In the field of AI, data annotation is a key link in model training, but high-quality annotation talents are scarce, and market supply and demand are unbalanced. Mercor has discovered this market gap with its AI recruitment platform, gradually transforming from providing contract talents for large data annotation companies to providing human data creation services directly to AI laboratories, becoming a strong competitor in the data annotation market. This article will provide an in-depth analysis of Mercor’s business transformation, market opportunities, core competencies, and challenges, and explore whether it can create the next Scale AI in the field of data annotation.
Mercor is on a critical and yet under-met intersection of supply and demand in AI: the demand for high-quality, vertically expert human data for next-generation AI models, and the supply-demand imbalance caused by the scarcity of relevant talent. Synthetic data is not a complete subsion for human data, especially when it comes to domain-specific knowledge and complex judgments. Breakthroughs in AI models rely heavily on “human intelligence input” from vertical experts.
Mercor was initially positioned as an AI recruitment platform. On top of this, it saw an opportunity to provide “contract people” to large data annotation companies like Scale AI as a BPO (business process outsourcing) service provider. In the process, Mercor discovered its ability to efficiently match and supply experts, with a strong PMF in the field of Human Data. As a result, Mercor began offering human data creation services directly to AI Labs, transforming from an upstream labor provider for Scale AI to a direct competitor in the data annotation services market.
This transformative PMF was quickly validated, with ARR reaching $75 million in early 2025. In February 2025, against the backdrop of tens of millions of dollars in revenue and abundant cash flow, Mercor still attracted capital to complete a $100 million Series B financing funding round supported by top institutions such as Felicis Ventures, Benchmark, and General Catalyst at a very low cost of only 5% dilution, with a valuation of $2 billion.
01. Investment logic
Mercor’s business model has evolved from a simple AI recruitment platform to a direct competitor in the human data annotation market, directly competing with industry giants such as Scale AI, and its business essence is to deliver data annotation results.
Mercor identified and exploited a “market failure” point from an existing market leader. It leverages its early talent acquisition experience to provide the best speed and flexibility for small, medium-sized, challenging data annotation projects (often under $50,000 budget), a gap in the market that Scale AI’s operating model struggles to cover efficiently.
B-end product manager’s ability model and learning improvement
The first challenge faced by B-end product managers is how to correctly analyze and diagnose business problems. This is also the most difficult part, product design knowledge is basically not helpful for this part of the work, if you want to do a good job in business analysis and diagnosis, you must have a solid …
View details >
Mercor’s marketing narrative will still emphasize the capabilities of its recruiting platform, but the real value it provides to AI Labs is the ability to quickly deploy expert-level human resources for complex, rapidly iterative tasks. But Mercor’s data quality still lags behind industry leaders. The core investment question is whether Mercor’s market is large enough and profitable, and whether it can compensate for data quality shortcomings before Scale AI adjusts its strategy to address its threats.
1. The explosion of demand for human data and the prominence of long-tail value
As large models delve into the fine-tuning and RLHF stages, synthetic data can no longer be completely replaced, and the bottleneck of unlocking and releasing the next generation of intelligence still lies in expert-level human data.
This has led to two major market opportunities:
- High Value of Long-Tail Projects: Large AI labs increasingly demand responsiveness and quality for small-scale projects under $50,000, while traditional large annotators have high thresholds and slow iterations, providing Mercor with room to enter the market.
- The explosion of professional vertical tasks: In professional fields such as healthcare, law, finance, and complex tasks that require subjective judgment, the demand for senior talent labeling is growing rapidly. High-quality human annotators are scarce in the market, especially in specialized fields, and large AI companies are willing to pay a high premium for this.
Mercor targets the pain points of traditional data outsourcing models, such as inefficiency, lack of transparency, and the high cost of self-built teams, providing a transparent, efficient, and flexible solution.
