Mercor vs micro1: How They Compare for AI Training Work (2026)

The short answer#

Mercor and micro1 do similar work. Both recruit specialists, such as software engineers, lawyers, doctors, and finance professionals, vet them, and place them on paid projects where they write, grade, and correct material that AI labs use to train and test models. Both started as AI-assisted hiring products and moved into AI training data. Both are young: Forbes says Mercor was founded in 2023,¹ and TechCrunch described micro1 as four years old in August 2026, which implies a start around 2022.²

As of October 2026, Mercor is the larger and more scrutinized of the two. It raised $350 million at a $10 billion valuation in October 2025, was reported in July 2026 to be in early talks at $20 billion, and has faced a 2026 security incident and a set of contractor class actions.³⁴⁵⁶ Forbes reported in September 2026, citing people familiar, that micro1 raised more than $100 million at a $4 billion valuation; micro1 declined to comment. On October 9, 2026 micro1 announced a $1 billion program to buy and license operational data from companies.⁷⁸

For a business owner, the more useful question is not which company is bigger. It is whether either one is a realistic buyer or channel for a company's own data. Both now say yes, with very different levels of published detail, and that is covered in the sections below.

At a glance#

What each company does#

Mercor#

Mercor describes itself as "organizing human intelligence to power the AI economy."⁹ Its core business is supplying experts, such as physicians, lawyers, engineers, and consultants, who evaluate and improve AI models. It also builds benchmarks that measure AI performance on professional work, branded APEX, and its site lists partner benchmarks with companies including Harvey, Ramp, Cognition, and Sierra.⁹ The site also lists human data and evaluation services, an Enterprise AI offering, and a data monetization program for businesses.⁹

The company began as an AI interviewing and hiring tool. TechCrunch described the pivot in October 2025: Mercor started off as an AI-driven hiring platform and quickly pivoted to providing companies with specialized experts for AI model training, and it was adding more software infrastructure for reinforcement learning.³

In July 2026 Mercor announced it was buying Deeptune, a company that helps train AI agents.⁴

micro1#

micro1 describes itself as a data lab that trains frontier models and evaluates AI agents, building "infrastructure for advancing intelligence through expert human data, real-world training environments, and contextual evaluations."¹⁰ Its site lists three product lines: Realm (reinforcement learning environments and frontier evaluations for AI labs), Cortex (an evaluation and visibility platform for companies running AI agents in production), and a robotics data offering built on expert demonstrations.¹⁰ It also lists benchmarks, including LegalReasonBench and Ivo-micro1 ContractBench, a contract review benchmark.¹⁰

micro1 grew out of a software agency that founder Ali Ansari started while at Berkeley. Because he had to vet engineers for that agency, he built an AI screener, which became the basis for micro1, according to a Stanford Daily interview.¹⁷ TechCrunch described micro1 as having begun as an AI recruiting startup, and said its AI recruiter Zara interviews and vets candidates who apply to work as contractors.¹²²

History and founders#

Mercor. Brendan Foody, Adarsh Hiremath, and Surya Midha are Bay Area high school friends who competed together on their debate team. They founded Mercor in 2023 and are Thiel Fellows.¹ Foody is CEO, Midha is board chairman, and Hiremath is CTO.¹ After the October 2025 round, Forbes described the three, then 22, as the youngest self-made tech billionaires.¹ Sources differ on some biographical details such as which universities the founders attended, so this page leaves them out.

micro1. Ali Ansari is the founder and CEO. Forbes, in September 2026, described him as 25 years old.⁷ According to a Stanford Daily interview, he started a college textbook resale company in middle school and an online math tutoring business in high school, and he attends Stanford classes part time while micro1 works with a research lab led by associate professor Stefano Ermon.¹⁷ The founding year is not stated by micro1 in the sources reviewed; TechCrunch called the company a three-year-old startup in September 2025 and a four-year-old startup in August 2026, which implies about 2022.¹²²

Funding and valuation#

Mercor#

• Series B: $100 million at a $2 billion valuation, led by Felicis Ventures.³
• Series C (October 2025): $350 million at a $10 billion valuation, confirmed to TechCrunch on October 27, 2025. Felicis led; Benchmark and General Catalyst, existing investors, took part, and Robinhood Ventures joined as a new investor.³
• Reported talks (July 2026): TechCrunch reported on July 9, 2026, citing Bloomberg, that Mercor was in early talks to raise at a $20 billion valuation, double the Series C valuation, and had told investors it already had a term sheet at that price. The talks were described as early stage.⁴

I could not find an official Mercor announcement confirming that a $20 billion round has closed as of October 20, 2026, so this page treats it as reported talks, not a completed financing.

micro1#

• Series A (September 2025): $35 million at a $500 million valuation, led by 01 Advisors, per TechCrunch and the Stanford Daily.¹²¹⁷
• Round reported September 2026: More than $100 million at a $4 billion valuation, according to a Forbes report published September 22, 2026 and citing two people familiar with the deal. micro1 declined to comment. The valuation is eight times the September 2025 mark.⁷ Forbes said two frontier AI labs and two xAI cofounders took part.⁷

I could not find an official announcement of the round, so this page treats it as reported, not confirmed. No primary source I found assigns it a series letter or names a lead investor, so this page does not.

