Mercor

What it is

Mercor is an AI-vetted talent marketplace: it interviews and scores candidates with AI agents, then matches them to paid work at customer companies. But its center of gravity has moved. Founded in 2023 by Brendan Foody, Adarsh Hiremath, and Surya Midha, Mercor started as a marketplace matching engineers to US startups and pivoted through 2024–2025 into the supply layer for AI post-training — recruiting PhDs, lawyers, bankers, physicians, and senior engineers to produce the expert data (RLHF, rubrics, evaluations, reinforcement-learning environments) that frontier labs use to train models. That pivot is the story. Mercor reached a $10B valuation on its October 2025 Series C and, as reported in July 2026, is in talks to raise $500M at roughly $20B; multiple outlets put its revenue above a $2B annualized run-rate by mid-2026, doubled from ~$1B in February. Its named customers are the frontier AI labs — OpenAI, Anthropic, Meta, Google — not corporate TA teams.

For a recruiting buyer, that reframe matters. Mercor is not a general hiring platform and it is not an ATS. It is an on-demand marketplace for AI-vetted specialist contractors, with an AI-interview layer bolted to the front — and its best-proven use case is supplying expert humans to AI labs, not filling your open reqs.

Why it shows up in Recruiting stacks

• AI conducts a structured first interview. Candidates complete an AI-driven interview that produces a depth-of-evaluation signal resume screening can’t reach — the mechanism that made Mercor’s vetting fast enough to scale to a large contractor pool.
• Pre-vetted specialist supply. 30,000+ active contractors across 45+ countries, weighted toward hard-to-source domain experts (ML research, quant/finance, medicine, law, senior engineering) rather than commodity labor.
• Speed on specialist needs. The match loop runs in days rather than the weeks-to-months of direct hiring or traditional staffing — the draw for time-sensitive contract work.
• The AI-data angle. If your org is building or fine-tuning models, Mercor is an expert-network / training-data vendor — evaluated against Scale AI and Surge AI — more than a recruiting tool.

Pricing

• Custom, marketplace-based. No published list price and no per-seat fee. Mercor bills clients a marked-up hourly rate for contractor time (cost-plus) and keeps the spread — reported gross margins around 35%.
• Contractor rate bands set the floor. Experts earn about $85/hr on average, with domain specialists (law, medicine, finance, senior engineering) reported up to ~$200/hr. Your client rate is that plus Mercor’s margin.
• Permanent placements. For full-time hires Mercor reportedly charges ~30% of first-year compensation — standard contingency-search economics.
• Real-world cost: an evaluation or data engagement is priced per expert-hour at volume (hundreds of hours), not per seat. Once margin is added, a single senior specialist routed through Mercor lands well above a $150–250/hr blended rate. Budget it as managed contractor spend, not SaaS.

Best for

• AI labs and AI-building teams sourcing domain experts for post-training data, model evaluation, and RL environments — Mercor’s strongest and best-documented use case
• Technical orgs needing on-demand specialist contractors (ML researchers, quant/finance, medical, security) faster than direct hiring or staffing delivers
• Teams that want an AI-vetting signal on a specialist contractor before committing hours

Watch-outs

• It is not a hiring platform for standard roles. Routing FTE engineering, GTM, or ops hiring through Mercor is using a contractor-supply marketplace for a job it isn’t built for. Guard: for full-time hiring, run an ATS (Greenhouse, Ashby) plus a sourcing tool (Juicebox, SeekOut); reach for Mercor only when the need is specialist contract supply or AI-training data.
• AI-interview scoring carries bias and validity risk. An AI evaluation that gates work has adverse-impact exposure. Guard: treat the score as one input, keep a human decision-maker, and run an adverse-impact check before it decides anything — especially under NYC LL 144, Colorado, and Illinois AI-hiring rules.
• Worker-classification and governance exposure. Mercor faces a California class action filed in October 2025 over contractor classification, and reported security and internal-fraud issues surfaced in April 2026. Guard: confirm data-handling terms, IP assignment, and classification indemnities in the MSA before routing sensitive work or PII through the platform.
• Customer-concentration and category risk. Mercor’s revenue leans hard on a handful of AI labs; if they build supply in-house or synthetic data displaces human labeling, demand could compress. Guard: don’t wire Mercor in as sole supply for anything business-critical — keep a second expert-network or staffing channel warm.

Alternatives

• Scale AI / Surge AI — the data-labeling and expert-network incumbents; pick these for high-volume annotation and established data-ops tooling. Mercor skews toward higher-end domain experts, and Scale’s mid-2025 destabilization after Meta’s stake is part of why Mercor captured displaced demand.
• micro1 — the fastest-growing AI-native entrant on the same vet-then-match model; weigh it when developer vetting and a global engineering marketplace are the core need.
• SeekOut / Juicebox — if what you actually need is to source and hire full-time employees, use a sourcing platform plus your ATS. Mercor does not replace that motion.

For standard engineering hiring at scale, Ashby or Greenhouse plus Juicebox sourcing delivers better long-term outcomes than routing full-time hiring through a contractor marketplace.