2. Differentiated competition with Scale AI
Although they differ in delivery methods and target projects, both companies end up vying for the same data labeling budget from AI Labs. Mercor is emerging as a lighter, flexible, and long-tail alternative to Scale AI.
• Niche is real: Mercor serves an overlooked but rapidly growing market segment: AI Labs and SMB customers with limited budgets, complex tasks, and speed sensitivity. Mercor is filling the gap for these customers who currently don’t have a good solution, Scale’s industrialization process is not covered (the system is optimized for large contracts), and the cost of building a self-built team is too high. Mercor’s core strength is its speed, which is extremely valuable during the experimental phase of AI models.
• The core trade-off is quality: This speed and flexibility comes at the cost of lower data quality (6-7/10 on Google and 8-9/10 on the Scale). Mercor believes that speed is more important than top quality for a particular market segment. If it can gradually close the quality gap over time and expand its talent pool, it will have the ability to expand into the high-end market that Scale currently dominates in the future.
3. Team and execution
The founding team is less than 21 years old, but they have demonstrated product acumen and execution that far exceeds their peers. From starting a business in the Harvard dormitory to achieving tens of millions of dollars in revenue, to quickly winning top AI customers, it has verified its insight and ability to implement pain points. The team has demonstrated compound capabilities in product design, technical architecture, and commercial operations. Speed and execution are Mercor’s biggest moats, and Mercor has an advantage in talent matching and project launch speed compared to large suppliers.
4. PMF Verification
In just two years, Mercor has achieved rapid growth from a dormitory startup to an ARR of $75 million and a valuation of $2 billion in February 2025, and has been favored by top investment institutions and individual investors such as General Catalyst, Benchmark, Peter Thiel, Jack Dorsey, etc., fully validating its product-market fit (PMF). AI Labs’ demand for human data has just started, and synthetic data cannot completely replace expert annotation, and there is no competitor that can compete with Mercor in terms of “quick order taking + small scale”.
Key risks
1. Price war between talent commercialization and data annotation
We believe that the movement of high-quality talents across multiple platforms to maximize revenue will inevitably lead to pressure on service prices and quality. If Mercor fails to build strong enough stickiness, the moat of its talent pool could be eroded.
Strategy: Mercor needs to establish a reasonable incentive mechanism and revenue-maximizing experience to build a platform moat. Successful experiences like Uber’s have enabled healthy scale through standardized tasks and strict quality control.
2. High commissions and “disintermediation” risk
Mercor has a commission rate of up to 30% in the recruitment module, which may face customer doubts about its sustainability in the long run, especially after customers have established long-term relationships with talent, and there is a risk of “disintermediation” bypassing direct cooperation with the platform.
Strategy: Mercor needed to continuously improve the irreplaceability of its AI platform in terms of talent matching, management efficiency, and quality assurance, making customers feel that the value of the platform far exceeded the commission cost. At the same time, more diversified charging models or value-added services can be explored.
3. Data annotation quality shortcomings
Mercor’s current data annotation quality (Google Customer Score 6-7/10) is still lower than industry leaders Scale AI and Surge AI (8-9/10), which may limit its ability to take on the most top-notch and sensitive tasks.
Strategy: Mercor has partnered with Scale AI and opened up its performance and quality control system, which provides it with an opportunity for quality alignment. Combined with its rapid recruitment and iteration speed, Mercor has the potential to bridge this gap quickly. The company needs to continue to invest in QA process optimization and “golden labeler” benchmark alignment to achieve industry-leading levels.
4. AI Agents as an alternative to the potential threat of junior engineers
In the long run, AI Agents may replace some junior engineer tasks and compress the freelance recruitment market.
Strategy: Mercor’s strategic focus is on high-quality, difficult, and edge case “human data,” tasks that are currently difficult for AI agents to perform and require expert-level human judgment.