Revenue and scale#

Mercor's CEO Brendan Foody said on X in July 2026 that its annualized revenue run rate had crossed $2 billion, which TechCrunch described as a doubling over four months.⁴ A TechCrunch piece from October 2025 said most of Mercor's revenue came from a few AI labs, and Foody said the company remained profitable.¹¹

For micro1, TechCrunch reported in September 2025 that the CEO said it was generating $50 million in annual recurring revenue, up from $7 million at the start of 2025.¹² In August 2026 TechCrunch reported, citing a person familiar with the company, that micro1's gross annual run rate had grown from $100 million to $500 million over eight months, and that micro1 keeps roughly 60 to 70 percent of the gross figure after paying experts, or a net run rate of about $150 million to $200 million.² Forbes repeated the $500 million figure in September 2026.⁷ The difference between a gross run rate and recognized revenue matters here: both companies bill a client for expert time and pass most of it to the expert.

The practical read: both are growing quickly from a small base, Mercor is clearly larger, and neither has published audited financials.

Business model#

Both companies are two-sided marketplaces with a service layer on top.

Mercor charges clients an hourly fee on top of the contractor's pay. TechCrunch described this in October 2025 as an hourly finder's fee and matching rate.³ It then adds products around that core: benchmarks, Enterprise AI, and a data monetization program.⁹

micro1 also sells vetted expert time and managed data projects, and has added software products (Realm and Cortex) and a robotics data line.¹⁰ Its pitch to labs is expert-generated data, environments, and evaluations; the AI interviewer Zara interviews and vets candidates who apply to work as contractors.¹²

A TechCrunch interview with Mercor's CEO in October 2025 explained a strategic premise that applies to this whole market. Brendan Foody said labs hire former employees of banks, consulting firms, and law firms through Mercor instead of signing data contracts with those firms, because their customers "don't want to give them data to automate large portions of their value chains."¹¹ That is a useful caution for anyone thinking about licensing data: some companies prefer not to sell, and labs have routes around that.

Who they serve#

Both serve frontier AI labs first.

Mercor. TechCrunch named OpenAI, Anthropic, and Meta as customers in October 2025, citing Mercor's CEO as its source for the claim that most revenue comes from a few labs.¹¹ Mercor's website positions it for enterprises as well, with partner benchmarks and an Enterprise AI offering.⁹

micro1. Forbes listed frontier AI labs, Microsoft, Amazon, and robotics companies like 1X as customers, and TechCrunch reported in September 2025 that micro1 had worked with Microsoft and several Fortune 100 companies.⁷¹² micro1 names no customers on its own homepage in the version I read.¹⁰ Because these customer lists come from press reports and not from customer announcements, treat them as reported, not confirmed.

How they source data and people#

Mercor#

Mercor lists a network of more than 30,000 experts on its site, with an average contracted expert rate of $107 per hour and daily payouts above $4 million (page as reviewed in October 2026).⁹ In October 2025 TechCrunch reported more than 30,000 experts earning above $85 per hour on average, and more than $1.5 million per day paid to contractors.³ In the same period Mercor said it paid up to $200 per hour for some expert work.¹¹ The site shows individual listings with rates, such as $350 to $450 per hour for an oncologist and $60 to $70 per hour for a sales and marketing expert.⁹

TechCrunch also reported that Mercor tells contractors not to upload documents from former workplaces, and that Foody acknowledged "there are things that happen" at its scale.¹¹ That is relevant to any company thinking about data provenance.

micro1#

micro1 uses an AI recruiter named Zara that interviews and vets candidates who apply to work as one of its contractors.¹² The company keeps roughly 60 to 70 percent of its gross billings after paying experts, per TechCrunch.² On pay, micro1's own jobs page, as reviewed in October 2026, lists no pay amounts, and I could not find an official micro1 page stating an average rate.¹⁸ Rates, vetting timelines, and project volume for micro1 contractors are therefore not documented in the sources reviewed.

Company data programs#

This section matters most to a company that holds its own business data. As of October 2026, both companies run programs to acquire or license data from businesses. Both programs are recent, and the sources are mostly the companies' own pages.

