Mercor Raises $750M at $20B Valuation | CEO Brendan Foody

Mercore's Mission and Motivation

• The founders explicitly reject financial exits, stating they prefer working "100 hours a week" to solve the critical problem of how humans contribute to a rapidly evolving economy where AI handles repetitive tasks.
• The company aims to become a "category-defining" firm by shifting the human role from executing repetitive work (e.g., customer support, redlining contracts) to training AI agents to automate those tasks.
• Founders maintain a "maniacal sense of urgency" driven by "varying chips on the shoulder," prioritizing first-principles thinking over mimetic competition to avoid being "second place."
• Key operational values include "high standards" in hiring, "time travelers" (long-term orientation based on 5-10 year futures), and "customer obsession."

Operational Metrics and Growth

• Daily payouts to experts have surged from $1 million annually (24 months ago) to over $5 million daily.
• The company has paid out approximately $1 billion historically with no major payment failures, utilizing a dedicated payments team and Stripe Connect infrastructure.
• Since September of the previous year, Mercore has grown by "more than 5x" on a mid-nine-figure base.
• The platform hosts over 5 million active users seeking work, with 70% of talent acquired through referrals.
• The company has conducted over 7 million interviews, leveraging an in-house AI interviewer built in April 2023 to scale talent evaluation across low-density talent markets.

Product Structure and Business Model

• Mercore operates as a data provider rather than a traditional marketplace (GMV); experts are ingredients for constructing Reinforcement Learning (RL) environments.
• Three core product pillars constitute the RL environment delivery:
  • Apps: Clones of enterprise software (e.g., Google Workspace, Salesforce) to serve as simulation environments.
  • Worlds: The specific content within those apps (documents, slides, messages).
  • Tasks: Prompts and rubrics used to grade model performance, similar to academic grading.
• The company's AI Productivity Index (APEX) benchmarks frontier model performance with partners like Harvey and Cognition.
• The business model leverages "Jevons Paradox," where increasing the efficiency of human data creators exponentially expands demand for their services.

Strategic Acquisitions and M&A

• The company acquired DeepTune to secure competency in building "Apps" (software clones), addressing a bottleneck in its RL environment production.
• DeepTune was a "Series A" funded company founded by Tim, with a customer-initiated partnership signal.
• The firm previously acquired Sepal; DeepTune is noted as the larger and more advanced acquisition to date.
• M&A is viewed as a "high leverage" strategy for building durable moats, focusing on acquiring mission-aligned founders rather than just technology.
• The founders have received acquisition offers allowing them to "walk away with a lot of cash" but consistently decline in favor of building a "legendary company."

Market Thesis and Future Outlook

• Data Moats vs. Commoditization: The founders argue that while models, compute, and app layers are prone to commoditization, data and brand moats remain the only durable competitive advantages for enterprise companies.
• AI Sovereignty: Enterprises will not allow frontier labs (e.g., OpenAI, Anthropic) to train on their proprietary data; instead, companies must build their own "sovereign AI stacks" to codify competitive advantage.
• Investment Thresholds for Enterprises:
  • <$10M inference spend: Companies are generally not focused on evals.
  • $10M–$100M inference spend: Companies should allocate ~10% of the budget to building evaluation datasets.
  • <blockquote>

    $100M inference spend: Companies must begin post-training (fine-tuning) their own models to amortize fixed costs across massive inference volumes.

    </blockquote>
• Human Labor Demand: The market for humans teaching models is bounded by the "superintelligence timeline"; as long as AI cannot perform all physical or digital tasks, human feedback remains a necessary stasis point for model improvement.
• Enterprise Transformation: By the end of the next decade, all Fortune 500 companies will function like "Neo-labs," integrating full-time employees and contractors to curate high-quality data for internal model training.
• IPO Status: An IPO is not expected within the next 12 months, though it remains "on the radar" as the company scales.

Competitive Landscape and Industry Observations

• Founders cite Jeff Bezos and Elon Musk as inspirations for long-term thinking and endurance through pain.
• The founders express concern for the constant criticism faced by figures like Sam Altman and Dario Amodei.
• They view the "inference layer" (e.g., Fireworks, Space10) as needing to build custom models to create switching costs, as raw GPU resale is highly commoditized.
• API-based businesses are identified as "most vulnerable" due to near-zero switching costs and overrated revenue quality by investors.
• Mercore partners with Harvey (legal AI), Cognition, and Labs to provide the expert data required for post-training frontier models.

Founder Routine and Culture

• Brendan (CEO) works a six-day week, waking at 7:45 AM, working out until 8:30 AM, and working in the office from 9:00 AM to 10:00 PM.
• His time is split roughly equally among hiring, customer interaction, and research/product.
• The company actively hires thousands of experts and full-time staff in engineering, research, and operations, contradicting the narrative that AI eliminates jobs.