Written by TFN Research Desk | covering startups, technology, venture capital, and business strategy.
While the AI industry focused on which company would build the smartest model, Brendan Foody was building the company that would own the humans those models needed to become smarter.
Brendan Foody is 22 years old. His company is worth $10 billion (Forbes, October 2025). He would have graduated college a few months before that number was announced. Instead, he is the CEO of one of the fastest-growing AI companies in history, the world’s youngest self-made billionaire, and the founder of a business that has become critical infrastructure for OpenAI, Meta, and virtually every major AI lab building smarter models. His co-founders, Adarsh Hiremath and Surya Midha, are the same age. Together they grew from zero to $10 billion in approximately 24 months by identifying the human bottleneck in a race everyone assumed was purely technical.
Topic tags: Founder Story • AI • Startup Strategy • Human Data • Youngest Billionaire
The 22-year-old who saw what the AI labs missed
Most people watching the AI race in 2023 were asking which company would build the most powerful model. Foody, Hiremath, and Midha were asking a different question: who would supply the human expertise those models needed to become intelligent?
The answer turned out to be Mercor, a company that started as a developer recruiting platform and pivoted into the critical infrastructure layer between AI labs and the domain expert networks they need. The pivot happened fast because the founders recognised what they had built accidentally before anyone else noticed it had value.
By February 2025, the company had $75 million in annualised recurring revenue (TechCrunch, February 2025). By September 2025, it was approaching $450 million (TechCrunch, September 2025). In October 2025, it raised $350 million in a Series C at a $10 billion valuation (TechCrunch, October 2025).
That is one of the fastest revenue ramps in startup history, built by three people who had not yet completed their undergraduate degrees.
Why this story matters
Mercor is important for Indian founders for a reason that goes beyond the headline valuation.
India has one of the largest pools of trained domain professionals in the world: doctors, engineers, lawyers, scientists, and researchers whose expertise is precisely what AI labs are willing to pay significant amounts to access. The market Mercor built in the United States by connecting verified domain experts to AI training workflows is waiting to be built for Indian talent.
Beyond the specific opportunity, the Mercor story proves two things that Indian founders should take seriously. First, the most valuable companies in the AI era may not be the ones building models. They may be the ones owning the layers the models need to become better. Second, the age premium in entrepreneurship is eroding faster than anyone expected. Antler’s research found that the average age of AI unicorn founders fell from 40 in 2021 to 29 in 2024 (CNBC, January 2026). Mercor’s founders are 22, seven years below that average, and the trend is still moving.
Quick facts
| Metric | Value | Source |
|---|---|---|
| Founders | Brendan Foody (CEO), Adarsh Hiremath (CTO), Surya Midha (COO) | Mercor |
| Founded | 2023 | Mercor |
| Headquarters | San Francisco, California | Mercor |
| Industry | AI recruiting and human data operations | Mercor |
| Valuation | $10 billion (October 2025) | Forbes via CNBC, October 2025 |
| ARR | Approaching $450 million (late 2025) | TechCrunch, September 2025 |
| Funding raised | $350 million Series C led by Felicis | TechCrunch, October 2025 |
| Domain expert network | 30,000+ professionals | Mercor |
| Daily earnings through platform | Over $1.5 million | Mercor |
| Average founder age | 22 | Mercor |
Background
Foody, Hiremath, and Midha grew up as friends in the Bay Area and competed together on the Bellarmine College Preparatory speech and debate team. They dropped out of Harvard and Georgetown respectively, accepted Thiel Fellowships, and moved to San Francisco in 2023 to build software connecting freelance programmers in India with technology companies in the United States.
The founding idea was straightforward: use AI to interview candidates at scale, assess their skills, and match them with employers faster and more accurately than any human recruiter could. The execution worked. Within nine months, the platform had reached a $1 million revenue run rate.
The pivot that made Mercor a $10 billion company was not planned. It was discovered.
In building their talent matching platform, the founders had accidentally assembled something the AI industry needed but had not organised: a vast, verified network of highly specialised professionals across medicine, law, engineering, science, and research. The kind of people whose judgment AI labs needed to make their models genuinely intelligent, not just statistically fluent.
“Everyone’s been focused on what models can do,” Foody told Fortune. “But the real opportunity is teaching them what only humans know: judgment, nuance, and taste.” (Fortune, November 2025)
The founders recognised the pivot before the demand had fully formed, and moved to build supply-side infrastructure before AI labs had articulated the need precisely. When demand arrived, Mercor was already there.
