Overview of 20VC: Mercor CEO Brendan Foody with Harry Stebbings
In this episode of 20VC, Harry Stebbings interviews Brendan Foody, co-founder and co-CEO of Mercor, about the company’s explosive growth, the implications of AI on software defensibility, the rising cost of talent and compute, and how agents are reshaping enterprise workflows. Foody argues that in AI, “the model is the product”, application-layer moats are thinning, and the biggest winners will likely be the model labs, infrastructure providers, and vertically integrated companies that can combine data, evaluation, and service delivery end to end.
Mercor’s Growth, Security Incident, and Business Model
Security incident and public rumor cycle
Foody addresses rumors around a hack and revenue slowdown, saying:
- There was a security incident, but it was contained quickly.
- Customer relationships remained strong, including with OpenAI.
- The company added security as a formal company value afterward.
- He says much of the online narrative was exaggerated by Twitter/X speculation and competitor-driven misinformation.
Growth and profitability
Key claims about Mercor’s scale:
- Mercor says it added $300M in net new ARR in 60 days.
- The company is profitable and has “more cash than it has ever raised.”
- Foody says Mercor has barely burned cash since its seed round.
Revenue model
Foody pushes back on the idea that Mercor’s revenue is just pass-through GMV:
- Mercor reportedly operates at 30%–40% gross margin.
- The company is not just a marketplace for experts; it provides:
- sourcing and hiring experts,
- the platform they work on,
- AI project management,
- quality checks,
- and end-to-end task delivery.
- He frames Mercor as a vertically integrated workflow company, not just a labor marketplace.
AI Market Thesis: Models, Infrastructure, and the Death of Software Moats
“The model is the product”
Foody repeatedly argues that:
- It is becoming very hard to build defensibility in software on top of frontier models.
- Many application-layer products can be recreated quickly by the model providers themselves.
- Over the next 12 months, he expects infrastructure and data companies to outperform application-layer companies.
Why infrastructure may win
He sees defensibility in:
- network effects
- data moats
- long R&D cycles in compute/chip infrastructure
- vertical integration around workflows and evaluation
Application-layer defensibility: his rebuttal
When challenged with examples like legal AI tools, Foody argues:
- Strong network effects can still create moats.
- But for many pure SaaS products, if the model can replicate the workflow, pricing power weakens.
- The real moat may be forward-deployed implementation and tacit knowledge, not just traditional GTM.
Agents, Evals, Token Spend, and Enterprise Adoption
Agents are changing internal operations
Mercor is already using agents across the business:
- AI project manager
- interview question generation
- candidate ranking
- accounting automation
- fraud detection
Foody says these systems are paired with evals that benchmark model performance for specific workflows.
Token spend will outgrow headcount spend
One of the strongest claims in the interview:
- Mercor is already spending more on tokens for internal agents than on employee headcount.
- Foody expects that in five years, the average enterprise will spend more on compute than on salaries.
- He views this as a form of Jevons paradox: cheaper, better models drive more total usage.
Data cleanliness and tacit knowledge
On enterprise adoption barriers, he argues:
- Data structure matters, but models will increasingly clean and structure data themselves.
- The harder problem is tacit knowledge trapped in employees’ heads.
- The next big job category may be training agents and codifying company knowledge so agents can operate effectively.
Hiring, Talent Wars, and the Cost of AI Researchers
Researchers are the hardest role to hire
Foody says the most difficult hiring challenge today is:
- AI researchers, due to extreme supply-demand imbalance.
Compensation has gone nuclear
He claims top AI researchers can command:
- tens of millions in stock per year
He also mentions one candidate had a $20M/year cash-equivalent offer from Meta’s TBD / superintelligence efforts.
Culture and scaling the company
Mercor has grown from around 40 people to several hundred very quickly, and Foody says:
- early-stage founders don’t fully appreciate how hard scaling gets,
- HR and people operations become necessary,
- but the company tries not to become bureaucratic,
- and the leadership works intensely without mandating fixed hours.
Capital, Valuation, and Why Mercor Stayed Independent
Funding rounds and valuation jumps
Foody walks through Mercor’s fundraising history:
- Seed: around $23M post-money
- Series A: around $250M post-money
- Series B: around $2B valuation
- Later round: around $10B valuation
He says each felt expensive at the time, but growth outpaced expectations.
Acquisition interest
He denies rumors that:
- Amazon tried to buy Mercor for $13B
- he would sell for $30B
His view:
- independence matters because Mercor is trying to build a legendary company in a new category of work.
Broader Economic and Policy Views
Jobs, displacement, and new categories of work
Foody believes:
- there will be more jobs in 10 years than today
- but there will also be significant job displacement
- new roles will emerge around:
- training agents
- managing agents
- data collection
- AI operations
- evaluation and compliance
Tax policy takeaways
He argues for:
- reducing or eliminating income tax on the bottom half of Americans
- shifting taxation toward things with less distortionary impact, such as:
- capital gains
- consumption
- carbon or other negative externalities
His thesis: jobs are a positive externality, so taxing labor may be economically backwards.
Notable Predictions and Takeaways
Foody’s biggest predictions
- Application-layer software moats will weaken
- Model labs will become among the most valuable companies ever
- OpenAI or Anthropic could eventually be worth $10T+
- Compute/inference spend will become a bigger budget item than headcount
- Open source and distilled models will capture much of enterprise inference usage
- AI security will become a massive growth market
- Services will be automated, and AI-enabled services will increasingly become software-like
Final impression
Foody comes across as highly conviction-driven and unusually candid about:
- the economics of Mercor,
- the fragility of application-layer moats,
- the changing nature of work,
- and the scale of the AI opportunity.
The interview is as much a snapshot of where AI business economics are headed as it is a company-specific update on Mercor.