Sep 15, 2025 · 1h 0m · 20vc
Mercor CEO & Co-Founder, Brendan Foody: How They Grew from $1M to $500M in 17 Months · 20VC with Harry Stebbings
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
This video features an in-depth interview with Brendan Foody, the 22-year-old co-founder and CEO of Mercor, exploring his journey from high school side hustles to achieving a historic $500 million revenue run rate in just 17 months. He shares key insights into the transition toward reinforcement learning environments, the strategic importance of high-quality human data, and his philosophy on corporate culture, capital efficiency, and the future of AI.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 26.8% of the talking time here. How this is scored →
Harry as informed peer 4.4 Guest teaching 4.0 Guest disagreement 2.2 Harry pushing back 3.4
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 14:13 Direct rejection of competitor claim on quality algorithms
Brendan flatly rejects Harry's quotation of competitor Edwin with 'That's not true at all', defending Mercor's proprietary evaluation models.
Hardest push from Harry ▶ 27:02 Harry directly challenges early valuation multiplesHarry pushes back on Mercor's $2B valuation relative to its early revenue run rate, bluntly calling it a 'fucking punchy price.'
Biggest teaching moment ▶ 20:43 Contrasting Mercor's expert pay model with incumbent crowdsourcingBrendan educates Harry on the structural shift in AI training data, contrasting Mercor's $95/hr expert compensation model against Scale AI's $30/hr crowdsourcing rates.
Harry holds his own ▶ 52:18 Harry offers seasoned VC strategy on predatory pricingHarry asserts his expertise as a venture capitalist, detailing how Brendan could aggressively leverage cash reserves to undercut and strangle competitors.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Early Entrepreneurship & Safe Donut Bets | 1 | 1 | 1 | 1 | Harry opens with light biographical banter based on intel from a mutual contact, asking about Brendan's early donut sales and his mother's reaction. Brendan lightheartedly shares his eighth-grade donut arbitrage business and middle school principal encounters. | |
| Ambition, College, and the Sneaker Hustle | 3 | 2 | 1 | 2 | Harry probes the founder mindset, asking if Brendan possesses a classic duality of a superiority complex paired with deep insecurity. Brendan responds humbly, explaining his early ambitions, high school sneaker reselling agency, and reluctance to attend college. | |
| The Value of Modern Education | 4 | 5 | 4 | 5 | Harry brings a sharp quote from competitor Edwin claiming everyone in the AI human data space is just a 'body shop.' Brendan explicitly rejects this premise, educating Harry on Mercor's role as a deep research partner sourcing high-skilled talent rather than low-skilled crowdsourcing. | |
| Human Data in the Era of AI Scaling | 5 | 5 | 3 | 5 | Harry pushes on supply-side bottlenecks and cites the Cohere founder questioning scaling laws. Brendan explains why RL environments with multi-tool integration expand human expert demand rather than diminish it. | |
| Market Differentiation & Proprietary Matching | 4 | 5 | 5 | 4 | Harry confronts Brendan with another competitor quote claiming no companies have algorithms to measure data efficiency. Brendan directly rejects the claim ('That's not true at all') and details Mercor's power-law expert matching systems. | |
| Revenue Concentration & Consolidating Spend | 5 | 4 | 3 | 4 | Harry brings insider info from a competitor's board member claiming labs multi-vendor to prevent vendor lock-in. Brendan reframes the market dynamics, arguing labs prioritize model performance over multi-vendoring and naturally consolidate spending over time. | |
| Customer Concentration & The Nvidia Analog | 5 | 4 | 3 | 5 | Harry presses on revenue concentration risks, drawing parallels to Nvidia's customer concentration. Brendan embraces the Nvidia comparison as validation and reveals Mercor's revenue run rate exploded from $1M to $500M in 17 months. | |
| Scale AI's Quality Decline | 4 | 5 | 2 | 3 | Harry bluntly mentions industry consensus that Scale AI suffered from poor quality. Brendan tactfully agrees on Scale's product/quality slump, contrasting Mercor's $95/hr expert compensation against Scale's $30/hr rate. | |
| Synthetic Data & Human Stasis Points | 4 | 5 | 2 | 3 | Harry asks if synthetic data will eliminate the need for human-annotated data over a 10-year horizon. Brendan introduces the 'human stasis point' concept, arguing human feedback remains indispensable for evaluating non-trivial real-world workflows. | |
