Mercor, the AI talent and data-labeling startup, has blown past $2 billion in gross annualized revenue, according to The Information. That’s a staggering number for a company that most people outside the AI training world have never heard of. The Information reports the milestone in an exclusive, underscoring just how fast money is flowing into the businesses that feed the world’s biggest AI models.
Here’s why that matters.
What Mercor actually does
Mercor sits in the plumbing of the AI boom. It connects human experts, think doctors, lawyers, coders, PhDs, with the labs training frontier models. Those experts label data, write evaluation questions, and grade model outputs so systems like the ones from OpenAI, Anthropic, and Google get smarter and safer.
It’s not a flashy consumer app. It’s a marketplace for human judgment. And right now, human judgment is one of the scarcest inputs in AI.
Why $2 billion is a big deal
Gross annualized revenue is a run-rate figure. It takes recent revenue and projects it across a full year. So this isn’t $2 billion banked in the bank. It’s the pace Mercor is now running at.
Still, the speed is what stands out. Mercor is a young company, and hitting a multi-billion-dollar run rate this quickly puts it in rare territory. For context, plenty of well-known software companies take a decade to reach numbers like that.
A few things worth noting:
- Gross vs. net. Gross revenue likely includes payments that flow through to the human experts doing the work. Mercor’s take-rate, the slice it actually keeps, is the number that really defines the business. The Information’s figure is the top-line, so read it with that in mind.
- Run rate can move fast. Run-rate numbers climb quickly in a boom and can soften just as fast if demand cools. It’s a snapshot, not a guarantee.
- This is a services-heavy model. Unlike pure software, a chunk of Mercor’s revenue goes out the door to contractors. Margins here look different from a typical SaaS company.
The bigger picture
The status quo used to be simple. AI labs spent their money on chips and compute. That’s still true. But a second spending wave has opened up, and it’s aimed squarely at data and human expertise.
As models get more capable, the cheap, low-skill labeling that powered early AI isn’t enough. Labs now want specialized knowledge. They’ll pay a radiologist to grade medical answers and a litigator to stress-test legal reasoning. Mercor’s growth is direct evidence of how much that expert data is now worth.
That puts Mercor in the same conversation as Scale AI and Surge AI, the other big names in the data-labeling market. Competition here is heating up, and the revenue numbers show why. Whoever controls the supply of high-quality human feedback holds real leverage over the labs racing to build better models.
What to watch next
A few signals will tell you where this goes:
- Fundraising and valuation. Numbers like these usually pull in new capital fast. Expect investor interest, and watch what valuation Mercor commands next.
- Customer concentration. If most of that revenue leans on one or two labs, the business is more fragile than the headline suggests. Diversified demand is the stronger story.
- Margins and take-rate. The next real test is how much of that $2 billion Mercor keeps, not just how much passes through.
- The durability question. Data-labeling demand is tied to how much the labs keep spending on training. If model training budgets shift, this market shifts with it.
For practitioners and founders, the read is clear. The AI gold rush isn’t just about who builds the smartest model. It’s about who supplies the picks and shovels, and human expertise has become one of the most valuable tools in the kit.
Mercor’s rise says the market for human-in-the-loop AI work is far bigger than most people assumed a year ago. You can find the full breakdown at the original report from The Information.