The AI startup fueling ChatGPT’s expertise is now valued at $10 billion
Mercor manages contractors around the world who help chatbots learn to think and speak like humans.
Mercor, a startup that has become a critical component in the ecosystem improving top AI models, is finalizing a new funding deal that would value the company at $10 billion, people familiar with the matter said.
That is five times the value the company had in February before it pivoted to one of the more lucrative arenas in the artificial-intelligence boom: hiring thousands of white-collar professionals to train the very machines that could one day replace and augment their work.
Mercor, which counts OpenAI and Anthropic among its customers, is set to raise $350 million, the people said. Co-founded in 2023 by three college dropouts, Mercor manages 30,000 contractors around the world who label images, write sentences, and provide expert feedback to help AI chatbots learn how to think and speak like humans.
Its rapid rise underscores the AI frenzy gripping venture-capital firms, where investors are eager to offer funding to founders at sky-high valuations because of fears of missing out on the next AI tech giant.
Menlo Park, Calif.-based venture firm Felicis, which led Mercor’s last funding round, will lead this one as well, some of the people said. Existing investors like Benchmark and General Catalyst will also participate.
Brendan Foody, 22 years old, Mercor’s chief executive, met his two other co-founders, Adarsh Hiremath and Surya Midha, while in high school in the Bay Area. Hiremath is the chief technology officer at Mercor, while Midha earlier this month said he was transitioning from chief operating officer to chairman of the board.
All three co-founders dropped out a few years into college to start Mercor. Soon after leaving school, they became Thiel Fellows, a program created by venture investor Peter Thiel that funds companies started by college dropouts.
Mercor initially started as a human resources and recruiting startup. It used AI to automate resume-screening, match candidates with the best roles, and screen candidates. It mostly focused on technical jobs, ranging from software engineers to mathematicians.
As part of this work, the company inadvertently compiled a mass network of specialized workers that the biggest AI companies have become eager to tap in order to train increasingly sophisticated chatbots and keep up in the AI race.
As AI companies sought to improve their models, Mercor quickly scaled up to hire contractors who could evaluate the quality of chatbot answers. To meet the needs of AI companies, Mercor began expanding its network of contractors to include lawyers, doctors, bankers and journalists. To support this strategy shift, Mercor recruited Uber’s former chief product officer, Sundeep Jain, as its first president in May.
The value of its network came into focus after Meta Platforms in June paid $14 billion for a 49% stake in Scale AI, one of the most well-known data-labeling startups. That pushed Scale’s valuation to an astounding $29 billion. As part of the deal, the company’s co-founder and CEO, Alexandr Wang, moved to the social-media giant to help lead its AI efforts.
Some of Scale’s customers and competitors raised concerns about the startup’s ability to remain neutral and protect customer data after Meta’s investment. As a result, the deal became a major boon for rival Mercor: Its revenue has quadrupled since Meta invested in Scale, according to some of the people familiar with its business.
The war for talent and business among data-labeling startups reached a fever pitch last month when Scale sued Mercor, accusing the company of allegedly stealing trade secrets. It also sued a former Scale employee, who had left to work for Mercor.
A spokeswoman for Mercor said the company would let the case play out through the legal process.
Doctors can now moonlight as data-labelers, earning $170 an hour and working a minimum of 20 hours a week in a six-week contract. Tasks include evaluating medical-related answers from AI and reviewing AI-generated medical studies, according to a contract listing posted by the company.
Another contract that Mercor is currently trying to fill: a progressive-politics specialist who, for $70 an hour, must identify answers from AI models that include “conservative framing inconsistent with liberal perspectives," read a listing posted by the company.
If a client pays Mercor $100 an hour for a data-labeler, Mercor will keep around 30% to 35% and pass the rest on to the contractor. The average hourly rate for contracts on Mercor is around $85 an hour.
Write to Angel Au-Yeung at angel.au-yeung@wsj.com
topics