| The Inefficiency of Academic Evaluations | 4 | 4 | 2 | 3 | Harry provocative asks if current AI evaluation benchmarks are 'bullshit.' Brendan agrees, highlighting the disconnect between academic benchmarks like Olympiad math and actual high-value economic workflows. | |
| Valuation Multiples & Capital Efficiency | 6 | 5 | 2 | 5 | Harry brings exact financial metrics to the table, calling Mercor's early valuation multiples 'fucking punchy.' Brendan walks through the rapid ARR expansion from $1.5M to $500M, proving early valuations were actually cheap on a forward basis. | |
| Signaling Through Private Financing | 5 | 3 | 2 | 4 | Harry questions why Mercor would need to raise capital given $500M ARR and strong profitability. Brendan explains the strategic value of funding as a market signaling mechanism rather than a cash necessity. | |
| Long-Term Orientation & Staying Private | 5 | 5 | 2 | 3 | Harry asks if an IPO is on the horizon given public vs private pricing dynamics and references a recent MIT study on AI ROI failure rates. Brendan quotes Jack Dorsey's advice on remaining private for long-term focus and critiques 'vibe spending' without clear PRDs. | |
| Sustainability of AI Revenues & Margins | 5 | 4 | 1 | 3 | Harry asks about application-layer revenue sustainability and weak gross margin structures caused by subsidies. Brendan shares venture heuristics on evaluating true user retention vs failing pilot programs. | |
| Capex and Overhyped Segments | 5 | 4 | 2 | 3 | Harry brings up Wall Street capex concerns regarding infrastructure investments. Brendan offers a bullish 10-year view and explains the elasticity of engineering demand driving tool usage like Cursor and Cognition. | |
| Model Creators & AI Talent Economics | 4 | 4 | 3 | 4 | Harry challenges Brendan on AI talent compensation, teasing that Meta's $100M compensation packages drain startup talent. Brendan counters that mission-driven purpose and uncapped startup equity upside retain top missionaries over mercenaries. | |
| Underappreciated Models & Generalization Mindset | 4 | 4 | 1 | 2 | Harry asks which model providers are underappreciated and whether specialized or general models win. Brendan highlights Google's Gemini Flash and admits changing his mind toward monolithic models after seeing O3's reasoning. | |
| Sovereignty & Work Culture Intensity | 5 | 4 | 3 | 3 | Harry probes sovereign AI models and brings up intense 996 work culture. Brendan clarifies that 996 was an informal descriptor for early team dedication rather than a mandatory corporate policy. | |
| Executive Hiring and Language Neutrality | 6 | 3 | 2 | 4 | Harry observes how startup language softens during executive scaling and asks what Brendan would do if he weren't scared. When Brendan ponders burning $100M, Harry steps in with aggressive VC strategy advice on using cash reserves as a weapon to strangle competitors. | |
| Peter Thiel's Influence & Capacity Constraints | 4 | 4 | 1 | 3 | Harry asks about board member Peter Thiel's view on capital discipline. Brendan shares Thiel's focus on fundamentals and discloses that Mercor turns away customer projects daily due to supply capacity constraints. | |
| Quick Fire Round: AI Super Intelligence and Evals | 4 | 4 | 2 | 2 | Harry conducts a quick-fire round covering AI misconceptions, OpenAI strategy, Sam Altman quotes, and dream investors. Brendan dismisses 3-year AGI predictions and outlines his vision for RL environments subsuming monotonous work. |
Statements from this episode (47)
Brendan Foody made hundreds of thousands of dollars in high school
“I made hundreds of thousands of dollars when I was in high school.”
Mercor hit a nine-figure run rate and quadrupled post-Scale deal
“We were already at a nine-figure revenue run rate, and the company quadrupled since the scale acquisition.”
Mercor scaled from $1M to $500M run rate in 17 months
“We scaled the business from one to five hundred million in the last 17 months, which is the fastest revenue growth of all time. One month faster than Cursor's time.”
Mercor pays $95 per hour compared to $30 at Scale and Surge
“Our average marketplace pay rate is 95 dollars an hour to put that in frame of reference, scale and search generally pay about 30 dollars an hour.”
Mercor has enough demand to double overnight if capacity allows
“We have the demand to, like, double overnight if we can meet capacity.”
Reinforcement learning environments will subsume the entire economy
“RL environments will subsume the entire economy, because it doesn't make sense that humans would be doing monotonous, redundant work.”