02. Market opportunities for expert data
The core of data annotation business is shifting from providing massive data to providing high-quality, expert-driven data. The market is divided into two parts: low-end tasks will be eroded by the model itself or synthetic data, while high-end, complex tasks are a market with high profits and barriers. The leaders of the future will be companies that can efficiently organize, manage, and deliver expert-level data services, not just labor-intensive data factories.
This is a macro trend, and Scale AI’s operating model objectively creates a niche for agile challengers like Mercor:
1. The “last mile” puzzle of AI capabilities
The bottleneck of the continuous advancement of models is no longer the amount of data, but the quality, diversity and complexity of the data. To address model generalization capabilities, handle edge cases, and perform complex logical reasoning, AI systems need data produced by human experts, not ordinary workers, that transcend the cognitive boundaries of existing models. This is the link with the highest value density.
2. Persistent demand for human experts
• Human Eval Demand Continuity: Judgment from frontline customers like Google suggests that the demand for high-quality human assessments will continue to grow over the next 5-10 years. Even if synthetic data is feasible, human experts are still required to evaluate the model to ensure that its performance does not degrade with iteration.
• Specialized and complex tasks: The manual need for simple tasks will decrease, but specialized fields (such as healthcare, law, finance) and tasks requiring deep logical reasoning will still rely on human experts for a long time. Mercor’s customer Google judged that large models would struggle to generate the same quality of professional domain data in the next two years.
3. Talent barriers formed by “high-quality premium”
• Talent scarcity: The demand for high-quality annotators far exceeds supply, especially in native English-speaking markets and specific areas of expertise (e.g., experts with PhDs).
• Willingness to pay: Hyperscalers have shown a strong willingness to pay a high premium for obtaining top talent, such as hiring a U.S. graduate student to annotate complex reasoning tasks.
4. The inertia of Scale AI’s operations has given rise to opportunities in the “long-tail market”
• Limitations of Scale AI: The operating model and cost structure make it difficult to fit small projects with budgets under $50,000, and its delivery and iteration rates are slow.
• Agile Needs in AI Labs: The exploration phase of cutting-edge models requires a lot of rapid, small-scale iterations. AI labs are increasingly demanding responsiveness and quality of delivery for small projects under $50,000.
• Mercor’s market entry point: This “small budget, high difficulty, fast cycle” market gap just provides Mercor with a unique entry space. It can take on these projects faster and more flexibly.
Market size estimation
Total Market (TAM)
According to Grand View Research, the global data labeling market was valued at approximately $3.7 billion in 2023. It is expected to reach $17.1 billion by 2030, with a CAGR of about 23.5%. Of these, about 20%-30% are high-complexity, human expert participation annotations.
LLM-Driven High-Value Submarkets (SAMs)
LLM development has given rise to new demands for RLHF, expert evaluation, and more. We estimate that around $5-700 million will be dedicated to such high-quality human data projects worldwide in 2024. This submarket is expected to grow to billions of dollars at a CAGR of 50-80%.
Demand side: More than a dozen leading AI labs, including OpenAI, Anthropic, Google DeepMind, and Meta, generally have annual budgets in the range of $5M–$20M for human evaluation and model alignment, with the top 10 alone exceeding $100M–$200M combined. This demand is spreading rapidly as enterprise customers (such as healthcare, finance, legal, etc.) begin to fine-tune their own models.
Supply side: Among annotation service providers such as Scale AI, Surge AI, and Mercor, Mercor’s ARR in 2025 alone will reach $75M, mainly from this type of task. Scale AI also estimated $50M+ in RLHF-related revenue, indicating that actual spending in this segment has reached $200M–$300M+ and is still accelerating.
Mercor has access to market (SOM) potential
The company currently has an annual ARR of approximately $75 million (February 2025), corresponding to a 10-15% share of the high-end Human Data submarket (2024 caliber). If the annual growth rate of >50% is maintained, the revenue can reach 2.5-30 million US dollars in the next two years, corresponding to a market segment of 20-30%.