Mercor's enterprise data licensing program#

Mercor's data page invites businesses in eight industries (legal, healthcare, finance and insurance, professional services, manufacturing, retail and e-commerce, real estate, and logistics and supply chain) to check whether they qualify.¹⁹ What it describes:

• Data: workflow data from connected tools such as Slack, Microsoft Teams, Gmail, Outlook, Confluence, Salesforce, GitHub, Google Drive, and Notion, with the page citing more than 100 supported tools.¹⁹
• Process: four steps: scope which tools to share, connect accounts, Mercor extracts and de-identifies the data into an asset, then payout.¹⁹
• Control and security: "No data leaves your systems until you approve it," with audited and logged access, encryption in transit, SOC 2 Type II, and keys managed in AWS KMS.¹⁹
• Compensation: based on "data volume, the tools you connect, and the depth of your data," with more tools generally meaning higher payout. No amounts are given. There are no upfront costs or fees deducted from the payment.¹⁹
• Buyers: described only as AI labs. The page says no lab buyer or third-party vendor sees raw data, and that extraction and anonymization are automated without manual review.¹⁹
• Timing: payment by wire after delivery of the anonymized dataset to the buyer, typically 2 to 4 weeks after the signed agreement. The process section describes payout "as soon as processing is complete," so the two statements are not identical.¹⁹

What the page does not address is data ownership after licensing, license terms, exclusivity, or how buyers are chosen.¹⁹ Mercor's homepage separately states that its anonymization masks more than 60 categories of sensitive identifiers.⁹

micro1's enterprise data partnerships#

On October 9, 2026 micro1 announced a commitment of $1 billion over 12 months to acquire and license operational data from companies, with capital from Citi and Hercules Capital.⁸²⁰ CEO Ali Ansari was quoted saying the initiative "gives businesses a new revenue stream from the operational intelligence they have built over years."⁸ micro1 says it builds reinforcement learning environments from de-identified company data, so AI agents can learn business workflows.⁸

micro1's data page adds detail:²⁰

• Data wanted: standard operating procedures, knowledge bases, internal documentation, CRM data, project histories, QA processes, decision-making patterns, and human feedback on AI outputs, across software, finance and accounting, customer support, sales and CRM, legal and contracts, and operations and logistics.
• Payment tiers: $100,000 and up for qualified enterprise data partnerships, $500,000 and up for large-scale operational datasets or participation across multiple teams, and $1 million and up for highly unique proprietary operational data. Compensation depends on quality, uniqueness, and value.
• Eligibility: operationally mature companies with 30 or more employees, established documentation, and English-language materials. US companies are the priority, then other Western markets, with strongest demand in the US, UK, and Canada.
• Process: an assessment, an evaluation, a discovery call covering data and privacy requirements, then an agreement.
• Privacy and ownership: personal information removed where relevant, isolated pipelines, optional synthetic rewrites, agreed retention periods, sample review before use, and deletion requests honored per the engagement terms. The page says companies retain ownership of their underlying data.

The Next Web's coverage said the article did not state who buys the output, only that micro1 helps AI research companies and enterprises train frontier models and evaluate agents.²¹ Neither micro1 nor Mercor, in the pages reviewed, names a lab that has paid for licensed enterprise data.

The Spirit Airlines bankruptcy data auction#

A real-world test of this market appeared in the Spirit Airlines bankruptcy in 2026. Both companies were involved as bidders for Spirit's internal records, according to press reports.

Reuters reported the price of Google's winning $10 million bid on August 17, as summarized by The Next Web. Mercor had countered at $5.2 million, then offered $7 million if it could take the raw data first and anonymize it itself, and remained the backup buyer at $7.5 million.²² Weeks after Google won the auction, micro1 offered $12.5 million for the data, according to CSO Online.¹⁶ CSO described the material as about 600 million email and chat records from 17,000 employees; other outlets have given different counts, so treat the volume figures as approximate.¹⁶

The sale drew objections. The Association of Flight Attendants-CWA objected, and unions sued to block the sale, saying former employees remain owed accrued vacation, sick leave, and other compensation.¹⁶ Under the sale agreement, third parties that de-identify the data must be acceptable to or designated by Google, which is solely responsible for the cost, per The Next Web.²² I did not find a report of a final court ruling on the sale or on micro1's late bid; CSO Online said the data's fate was still undecided.¹⁶

The episode shows two things. Both firms want business records at scale, and employee data in particular raises consent and labor questions that can stall a deal.