Timeline
| Year | Milestone |
|---|---|
| 2023 | Foody, Hiremath, and Midha found Mercor in San Francisco; begin with developer recruiting platform connecting Indian programmers to US tech companies |
| Early 2024 | Platform reaches $1 million revenue run rate within nine months of founding |
| Mid 2024 | Founders identify that their expert network is more valuable as AI training infrastructure than as a recruiting platform; begin pivoting |
| February 2025 | ARR reaches $75 million; Mercor raises at $2 billion valuation (TechCrunch, February 2025) |
| March 2025 | CEO Brendan Foody publicly announces ARR has crossed $100 million |
| September 2025 | ARR approaches $450 million (TechCrunch, September 2025) |
| October 2025 | Mercor raises $350 million Series C led by Felicis at $10 billion valuation (TechCrunch, October 2025) |
| October 2025 | All three founders become the world’s youngest self-made billionaires; Scale AI files suit alleging trade secret misappropriation (Fortune, November 2025) |
How it happened
Move 1: Accidentally building the right network before the demand arrived
The most consequential thing Mercor did was not strategic. It was accidental, followed by a fast and correct response to what the accident revealed.
In building an AI-powered recruiting platform for programmers, the founders had to solve a real technical problem: how do you verify that a candidate actually has the skills they claim? Their answer was AI-assisted interviews and skills assessments administered at scale. That capability worked for programmers. It also worked for doctors, lawyers, engineers, and scientists.
By the time the founders recognised that their platform had assembled a verified network of 30,000-plus domain experts, they also recognised that this asset was more valuable to AI labs than to corporate recruiting departments. AI labs were running into the same problem at scale that Mercor had solved in miniature: how do you evaluate whether a model is giving a medically sound answer about a complex diagnosis? You need doctors. Verified, available doctors who can assess model outputs at speed.
Mercor had the supply. The labs had the demand. The pivot was not a creative leap. It was a recognition of what the company had already built.
Move 2: Becoming infrastructure before anyone else understood the category
The category Mercor entered did not have a name when it entered it. “Human data operations for AI training” was not a job description, a VC thesis, or a conference track in 2023. It was a problem that every major AI lab was solving manually, expensively, and at insufficient scale.
Mercor entered that space before the labs had articulated the need clearly, which meant Mercor had to convince customers they had a problem whose shape they had not yet defined. That is a harder sell than addressing a recognised category, but it comes with a structural advantage: you build the network before competitors understand what network to build.
By the time Scale AI, the incumbent AI training data platform, identified Mercor as a competitive threat, Mercor had a 30,000-person verified expert network that could not be replicated quickly (Fortune, November 2025). The expertise verification, the relationship infrastructure, and the operational processes for deploying domain experts at AI-lab speed were all already built.
Scale AI subsequently filed a lawsuit against Mercor alleging misappropriation of trade secrets. Incumbents do not sue companies that are not winning.
Move 3: Moving without credentials or permission
The three Mercor founders had no corporate experience, no prior exits, no MBA, and no industry reputation when they started. They had a specific insight, a speech and debate team’s comfort with argumentation, and Thiel Fellowships that gave them financial runway to try.
What they did not do is wait for permission to enter the market. They did not wait for a VC to tell them the opportunity was real. They did not wait for the AI labs to publish an RFP for domain expert services. They built the supply side of the market before the demand side had fully formed, which is the only reliable way to own a new market.
When Hiremath reflected on the speed of the company’s growth, he said: “The thing that’s crazy for me is, if I weren’t working on Mercor, I would have just graduated college a couple months ago. My life did such a 180.” (Fortune, November 2025)
Business model breakdown
Mercor operates as a two-sided marketplace connecting AI labs with verified domain experts. On the supply side, the company has built a network of 30,000-plus professionals across medicine, law, engineering, science, and research. These experts are verified for credential accuracy and assessed for the quality of their domain judgment, not just their availability.
On the demand side, Mercor’s customers are AI labs and enterprise AI teams that need human expert evaluation to improve their models. These are not small contracts. A lab training a medical reasoning model needs hundreds of doctors to evaluate model outputs at consistent quality and speed. Mercor’s platform manages that workflow, the expert matching, the task assignment, the quality control, and the compensation.
The platform generated over $1.5 million in daily earnings for its expert network by late 2025 (Mercor), which indicates both the scale of the demand and the stickiness of the supply. Experts who earn meaningful income through Mercor have no incentive to defect to a competing platform. The supply-side moat compounds with scale, not against it.
Approaching $450 million in ARR by late 2025 (TechCrunch, September 2025), the revenue ramp implies that AI labs are treating Mercor as recurring infrastructure spend, not project-by-project procurement. That is the revenue quality that justifies a $10 billion valuation on a two-year-old company.
and a verified expert network.