03. Business evolution: from recruitment platform to data annotation
Business model
Mercor was initially positioned as an AI recruitment platform. On top of this, it saw an opportunity to provide “contract people” to large data annotation companies like Scale AI as a BPO (business process outsourcing) service provider. In the process, Mercor discovered that its ability to efficiently match and supply experts had a strong market demand (PMF) in the field of Human Data. As a result, Mercor began offering human data creation services directly to AI Labs, transforming from an upstream labor provider for Scale AI to a direct competitor in the data annotation services market.
Core Business Lines:
Currently, Mercor operates two main business lines, which share its underlying AI talent matching technology and expert network:
1. AI Recruitment and Mobility: In traditional businesses, Mercor offers AI recruitment services for tech companies, helping them find and hire full-time or contract engineers. For example, Epsilon Labs hires overseas MLOps contract engineers through Mercor, from which Mercor receives a salary-based commission (like 30%).
2. Human Data Services: This is the company’s current core growth engine and an area that competes directly with giants like Scale AI. Mercor provides expert human data annotation services for model fine-tuning, evaluation, and RLHF for AI labs, such as the world’s top five AI labs.
Mercor built an end-to-end human data delivery system on three lines:
1. The Asset: Mercor’s core asset is a deeply indexed and screened talent network of over 300,000 experts. Unlike traditional outsourcers who hide the background of the annotator, Mercor presents candidate profiles (education, work experience, skills assessment, etc.) to clients in full transparency.
2. Flexible Workflow Integration: Mercor’s product format is highly flexible, effectively reducing customer adoption resistance.
• Lightweight access: Customers can continue to use their own labeling platform, with Mercor acting only as a “plug-and-play” provider for high-quality talent.
• Complete Solution: For customers without off-the-shelf tools, Mercor offers its own platform to quickly set up a complete workflow from data pipeline to quality control.
3. Structured Quality and Incentives (The Framework): Mercor abandons the traditional piecework model (which is naturally oriented towards sacrificing quality) and instead pursues a “pay by the hour” model. For top AI labs, the value of R&D iteration speed and data quality far outweighs the cost savings of labor alone. On-time billing directly incentivizes high-quality investment and flexible adjustment, which perfectly meets the needs of the cutting-edge model exploration stage.
Mercor also assumes all the legal and compliance risks associated with classifying labelers as independent contractors. This removes a critical operational hurdle for the client, allowing them to quickly launch and scale their teams without worry.
Candidate perspective
The vast majority of the positions offered by Explore are contractor, with the exception of a very small number of full-time salaried employee positions. The platform provides candidates with a rich AI interview experience with detailed feedback. In some specific projects, there is also a corresponding customized interview process.
Company perspective
When using Mercor, companies can easily browse candidate profiles, including resumes, interview records, and online information, to quickly identify the best candidates. The platform also supports one-click appointment for interviews, or directly issue contracts to candidates to achieve same-day employment; All payments can also be completed in one go through the Mercor platform, ensuring a compliant global payment process.
Technical architecture
Mercor’s technology roadmap is based on insights: the value created by the top talent (Top 1%) in knowledge-based jobs follows a power law distribution, and at the heart of its technology architecture is to build an insight engine that can accurately identify and predict these “critical minorities”.
Hierarchical AI Architecture: Mercor employs a hybrid architecture of “common foundation model + vertical domain model”. The bottom layer uses large models such as GPT4 to handle general tasks such as resume parsing, and the upper vertical domain models include:
1. Core IP: Based on real job performance data from more than 100,000 customer feedback, the “Job Competency Prediction Model” is trained through reinforcement learning (RLHF).
2. Domain knowledge injection: The model is connected to industry-specific knowledge bases (such as medical UMLS and manufacturing ISO standards), which has industry scalability.