Controversies and legal issues#

Mercor#

Data breach and class actions. In late March 2026, Mercor was affected by a supply-chain attack on LiteLLM, an open-source interface used for several AI platforms. Mercor told TechCrunch it was one of thousands of companies affected, and the extortion group Lapsus$ claimed responsibility on its leak site; Mercor did not confirm whether the incident was connected to those claims.⁵ In April, TechCrunch reported that a hacker group had claimed to have obtained 4TB of stolen data.¹⁴ In a post dated June 25, 2026, Mercor said its investigation with outside forensics experts was complete, that only a very limited subset of its experts had sensitive information affected, and that it had seen no evidence of fraudulent use.²³

Several lawsuits followed.⁶ Plaintiffs, described as independent contractors who work as experts on the platform, filed class actions starting April 1, 2026 in the US District Court for the Northern District of California. One case is Esson v. Mercor.io Corporation. The suits allege negligence, unjust enrichment, breach of implied contract, breach of privacy, and violation of California's Unfair Competition Law.⁶

A separate complaint, Ananthula v. Mercor.io Corporation, No. 3:26-cv-03362, filed April 21, 2026 by Hausfeld LLP and another firm, alleges exposure of recorded AI interviews, facial biometric data, identity verification materials, and tax and banking information, and alleges that a LiteLLM compromise led to the exfiltration of about four terabytes. It also challenges Mercor's AI-driven scoring tools under the Fair Credit Reporting Act. These are allegations, and the Hausfeld page reports no ruling or settlement.¹³ I did not find a published detailed response from Mercor to these allegations; HR Dive reported that Mercor did not immediately respond to a request for comment.⁶

Client pause. TechCrunch reported in April 2026, citing sources who spoke to Wired, that Meta had paused its contracts with Mercor indefinitely; Mercor declined to comment.¹⁴

Scale AI suit. In September 2025 Scale AI sued Mercor and former Scale employee Eugene Ling in federal court, alleging trade secret misappropriation by Mercor and breach of contract by Ling. Scale alleged Ling took more than 100 confidential documents concerning Scale's customer strategies and tried to pitch Mercor to one of Scale's largest customers before leaving. Mercor co-founder Surya Midha said Mercor has no interest in Scale's trade secrets, that Ling had old documents in a personal Google Drive, and that Mercor never accessed the files.¹⁵ I found no report of a resolution.

Contractor pay dispute. After a Mercor project reviewing video content for Instagram and Facebook was canceled in November 2025, contractors told Forbes, as reported by PEOPLE, that they were offered similar work at a lower rate, $16 per hour versus $21 per hour. Mercor told Forbes the claims were inaccurate, and a company email framed the change as aimed at "greater earning stability and consistent access to work."²⁴

I did not find a misclassification lawsuit against Mercor in the sources opened for this page.

micro1#

I did not find lawsuits, regulatory actions, or a security incident involving micro1 in the sources opened for this page. The Spirit bid drew public objections from employees' representatives to the sale of the records generally, not to micro1 specifically.¹⁶

Strengths and weaknesses#

Mercor strengths - Larger scale and a longer public record, with named lab customers and a profitability claim from its CEO.¹¹ - A broader product set: benchmarks, RL environments, enterprise AI, and a data licensing program with stated security controls.⁹¹⁹ - Higher stated average expert pay on its site ($107 per hour), with specialist listings above $350 per hour.⁹

Mercor weaknesses - A 2026 security incident with multiple class actions and a reported Meta pause.⁶¹³¹⁴ - Heavy customer concentration in a few labs.¹¹ - No published payout amounts for its data licensing program.¹⁹ - Contractor complaints about project cancellations and pay changes.²⁴

micro1 strengths - Rapid reported growth, from about $7 million ARR in early 2025 to a reported $500 million gross annual run rate.¹²² - Published payment tiers and eligibility rules for enterprise data.²⁰ - Financing for its data program from named institutions (Citi and Hercules Capital).⁸

micro1 weaknesses - Revenue and round details rest on people familiar with the company, and I found no official announcement of the round.⁷² - The enterprise data program is days old, with no named buyers or completed deals in the sources reviewed.⁸²¹ - No average contractor pay published in the sources reviewed.¹⁸

Who each suits#

Experts and contractors. Mercor's site lists an average contracted rate and a larger network, while micro1 publishes no average that I could find and screens applicants with an AI interviewer. Both can pause projects or end them early, and both treat workers as contractors. Anyone applying should read the contract terms, check payment timing, and not rely on advertised rates as expected earnings.

AI labs and enterprises buying human data. Mercor has the larger, more established network and the benchmark and RL environment work, but a lab or company weighing it should factor in the 2026 security incident and the Meta pause. micro1 offers expert data plus evaluation and agent monitoring products, with less public history.

Companies with data to license. Both programs are early. The comparison is below.

What this means for data sellers#

Do not let one buyer set your price. Any single buyer's published range, program page or first offer is that buyer's number, not the market's. The sensible move is to get competing offers on the same data, compare scope, exclusivity, indemnity and payment timing side by side, and have someone independent look at the terms. Our guide to selling direct vs using a broker explains how, and if a buyer is already talking to you, we review offers free.

For more on what business data AI companies want, see data categories, the buyer landscape, and how selling works.

Facts in this page were checked on October 20, 2026. Funding talks, lawsuits, and programs described here are changing, and several figures are self-reported.