Mercor Raises $750M Series D at $20B Valuation to Scale Enterprise AI Agents

Mercor has raised $750 million in Series D funding at a $20 billion valuation to expand its AI training, evaluation and post-training infrastructure. The San Francisco-based company is targeting enterprises that want to develop AI agents capable of performing complex professional tasks reliably.

The round includes participation from Benchmark, Felicis, Fundomo, funds managed by affiliates of Apollo, G Squared, General Catalyst, Glade Brook Capital, NVIDIA, Salesforce, Sapphire Ventures, StepStone Group, XTX Ventures and Zoom.

The latest investment doubles Mercor's valuation from its $10 billion Series C in October 2025, highlighting growing investor interest in infrastructure that helps organisations train, test and deploy AI systems.

Mercor plans to use the funding to accelerate research and develop realistic environments where AI agents can learn professional workflows. Its broader objective is to help businesses measure how reliably AI agents complete tasks and understand the costs associated with their performance.

What does Mercor do?

Founded in 2023, Mercor initially connected AI companies with professionals who provided specialised knowledge and expertise to train advanced AI models.

Its business has since expanded beyond supplying human expertise for training data. The company is building infrastructure that helps organisations evaluate AI models, improve their capabilities and prepare agents for real-world professional workflows.

Mercor says its network includes more than five million experts and serves frontier AI laboratories, technology companies, AI startups and Fortune 500 businesses.

The company also reports having approximately 700 employees.

Its expansion reflects a broader shift in enterprise AI: businesses increasingly need systems that can perform specific professional tasks rather than simply generate text or answer general questions.

Why enterprises need customised AI evaluations

AI models can perform well on public benchmarks but still struggle with the specific systems, policies and operational requirements of individual businesses.

For example, an AI agent might successfully complete a general research task but fail when navigating a company's internal software, interpreting its procedures or handling an unusual customer request.

Mercor addresses this challenge through customised evaluations, commonly known as evals.

These evaluations measure AI performance against an organisation's own requirements and workflows. Instead of relying exclusively on general benchmarks, companies can assess whether an AI system performs the tasks it is actually expected to handle.

Mercor's approach includes:

• Customised benchmarks: Evaluating AI systems against company-specific workflows and standards.

• Expert-driven training: Using professional knowledge to improve AI capabilities.

• Post-training infrastructure: Identifying where models need further training to improve their performance.

• Realistic working environments: Allowing agents to practise complex tasks before deployment.

• Performance evaluation: Measuring task completion reliability and associated costs.

The company also operates APEX, its public benchmark series designed to measure AI performance on economically valuable professional work.

Building realistic environments for AI agents

An important part of Mercor's strategy is helping AI agents practise tasks in environments that resemble actual enterprise software and workflows.

Earlier in 2026, Mercor acquired Deeptune, whose team had spent two years recreating hundreds of enterprise applications as training environments.

These simulated environments allow AI agents to practise complex workflows before being deployed in real business operations.

This is relevant because enterprise software environments often contain interconnected applications, permissions, exceptions and processes that general-purpose AI benchmarks do not fully capture.

By combining professional expertise, customised evaluations and simulated work environments, Mercor wants to help organisations move from experimenting with AI agents to deploying them in operational settings.

Three founders behind Mercor's growth

Mercor was founded in 2023 by Brendan Foody, Adarsh Hiremath and Surya Midha.

Brendan Foody and Adarsh Hiremath serve as co-CEOs of the company. The founding team has built Mercor around connecting professional expertise with the development and evaluation of advanced AI systems.

Foody believes that businesses will increasingly require AI agents tailored to individual professional roles, alongside systems that can train and evaluate those agents using actual company workflows.

Hiremath has similarly emphasised that AI performance must be measured against the requirements of each organisation rather than relying solely on general-purpose benchmarks.

The latest funding will help the founders expand Mercor's enterprise offering beyond its initial customer base of AI laboratories and technology companies.

Mercor's growth and $20B valuation

The Series D follows Mercor's $350 million Series C funding round in October 2025, which valued the company at $10 billion.

The latest $750 million investment doubles that valuation to $20 billion in approximately one year.

In July 2026, Mercor announced an annualised revenue run rate exceeding $2 billion. However, the supplied reporting distinguishes this figure from net revenue, noting that contractor payments account for a substantial portion of gross billings.

The company has also reported a network of more than five million experts, although the earlier reporting clarifies that this refers to candidates vetted rather than five million active contractors.

These distinctions matter when evaluating Mercor's actual commercial scale. Its reported revenue run rate, expert network and customer reach indicate substantial activity, but they should not be interpreted as equivalent to net revenue or active workforce size.

Competition from Scale AI and Surge AI

Mercor operates in a competitive market involving companies that provide AI training data, human feedback, evaluation tools and specialised datasets.

Scale AI is one of its major competitors. The company raised $1 billion in Series F funding in 2024 at a valuation approaching $14 billion. Meta subsequently invested $14.3 billion for a 49% stake.

Surge AI is another major competitor in expert-driven AI training data. The supplied report says it was reportedly considering its first external funding round at a valuation above $25 billion.

Other companies are also expanding into this market:

• Snorkel AI: Raised $350 million at a $3.5 billion valuation to expand expert-built datasets and simulated environments.

• micro1: Raised $35 million at a $500 million valuation to expand its expert network.

Mercor's differentiation lies in combining its professional expert network with customised evaluations and realistic training environments. This positions the company beyond traditional data labelling and towards infrastructure for training and assessing enterprise AI agents.

What the $750M funding means for Mercor

The new Series D gives Mercor additional resources to expand its research, evaluation and post-training capabilities.

Its strategy reflects an important development in enterprise AI. As businesses move beyond AI experimentation, they need ways to determine whether agents can complete actual work accurately, consistently and at an acceptable cost.

Mercor is positioning itself as a platform that helps organisations train, evaluate and improve those systems using professional expertise and realistic workflows.

However, competing with established AI data providers will require continued investment in technology, enterprise relationships and evaluation capabilities.

Mercor's $750 million Series D at a $20 billion valuation marks a major investment in enterprise AI infrastructure. Its next challenge is to turn its expert network and evaluation technology into the trusted platform businesses use to deploy AI agents across professional workflows.