3. Signal networks beyond resumes: Analyze candidates’ dynamic, unstructured “text signal networks” such as GitHub code quality, technical blog depth, and decision-making logic in interviews to build strong correlations with job performance. Evaluate “what can actually be done” and “what the way of thinking is”.
business model
Mercor’s business model is built on its two core businesses: “AI Recruitment Platform” and “Human Data Services.” Its AI recruitment capabilities provide a stable and high-quality talent supply for its human data services, forming a unique synergy. Customers, especially top AI labs like Google, buy the end result delivered by it, valued in quality, speed, and security, not the process itself.
Talent Competition: Both sides are competing for high-quality talent such as Indian/North American remote freelancers, programming competition players, etc. Mercor attracts talent by offering higher hourly wages, while Scale AI offers more flexible hours and diverse project opportunities.
Lab Selection Preferences: The choice of AI Labs depends on the specific needs of their project
1. Choose Mercor’s scenario: The project budget is small (less than $50,000), requiring extremely high iteration speed and flexibility; The task is difficult and requires the participation of specific, screened experts; Accept delivery quality that is “good enough” rather than “top of the industry” in exchange for speed of development.
2. Choose Scale AI’s scenario: The project budget is sufficient and requires large-scale, standardized data production; Extremely stringent requirements for data quality, QA processes, and compliance; The project cycle is long, and the speed of a single iteration is not high.
Mercor’s model is more focused on “talent-driven, agile delivery.” The core selling points are speed, flexibility, and talent transparency. It essentially uses high-end talents as a plug-and-play service that can be flexibly scheduled to solve the small-scale and fast-cycle R&D needs of customers.
Scale AI focuses on “process-driven industrial production”. It sells standardized delivery quality, predictable output at scale, and end-to-end compliance. The customer buys a reliable, worry-free “black box” for large-scale data production tasks.
This is also a dynamically evolving market, and Mercor may use its strengths in high-end talent services to gradually encroach on some of the larger projects that require greater flexibility. Scale AI may also iterate on technology to improve its efficiency in handling small-scale tasks.
04. Mercor’s core customer and user feedback
Core customers and two use cases
Mercor’s customers use its platform in two main ways, depending on their needs:
Use case 1: Human Data Labeling (core growth business)
Customer group: Leading AI Labs (OpenAI, etc., also including competitor Scale AI). The primary need is to provide expert-level human assessment and complex data annotation for the fine-tuning and RLHF phases of GenAI models. Mercor provides these labs quickly, in part, a team of expert contractors, who are billed by the hour to participate in model regression testing, output high-quality evaluations, and handle complex annotation tasks for edge cases.
Use case 2: AI recruitment and relocation
Customer Group: Small and medium-sized AI startups. For example, Epsilon Labs recruits overseas MLOps contract engineers through Mercor to be responsible for pipeline construction, training, monitoring, and assistance. Axion Ray saved time on resume screening with Mercor AI interview system and tried to match engineers, but switched to traditional channels because the candidate’s expertise didn’t match their needs.
An executive from Scale AI commented that Mercor AI video interviews improved initial screening efficiency and consistency, with faster delivery than the industry average. However, the limitation is that difficult tasks still need to be manually reviewed, and the quality cannot be fully guaranteed; Moreover, some candidates produce low-quality data after passing the screening, resulting in the platform being banned and re-screened. Scale now opens up data from its performance and quality control systems to help Mercor optimize the quality of its talent pool, and Mercor continuously improves based on Scale’s data to keep up with labeled quality.
As a major customer of many data annotation service providers, Google provides the most objective horizontal comparison. Google’s Machine Learning team said that they will first test multiple suppliers with “golden labeler”, and then distribute tasks in batches after the quality is qualified, Mercor currently accounts for a small proportion, although it is growing rapidly, but the absolute annotation volume is much lower than that of Scale AI and Surge AI, Scale/Surge AI maintains 8–9 points all year round, and Mercor currently has about 6–7 points, which still needs to be aligned multiple rounds. But in terms of delivery speed, Mercor is the fastest. In terms of professional assessment, Google’s ML team believes that in fields that require deep vertical knowledge, such as healthcare, Mercor still needs to expand its professional talent pool to compete with relatively vertical specialized platforms such as Turing and CertAI.
