the company's own PR-wire claims (₹100 crore-plus, with implied ₹50 crore PAT). A real business built on this insight would typically show headcount scaling with claimed revenue and named pilot clients disclosed ahead of any listing process, not a self-published press release and a hockey-stick brochure chart that doesn't match the filed numbers.
• No named clients exist anywhere in public materials, despite bold claims about serving frontier AI labs. That's a meaningful gap for a company whose entire investment case rests on B2B relationships with the world's most scrutinized AI companies.
• The paid-up capital versus market pricing disconnect is severe: roughly 24,350 shares outstanding, trading at a price that implies a market cap approaching ₹800 crore, is a textbook thin-float setup where price reflects what a small number of counterparties want it to say, not broad market consensus.
• Governance is closely held: the board is essentially built around one family (Rohit Manglik, Pushpa Manglik) plus co-founders, with no visible independent institutional director on record, despite the public-limited conversion.
• The timing pattern is itself worth noting: this is a company riding its second narrative wave on the same underlying entity (first vernacular ed-tech, now AI infrastructure), with each pivot roughly coinciding with that theme's peak retail investor interest. That pattern, observed across two unrelated hype cycles on one shell, is a more specific tell than any single number in isolation.

Bottom line: the AI-training-data theme this company is riding is real, and there's a legitimate version of this story where a company with real content assets successfully repositions into a valuable AI infrastructure business. But the evidence currently available doesn't support underwriting InfoBay AI specifically as that company: the claimed footprint (frontier-lab clients, ₹100 crore-plus revenue, near-50% margins) is well ahead of anything independently verifiable, and the extreme thin float means the quoted grey-market price carries very little information about fair value. This reads more as a stock story for the unlisted retail/HNI channel than a company-conviction bet. Anyone specifically wanting exposure to the AI-training-data theme has better-evidenced alternatives with disclosed clients, institutional governance, and verifiable financials.