Mercor

Does Mercor have a moat?

Yes.

63/100 · Real moat

“Matching experts to tasks is a weekend build. Winning 5M vetted experts and OpenAI, Anthropic and Google as buyers takes years and a billion dollars.”

Code to copy

3 months

Moat to cross

2+ years

Market

Crowded market

Share on X

Moat depth

MoatlessShallowReal

Revenue not verified. Founders can list on TrustMRR to show it here.

01/The verdict

Two-sided marketplace and managed service. Vetted experts (doctors, lawyers, bankers, engineers) do AI training, evals and RL tasks. OpenAI, Anthropic, Google and Meta pay.

The software is simple: an AI interview, matching, task tools and payouts. The hard part is a supply of nearly 5M vetted experts, a delivery operation labs trust, and funding. Sacra puts gross revenue at about $2B annualized in June 2026, and Mercor has raised about $486M at a $10B valuation. Customer lock-in is weak, because labs buy from several vendors. About 91% of H1 2026 revenue came from AI labs, and the March 2026 breach hurt trust.

02/4 moats

Marketplace liquidity

Deep

~30 mo to cross

Mercor says it has nearly 5M experts. Its payouts are over $1.5M a day, and it works with former senior staff of banks, consultancies and law firms. Labs come for the experts, and experts come for the paid work.

To cross it: Recruit and vet thousands of credentialed experts at $85 to $200 an hour while having no lab contracts to pay them with.

Funding

Deep

Time won't do it

The $350M Series C (Felicis, Benchmark, General Catalyst, Robinhood Ventures) valued it at $10B, about $486M raised in total. Sacra reports about $2B annualized gross revenue in June 2026. Pay of $100K per referred data licensor shows how much it can spend.

To cross it: Raise hundreds of millions and fund expert payroll months before labs pay.

Brand & trust

Real

~24 mo to cross

OpenAI, Anthropic, Google DeepMind, Meta, Reflection AI and Thinking Machines Lab are named as customers. APEX benchmarks are in use. The LiteLLM breach (about 4TB claimed, Meta paused contracts, class actions filed) shows how fragile that trust is.

To cross it: Deliver quality at scale for a frontier lab with no track record, and pass its security reviews.

Real-world operations

Real

~18 mo to cross

Mercor runs the project management, quality control and contractor payouts across 100+ countries. It also runs AI interviewing and vetting. This is a human workforce doing paid work, not only software.

To cross it: Build the vetting, QA and global payments process, and hire delivery staff who can run lab-grade projects.

03/The gap

Easy to copy

The AI interview, resume parsing and expert matching are easy to rebuild with Claude Code.
The task tools and annotation interface are standard.
Labs multi-source vendors, so there is no real lock-in on any one project.
The APEX benchmark can be copied with enough experts.

How you'd go after it

Go vertical: one domain such as radiology, tax or chip design, with deeper experts and better quality control than a generalist.
Win on price or speed with a smaller operation, taking lab work that Mercor treats as low priority.
Pitch labs worried about the 2026 breach on security, with SOC 2, isolated data and a clean audit record.
Serve enterprises building their own agents with a managed expert pool.

04/Build against it?

Only with an edge

A direct clone loses to Mercor's roughly $2B annualized gross run rate and its supply of experts. Only try it with a deep niche, such as one expert domain you already have access to, or an existing lab relationship.

Already selling this:Surge AIScale AITuringHandshake AImicro1Labelbox

05/Receipts

Checked 2h ago by Claude Sonnet 5.5 on high effort, with web search. Confidence: medium.

Does yours?

Paste your website. You get your score in about a minute.