User perspective
We also looked at user feedback on reddit. Many freelancers report that Mercor’s salary is attractive (about $50/hour for undergraduates, $100–200/hour for master’s and doctoral degrees), and it is 100% remote; Moreover, the current combination of AI interviews, behavioral screening, and technical assessments does improve matching efficiency.
Mercor is also very controversial, saying that Mercor is inviting users to write about Reddit experiences, although it is not limited to good or bad, and has also raised some questions about the “real nature of work”. Some users also reported that “there are more monks and less porridge”, the task volume is unstable, and there is a phenomenon that “it is difficult to quickly obtain projects after launch”.
05. Differentiated competition: Mercor is filling the gap left by Scale
In the data annotation market, Scale AI has an absolute advantage. Mercor takes a differentiated approach, targeting small-budget, challenging projects that Scale AI cannot cover.
Medium and large projects
Scale AI
1. Market positioning: With high-quality and large-scale delivery capabilities, it is a market standard setter for major cloud vendors and AI laboratories. Its subsidiaries Remotasks, Outlier and other crowdsourcing platforms manage a global network of freelancers (more than 100,000 people).
2. Advantages: Large scale and flexibility, able to handle diverse and complex projects (e.g., audio, browser tracks), thanks to a large network of contributors and engineering capabilities, its technology platform, quality control processes and global talent network are core assets. However, its operating model and cost structure are not suitable for smaller projects with budgets under $50k, creating a market opportunity.
3. Latest developments: 1/ Previously, it served multiple fields (such as autonomous driving, maps, e-commerce), but in the past two years, the focus has shifted to GenAI; 2/ Scale ai is also working on high quality and diversity, and it has already worked well in some vertical fields.
4. Team and Financing: Scale AI has raised a total of $1.6 billion, with the latest round being a $1 billion Series F in May 2024, valued at $13.8 billion. In March 2025, it was reported that it was seeking financing at a valuation of up to $25 billion through a tender offer.
Surge AI
1. Market positioning: Regarded as number one competitor by Scale AI, second only to Scale AI in terms of scale. Serving Anthropic, Google, OpenAI, and more, offering RLHF and human data platforms. At present, the market share, overall quality, recruitment and iteration speed are acceptable. Its crowdsourcing platforms DataAnnotation.tech, Taskup.ai, and Gethybrid.io enable flexible, high-quality universal task delivery.
2. Advantages: Surge AI is much smaller than Scale AI but offers similar comprehensive and flexible services, capable of handling projects of different sizes (suitable for small, medium and large projects, competitive in price-sensitive projects) and excels in the field of Gen AI.
3. Team and financing: Founded by the data team of the former major manufacturer, it completed a $25 million Series A financing round on July 1, 2020 alone.
•Founder and CEO: Edwin Chen, former head of machine learning and content moderation teams at Google, Facebook, Twitter, with a background in mathematics and linguistics at MIT.
• Engineering Team Leader: Andrew Mauboussin, former Twitter Machine Learning Engineer with a Harvard computer science background, responsible for real-time human computing APIs and international data collection in 30+ languages.
• Head of Product & Growth: Bradley, former head of data operations at Facebook, graduated from Dartmouth, leading product development and growth strategy.
Vertical Professional Mission
These companies are Mercor’s rivals in the high-quality segment. They have established strengths by deeply cultivating specific areas: Surge (RLHF), Labelbox (CV tool), Turing (certified experts). This proves that verticalization is an effective defense strategy. Mercor needs to prove that its choice of “difficult general-purpose tasks” is an equally solid market segment.