Comparison table
| Mercor | Scale AI | Traditional recruiters | |
|---|---|---|---|
| Core offering | Domain expert matching for AI training | Large-scale data labelling and annotation | Candidate sourcing and placement |
| Expert verification | Deep credential and judgment verification | Task-based quality control | Resume and reference checks |
| Primary customers | AI labs needing domain reasoning evaluation | AI teams needing large-scale data annotation | Corporate HR departments |
| Expert network | 30,000+ domain specialists | Large contractor pool, generalist-weighted | Variable by firm |
| Differentiation | Specialised domain judgment at lab speed | Scale and breadth of annotation | Human relationship networks |
| Founded | 2023 | 2016 | Legacy industry |
What competitors missed
Scale AI, the incumbent in AI training data, built its business around large-scale data labelling and annotation. It was a strong and profitable business. But Mercor entered a more specialised position: not data labelling, but domain expertise evaluation. The difference is significant and structural.
A generalist labelling platform cannot evaluate whether a doctor’s reasoning about a complex oncology case is medically sound. It cannot assess whether a lawyer’s analysis of a contract clause reflects current case law. It cannot verify whether an engineer’s structural calculation is correct. Those evaluations require domain experts, not annotators. Mercor’s network can do what Scale AI’s cannot.
The competitive blind spot across the industry was the assumption that AI training was a data problem. Foody, Hiremath, and Midha identified it as a human capital problem. That framing difference, combined with their speed in building the supply-side network, produced a company that sits at the centre of the AI training economy while competitors are still largely working on the periphery.
Risks and challenges
- Synthetic data substitution. AI labs may eventually train models to evaluate their own outputs using synthetic data, reducing the need for human evaluators. The timeline for that shift is uncertain, but the risk is real and Mercor’s long-term defensibility depends on how slowly synthetic evaluation improves.
- Scale AI litigation. The trade secret lawsuit from Scale AI is a cost and distraction, though incumbents suing fast-growing competitors is a well-documented pattern and rarely the fatal threat it appears at filing.
- Expert network quality at scale. Maintaining the credential accuracy and judgment quality of a 30,000-plus person network is an operational challenge that becomes harder as the network grows. Quality dilution would erode Mercor’s core differentiation.
- Customer concentration. If a significant portion of Mercor’s ARR comes from a small number of major AI labs, losing one relationship would create meaningful revenue risk.
- AI lab consolidation. If the AI lab landscape consolidates to a smaller number of dominant players, Mercor’s customer base contracts, even if demand per customer remains high.
What founders can learn
- Look for the human bottleneck in every AI workflow. The most valuable companies of the next decade may own the layer that AI cannot replace by itself. Mercor found the bottleneck in AI training and built infrastructure around it before the category had a name.
- Accidentally building the right network is valid strategy, if you recognise the pivot before the moment passes. Mercor started as a recruiting platform and became AI training infrastructure when the data showed them what they had built. The founders had to see the difference and act on it fast.
- Move without credentials. Foody, Hiremath, and Midha had no corporate experience, no prior exits, and no MBA. They had a specific insight and the willingness to act on it before they had permission. That speed was itself the competitive advantage.
- Build supply-side infrastructure before demand is fully formed. Mercor built its expert network before AI labs had clearly articulated the need for it. Waiting for customers to specify the need means someone else has already built the supply.
- Do not confuse the initial product with the real business. Mercor’s initial product was developer recruiting. Its real business was domain expert matching for AI training. The founders had to make that distinction under time pressure and act on it decisively.
Expert analysis
Bull case. Mercor operates at the intersection of two exponentially growing forces: the demand for AI training data and the scarcity of verified domain expertise at scale. As AI models become more specialised and enter healthcare, law, finance, and scientific research, Mercor’s expert network becomes more valuable, not less. The $450 million ARR approaching by late 2025 (TechCrunch, September 2025) is the beginning of a much larger number, and the expert network moat compounds with every new customer relationship.
Bear case. Synthetic data and self-evaluating AI systems may reduce the need for human domain experts faster than Mercor’s current valuation assumes. If frontier models reach a capability threshold where they can reliably self-assess output quality, the recurring need for Mercor’s expert network shrinks. The timeline for that shift is uncertain, but betting on it being never is not a conservative position.
Contrarian view. The most important thing about Mercor may not be the valuation but the proof that the most defensible asset in the AI economy is a human network, not a technology stack. The technology Mercor uses to match and manage experts is replicable. The verified expert network and the operational relationships with major AI labs are not. Founders building in the AI era should ask which of their assets are truly network-based and which are merely software.
Future outlook
The demand for verified domain expertise in AI training is structurally tied to the increasing ambition of the models being built. As AI labs push models into specialised professional domains, including medical diagnosis, legal reasoning, and financial analysis, the need for expert evaluation does not decrease. It becomes more specific and more difficult to satisfy at scale.
Mercor is positioned to expand in two directions from its current position. The first is depth: more specialised expert networks for more narrow domains, increasing the quality and difficulty of the evaluation work it can supply. The second is geography: building expert networks outside the United States, where pools of credentialed professionals in India, Southeast Asia, and Latin America are both large and underutilised by AI training infrastructure.