College is no longer educationally valuable due to free online information
“I think that so much of the reason that college is no longer valuable from an educational standpoint is that that information is all available online.”
Scale AI used Mercor's platform to hire thousands of workers
“Scale AI came to us and they used our platform to hire thousands of people.”
Human AI data TAM is bounded by human-model capability gaps
“The total addressable market is limited by the amount of things that humans are better at than models.”
Human data verifiers remain essential for expanding AI model capabilities
“So long as there's things that the human is able to do, the model's not able to do, and we want those capabilities in the model, whether it's to schedule a meeting or write emails for you or whatever it is, we need humans that help to create those verifiers an…”
AI model capabilities are not plateauing despite recent industry concerns
“I don't think that Models are plateauing. Like, if we look at the last 12 months of progress in models, I've been blown away.”
AI data value follows a power law driven by top contributors
“The outcomes of data and the people that contribute to it are extremely power lot, similar to a company where if you have a hundred people on a project, oftentimes majority of the model improvement is coming from the top 10 to 20% of people, right? Just like s…”
Mercor works with all top frontier AI research labs
“We work with All of the top research labs at the frontier of model capabilities, and we're not like the crowdsourcing companies in that we try to hide all the people on the platform, pay them low rates, et cetera.”
AI labs eventually consolidate almost all human data spend to Mercor
“We've definitely found those stories of customers where I think they start out multi-vendoring, working with a bunch of different customers, with different vendors, but ultimately get to the point where they realize that they're going to be making a trade-off …”
AI data markets naturally consolidate around early leaders via scale economies
“If you look at a lot of the analogs and markets, they often start very fragmented with many different players, but consolidate over time. And so much of that reason for consolidation is that there's structural advantages and economies of scale to being the fir…”
Mercor's customer concentration breakdown is similar to Nvidia's
“Our largest customer, I can't share the exact percentage, but the breakdown is relatively similar to Nvidia.”
Mercor is growing faster at $500M run rate than ever before
“In fact, the company is growing faster now at 500 than it's ever grown before, and so the growth continues accelerating.”
Scale AI lost its focus on product and quality
“But in some ways, scale lost the focus on product on scaling quality. And that was one of the largest challenges of the business.”
Treating marketplace talent well is critical for frontier AI development
“But actually if I had to choose the most important thing, it would be the internal link to quality, which is that having phenomenal people that you treat incredibly well is the most important thing in this market and getting those people to refer all of their …”
AI models will still require human training data in ten years
“I very much believe so, and the reason is that the question comes down to when we'll have super intelligence. Once we have super intelligence and models are better than humans at everything, then, of course, that means that humans won't be able to contribute t…”
AI beats PhDs at reasoning but fails at basic workflow tasks
“Like, these models have gold medals in Olympiad math, and they're better than the best PhD at reasoning, but they can't Draft an email for me. They can't schedule a meeting. They can't do so many of the basic things of just using a handful of tools to do a tas…”
Academic AI benchmarks are wholly disconnected from enterprise needs
“One of the largest inefficiencies in all of AI research is that the evals that people have been going on Of humanity's last exam and PhD level reasoning or Olympiad math are wholly disconnected from the outcomes that consumers and enterprises actually care abo…”
Benchmark offered Mercor a $250M valuation at a $2M run rate
“When we met Victor, we were at. 1.5 million dollars in revenue run rate. He gave us the term sheet and we were at a little over two million dollars in revenue run rate. So over a hundred X multiple on revenue.”
Mercor raised its Series B at a 100x revenue multiple
“When Felicis gave us the term sheet, we were at twenty million in revenue run rate. So it's a hundred X multiple on the revenue”
Mercor is highly profitable and does not need additional external financing
“And so now we're 25 times larger in revenue scale than we were at the Series B but in a spot where the business is so profitable that we don't really need to go out for financing or spend too much time thinking about financing while we often get a lot of offer…”
Mercor averaged 54% month-over-month revenue growth around its Series B
“We, we're growing over 50% month over month. Like we averaged 54% month over month growth for a while at that time period.”
Mercor will likely complete a low-dilution financing round soon
“I think that it's likely we'll do a financing soon to answer your question with low dilution, largely because there's a lot of benefits to signaling ourselves as the market leader and RL environments and all of the high complexity data that we produce.”
Jack Dorsey invested in Mercor and advised keeping the company private
“Like I remember when I was talking with Jack Dorsey before he invested, one piece of advice he gave me was that we should stay private as long as possible.”