Turing
1. Market positioning: It is a major vendor in the coding field for OpenAI and other LLM producers. Initially a software developer marketplace platform that matched enterprises and developers, it turned to data annotation and AI training due to customer needs such as OpenAI. In 2023, the software engineering market will slow down due to layoffs, but the demand for data annotation and basic model support will surge, Turing has not completely abandoned the recruitment business, but has increased investment in the field of AI data annotation, and has advantages in the field of code-related data annotation and software engineering, with customers including top AI labs and technology companies such as Anthropoic, Google, Nvidia, and OpenAI.
2. Advantages: 1/ Provide top talent, data, and tools to train cutting-edge models, help AI labs improve model performance, and are good at recruiting expert talents in specific vertical fields according to user interviews. 2/ According to an interview with Scale AI, only code is done and the code field is very good, and the solution is vertical and in-depth (although it cannot be completely monopolized); 3/ The platform has a huge database of 3 million developers, mainly from developing countries such as India, but is not a real “talent network”.
3. Recent Developments: Turing plans to expand into reasoning, mathematics, and STEM fields, as these are natural extensions of code and increased demand for multimodal models; Turing is not satisfied with low-end data annotation, aiming to elevate the value chain through AI consulting and AI application development.
4. Financing: A total of approximately $247 million has been raised through six rounds of financing, with Turing valued at $2.2 billion and an ARR of approximately $300 million after the Series E funding round in March 2025. Investors include WestBridge Capital, Foundation Capital, Khazanah Nasional Berhad, StepStone Group, Scott Banister, Adam D’Angelo, and others.
5. Team: The two founders, Jonathan Siddharth and Vijay Krishnan, both graduated from Stanford University and co-founded the AI startup Rover (later sold), along with AI technology experts from Meta, Google, Microsoft, Amazon, Stanford, Caltech, and MIT.
Labelbox
Valued at approximately $500 million in 2024, it offers flexible tools to support image, text, and video annotation, widely used in healthcare, retail, and autonomous driving.
1. Market Positioning: Focuses on providing flexible and efficient annotation tools for AI startups, small and medium-sized enterprises, and research institutions, supporting general-purpose tasks and multimodal data (text, images, videos). Its platform is known for its user-friendly interface and integrated workflows, making it suitable for rapid project iterations. Labelbox is more of a tool provider than an end-to-end service and may be less competitive on high-complexity tasks such as RLHF.
2. Advantages: Compared with providing data annotation, it pays more attention to data management, provides a self-service annotation platform, customers can customize workflows (such as sentiment analysis templates), support small batches, highly customized tasks, and reduce dependence on external annotation teams; In data annotation, CV images have an advantage; Google customer interviews point out that labelbox can also recruit high-quality talent.
3. Team and Financing: Cumulative funding of $188.9 million, with a valuation of approximately $1 billion after Series D in January 2022. Investors include SoftBank Vision Fund II, Andreessen Horowitz, B Capital Group, Kleiner Perkins, Gradient Ventures, Databricks Ventures, Snowpoint Ventures, In-Q-Tel, and others. The founding team is a Stanford graduate with a rich data background, founder Manu Sharma is a former Microsoft engineer, and Brian Rieger previously worked in data at Boeing and other companies.
Mercor’s market entry point is to serve customers that large vendors overlook. The current positioning is closer to Surge AI, SuperAnnotate, Turing. Mercor took on projects that Scale AI was reluctant to do because it was too small. These projects usually require high quality and difficulty, but the amount of data is small and requires more flexible delivery. This is a pragmatic initial market strategy. The question is whether this part of the demand can be converged into a large-scale, high-profit business.
In addition, in the field of recruitment, Mercor also competes head-on with popular platforms such as Upwork and LinkedIn, as well as vertical tools such as Juicebox and micro1.
