Mercor raises $750 million in Series D round to train AI on all types of tasks
Read more
Ventureburn
ventureburn.com

Mercor raises $750 million in Series D round to train AI on all types of tasks

A week ago, OpenAI raised $11 billion at a valuation of $1.25 trillion, demonstrating the continued influx of capital into the artificial intelligence sector. However, this week market attention shifted to Mercor, which successfully closed a Series D funding round, raising $750 million.

Under the terms of the deal, the San Francisco-based startup is now valued at $20 billion, representing a doubling of its worth compared to just over a year ago. The capital raised will be used to train AI systems to perform any task while compensating real people for instructions on executing practical operations.

The round was announced on November 10th. Although the lead investor was not disclosed, the list of major participants included Benchmark, Felicis, General Catalyst, Sapphire Ventures, NVIDIA, Salesforce, and Zoom. Other investors included the Apollo fund, G Squared, StepStone Group, and XTX Ventures, as well as some funds derived from employee stock option offerings.

Mercor previously raised $350 million in October 2025 when its valuation was $10 billion. In a short period of 13 months, the company increased its market value to $20 billion.

Mercor was founded in 2023 by three friends: Brendan Fudi, Adarsh Hiremath, and Surya Midha. Initially, the company was conceived as an AI-powered recruitment solution, but the founders discovered a larger problem within the AI industry.

Human experts are critically important for creating and testing AI models, including specialists such as lawyers, doctors, bankers, and engineers. Mercor connects these experts with laboratories that require their expertise.

Today, the company claims to have a network of over 5 million experts across more than 800 professions in the United States. As of July 2026, it reports annual revenue exceeding $2 billion. It should be noted, however, that this is the total billed amount, and contractors receive a significant margin.

As Fudi notes, 'AI can only be as useful as the human brain behind it.' Until recently, Mercor's main clients were cutting-edge AI labs. The new funds are intended to expand the company's services to the business community.

The plan is simple: Mercor develops custom tests based on a specific company's workflows. These tests demonstrate the model's performance using the company's own data. The results are then used by Mercor for further model training. Hiremath explains that this is important because 'a model might score highly on a public benchmark but fail.'

Every business has its unique characteristics and rules. This is why Mercor acquired Deeptune earlier this year. The Deeptune team spent two years restoring hundreds of corporate applications. Now users can practice with these copies using AI agents without affecting real systems.

AI agents are approaching the point of taking over business processes. The main question is whether they can function. Nevertheless, investors have doubled down. The market for training AI on all tasks is rapidly developing, and Mercor aims to take a leading position.

Similar stories

Remanence raised $30 million to help companies train and own their own AI models
Read more
ventureburn.com

Remanence raised $30 million to help companies train and own their own AI models

Every time a business uses an artificial intelligence agent, it pays for the services in the form of credits, and the more the agent is used, the higher the bill becomes. To solve this problem, the startup Remanence has raised $30 million in a seed round.

This Parisian startup helps businesses train and own their own AI models. It offers a simple solution: instead of renting, companies can start owning their own models. This round is significant for teams developing products using AI agents.

The round was led by EQT Ventures and FirstMark. Other participants included Zero Prime Ventures, Emblem, Drysdale Ventures, Global Founders Capital, and Yellow. Prominent figures such as Olivier Pomel (Datadog), Philippe Corrot (Mirakl), Romain Niccoli (Pigment), and Mehdi Ghissassi (AI71) also participated in the round.

Information about Remanence's seed round was leaked beforehand because the founders were seeking $30 million. This is a substantial amount for seed funding for a company just emerging from stealth mode.

Most companies today use AI models that run on rented services. The more intensively an AI agent is used, the higher the costs. Furthermore, the data generated by these agents often remains unused. Remanence views this data mass as a valuable asset.

The company claims that businesses can independently train their models based on their own workflows and become owners of them. Remanence's mission is to ensure high performance, low latency, and low costs.

This is achieved by specializing a general model for a specific application. However, Remanence's goal is to make the process of creating and owning AI agents, which was previously complex for most companies, more accessible and cheaper.

Remanence integrates with existing business AI systems. It captures the trace logs of these systems and converts them into individual training configurations. Then, Remanence applies reinforcement learning—a trial-and-error method that rewards good outcomes to improve open models. The parameters of these models are published, making their operation less expensive compared to closed models.

The more agents work, the smarter the proprietary model becomes. Data remains relevant instead of companies simply storing their data in archives. The founders are familiar with the AI field. Three founders previously worked at the French AI company H Company. Cedoz is a Google DeepMind alumnus. Hamza Bensekroun, a former leader at InstaDeep, is a Moroccan entrepreneur. The team is supplemented by Axel Moyal.

Remanence is not the only player in this race. Recently, investors have begun betting that model ownership will be the next step in AI development, as businesses prefer to own rather than rent. In August, River AI raised $1.1 billion to help companies train and own their own models.

To compete, Remanence will direct new capital towards computing infrastructure and hiring additional engineers. Funds will also be used to open offices in London and San Francisco, although the company is currently based in Paris. There are no officially named clients or pilot projects yet, so Remanence's proof is still ahead. Nevertheless, if you have paid for AI at work or created products based on it, this round is a preview of where costs and control may move in the future.

Popular