Extraordinary AI startups built today will look cheap in ten years
“It wouldn't shock me if, like, we feel like things are frothy and it's a crazy time, but if we're evaluating things on a ten-year time horizon all of these extraordinary businesses that are being built will look like a discount”
If an AI model is the product, its eval is the PRD
“And if we think about the model as the product, then the eval is the PRD. And so many people have been sort of just like, you know, vibe spending on AI without actually writing the PRD of what do they want to implement and how do they measure that it's going t…”
Mercor has positive growth in net margins unlike most AI startups
“Like, we have very positive growth in net margins, unlike most AI companies.”
Subsidizing AI product margins fails in markets with low switching costs
“But I think the case where I would be hesitant is when there is very competitive markets with low switching costs so that people are pumping hundreds of millions in subsidies maybe billions in subsidies and then all of a sudden they switch over, the customers …”
AI coding and foundation model hype is matched by real value
“Nothing jumps out to me on that, because obviously, like, I think the things with the most hype are code and foundation models and maybe starting to be use cases in finance, and I feel like The value being created is also very real.”
Mercor engineers rely heavily on Cursor, Claude Code, and Cognition
“Like, the amount of utility that our engineers get from Cursor and Cloud Code and Cognition is incredible, and same thing for-”
Tech companies will employ more software engineers in five years, not fewer
“I think more. And the reason is that engineering is such an elastic role, right? Where if we could build a hundred times more software or say we make engineers 10 times more efficient, We would probably build a hundred times more software, right? And so far as…”
The dominant AI foundation model creators have likely already been established
“I think the largest model creators have already been yeah, already exist, but I'm not a hundred percent sure about that. Like, I definitely have, you know, some error bars about it. My expectation for why The largest model builders exist is just obviously the …”
Google DeepMind's Gemini Flash models are extraordinary and underappreciated
“I definitely feel like a lot of the Gemini Flash models are also extraordinary and underappreciated on, on evals especially their small models I'm always amazed with. So if I had to choose, maybe not a company, but especially Sadov, Models. I think the DeepMin…”
Enterprise model customization will be a major investment area this decade
“I still think that we're just in the first inning of model customization of every enterprise wanting models to know how to use their own set of tools of knowing how to use all of their own like knowledge bases, et cetera, and the processes that they've codifie…”
Sovereign AI models will not become the largest general-purpose players
“Maybe wins in a scoped part of the market. Like, I could see why, for example there would be a lot of benefits to having Mistral be an expert in European law that might have nuances from other kinds of law, and they've just invested far more in having the best…”
Mercor never mandated 996 work hours for its employees
“Well, not exactly. I need to offer a key clarification, which is that we've actually never mandated hours. It was more so when we were talking about nine nine six, it was a description of how the early team worked. Like, in fact, the reason we talked about nin…”
Working hours and employee obsession decouple as startups scale
“I think the truth is that all along it's been much more about hiring people that like give a shit and love what they do and are obsessed with it and the way that we are rather than specific hours. And early on those were highly correlated, but I think that as …”
The current AI market cycle resembles 1996 or 1997, not a bubble
“And don't get me wrong, I'm still, like, incredibly bullish on the market and AI. I think we're much more at, like, 96 or 97”
Believing AI superintelligence will arrive in three years is totally wrong
“That we'll have super intelligence in three years. That's better than humans at everything. I think it's totally wrong.”
Selling basic AI APIs is a bad business with low switching costs
“I think model customization is a really exciting opportunity because API will have low switching costs, not much pricing power, it's not a good business and focusing more on model customization is a really exciting opportunity.”
AI startup durability depends on whether better models improve or ruin them
“I really like the thing Sam Altman says of, will models being dramatically better in one to two years improve your business or worsen it? I think that that is, in so many ways, the most important question to see if you're building a business that's durable and…”
Shorts cut from this episode
▶ $100K+ in High School 💵 · 20VC with Harry Stebbings (@0:00)
▶ The Most Underrated AI Model · 20VC with Harry Stebbings (@44:41)
▶ Mercor Founder’s Early Donut Side-Hustle · 20VC with Harry S (@1:33)
▶ Why We’ll Have More Engineers in 5 Years · 20VC with Harry S (@40:35)
▶ The Power of Big Customers · 20VC with Harry Stebbings (@17:39)
▶ From $1M to $500M in 17 Months 📈 · 20VC with Harry Stebbing (@0:00)