06. Mercor’s talent structure and atypical financing story
Core team background
Mercor has three young, complementary and entrepreneurial talented founders: Brendan Foody (CEO), Adarsh Hiremath (CTO), and Surya Midha (COO). The average age is only about 20 years. CEO Brendan Foody was born in 2004, from trying to sell donuts and reselling sneakers in middle school to setting up an AWS consulting firm in high school, this entrepreneurial “talent” drove Mercor to start in a Harvard dorm room and grow his business to millions of dollars ARR without external financing. Brendan Foody dropped out of Harvard’s computer department after two years to go all in Mercor without financing. CTO Adarsh Hiremath has developed a close relationship with Brendan in a past project (a consulting firm for high schools), where he is responsible for engineering and AI capabilities. COO Surya Midha has a background in policy debate and operations management. The three won the policy debate competition together in the high school debate team.
Core operations and growth executives from Scale AI and OpenAI are experienced in data annotation, supply chain construction, customer delivery, and government relations, shortening Mercor’s learning curve on complex B2B service processes:
Shaun VanWeelden (ex OpenAI、Scale AI)
Mercor’s director manager increased the company’s annual revenue from $1 million to $100 million in 11 months. As the head of Human Data Operations at OpenAI, he understood RLHF’s data operations and quality standards and made up for the team’s shortcomings in enterprise operations.
He previously served as Head of Human Data Operations at OpenAI, where he led teams to provide RLHF data for AI research, optimize processes and supplier management, and represent the company in congressional hearings to explain human participation mechanisms in models. Previously, at Scale AI, he was promoted from customer engineer to director of computer vision projects, leading autonomous driving data annotation projects for customers such as GM and Toyota, and facilitating the first federal government cooperation contract ($5 million).
Sidharth Potdar (ex Scale AI)
Mercor Head of Operations, where he previously served as the Head of Growth at Scale AI, where he was responsible for product operations, strategic projects, and growth teams, focusing on AI data annotation and autonomous driving technical support. Previously worked as an analyst at McKinsey, participating in a number of strategic consulting projects.
Mercor’s organizational culture is built on strong founder and builder genes. With more than half of the team members being former founders and a median age of just 22, we have reason to trust the team’s speed of execution and owner-ship. However, the founding team is young, starting a business for the first time, and it is a B2B business, which is different from product-focused companies such as Facebook and Cursor, which face more challenges in managing companies in the growth stage.
Financing history
Mercor’s financing journey is not a traditional “fundraising-development” path, but an “active pursuit” of top capital.
1. Seed Round (2023)
The company was founded with a $3.6 million seed round led by General Catalyst and attracted individual participation from NEA Chairman Scott Sandell, as well as Soma Capital and Link Ventures. The three founders were only 19 years old at the time, and the company was still in its infancy.
2. Series A (September 2024)
Benchmark led a $30 million Series A funding round, with its valuation soaring to $250 million.
This round of financing brings together heavyweight individual investors such as Peter Thiel, the “Godfather of Silicon Valley Investment”, Twitter co-founder Jack Dorsey, director of OpenAI and CEO of Quora, and former U.S. Treasury Secretary Larry Summers. The addition of these strategic investors not only brings funding but also indicates the empowerment and resource introduction of Mercor’s future business development, especially the cooperation with top AI Labs.
The addition of Victor Lazarte, a partner at Benchmark, to the board reflects the fund’s deep investment and long-term support for Mercor, not just a financial investment. Benchmark’s intervention is an active capture of Mercor. (Victor convinces Brendan to have an initial exchange, after which Peter Fenton, one of Benchmark’s partners, invites Brendan to join him in a helicopter ride.) Brendan then talked to Victor and the Benchmark team a few times)
3. Series B (February 2025)
With the company’s business revenue reaching tens of millions of dollars and not in a hurry to raise funds, Mercor still completed a $100 million Series B funding round valued at $2 billion, led by Felicis Ventures, with participation from existing investors Benchmark, General Catalyst and DST Global. The financing diluted only 5% of the equity, securing a huge amount of money at a very low dilution ratio.