The Indian market is particularly relevant. A Mercor-equivalent platform built around Indian domain professionals, doctors trained in AIIMS, lawyers qualified in the Bar Council of India, engineers from IITs, would address a genuine gap in AI training infrastructure while creating significant earning opportunities for Indian professionals. That opportunity is visible and not yet owned.
The bottom line
Mercor’s three founders did not build the world’s youngest self-made billionaire story by chasing valuations. They built it by finding the human bottleneck in the AI era and owning it before anyone else understood the category existed.
Key takeaways
- Mercor’s three 22-year-old founders became the world’s youngest self-made billionaires in October 2025 when their company reached a $10 billion valuation (Forbes, October 2025).
- The company pivoted from developer recruiting to AI training infrastructure by recognising its true asset: a verified network of 30,000-plus domain experts.
- ARR went from $75 million in February 2025 to approaching $450 million by late 2025, one of the fastest revenue ramps in startup history (TechCrunch, September 2025).
- The core insight: the bottleneck to smarter AI is not computing power but human domain expertise at scale.
- For Indian founders specifically, the market Mercor built in the US, connecting verified domain experts to AI training, is waiting to be built for the Indian talent pool.
Conclusion
The Mercor story is a clean proof of a principle that the AI industry keeps rediscovering: the infrastructure layer around a technology wave is often more valuable than the technology itself.
Every AI lab in the world is competing to build the most capable model. The models are what the press covers, what the conferences celebrate, and what the capital chases. Mercor found the unglamorous problem behind the glamorous race: someone has to teach these models what good judgment looks like, and that teaching requires verified human expertise that does not come from the internet.
Three 22-year-olds who had competed on a speech and debate team saw that before anyone else had named it. They built the supply-side infrastructure before the demand had fully formed. And when the demand arrived, Mercor was already there with 30,000 verified experts and the operational capability to deploy them at lab speed.
For Indian founders, the takeaway is not to copy Mercor’s model. It is to ask the same underlying question in every category they are building: what is the human bottleneck in this workflow, and who is going to own it?
TFN LENS
Mercor’s story is directly relevant to Indian founders in a way that most global AI success stories are not.
India has the professional talent pool that AI labs need: doctors, engineers, lawyers, scientists, and researchers whose domain expertise is precisely what these models require to become intelligent in specialised domains. What India does not yet have is a Mercor-equivalent platform that aggregates, verifies, and deploys that talent for AI training at scale.
That is a real market gap. It is also a market gap that Indian founders are better positioned than anyone to fill, because the supply side is already here. The question is which founding team will move first to build the infrastructure around it.
Building something of your own? Follow The Founder Nation and NamasteVC for curated startup funding news, grant alerts, and founder stories from India’s startup ecosystem, delivered straight to your feed, every week.
Frequently asked questions
Who founded Mercor?
Mercor was founded in 2023 by Brendan Foody (CEO), Adarsh Hiremath (CTO), and Surya Midha (COO). All three were 22 at the time of the company’s $10 billion valuation and had previously competed together on the Bellarmine College Preparatory debate team before attending Harvard and Georgetown, from which they dropped out to take Thiel Fellowships.
What does Mercor do?
Mercor operates a two-sided marketplace connecting AI labs and enterprise AI teams with verified domain experts, including doctors, lawyers, engineers, and scientists. These experts evaluate AI model outputs to help labs train more accurate, reliable models in specialised professional domains.
What is Mercor’s valuation?
Mercor reached a $10 billion valuation in October 2025 following a $350 million Series C round led by Felicis Ventures with participation from Benchmark, General Catalyst, and Robinhood Ventures (TechCrunch, October 2025).
How fast did Mercor grow?
Mercor’s ARR grew from $75 million in February 2025 to approaching $450 million by late 2025 (TechCrunch, September 2025). That pace makes it one of the fastest revenue ramps in startup history.
Is Mercor competing with Scale AI?
Yes. Scale AI filed a lawsuit against Mercor in late 2025 alleging misappropriation of trade secrets involving a former Scale employee. Mercor competes in a more specialised segment of the AI training market, focused on verified domain expert evaluation rather than large-scale generalist data annotation.
Why is Mercor relevant for Indian founders?
India has one of the world’s largest pools of credentialed domain professionals. Doctors, engineers, lawyers, and scientists whose domain expertise is exactly what AI labs pay Mercor to provide. Building a Mercor-equivalent platform for Indian domain expertise would address a real market gap while creating significant economic opportunity for Indian professionals whose knowledge has historically been undermonetised in the global AI training economy.
©️ The Founder Nation | All rights reserved | Written by TFN Research Desk | Word count: ~3,627 | Read time: ~19 minutes |
