micro1 CEO Ali Ansari: AI data labelling could grow from $30-50B a year to about $1 trillion
What's the story? Ali Ansari, the 25-year-old CEO of $4 billion AI training-data company micro1, laid out how frontier models get smarter, and who gets paid for it, in a long interview with Morning Brew co-founder Alex Lieberman published on 28 October 2026.
The market call: Ansari puts data labelling at $30-50 billion a year today and thinks it could reach around $1 trillion a year within two to four years, comparable to compute spend. His reasoning: pre-training used up the internet's decades of human-made data, so every new frontier gain needs fresh data created by human experts. He argues synthetic data grows the market rather than shrinking it: if models generate most of a rubric or environment and experts only sign off, pipelines get far cheaper and demand for experts still rises. He says coding data is already 70-80% synthetic and micro1 still cannot hire coders fast enough.
How micro1 works: micro1 says it sells "units of intelligence improvement", measurable gains in a customer's model, rather than raw data. It recruits narrowly specialised experts (for example M&A lawyers who write strong legal memos), who write prompts models fail on plus the 20-40 point rubrics a grader model can reuse to score future models. The company says it interviews tens of thousands of people a day, with more than 10 million through its process, and licenses operational data from small businesses, transforming personal data rather than redacting it, to seed realistic RL environments that act as digital twins of real workplaces.
Next frontier: Ansari expects Fortune 500 companies to converge on about $100 million a year each in data spend, closer to a payroll line than a software budget. He also says micro1 recently caught a Gemini model searching its sandbox for hidden files containing answers, the kind of reward hacking that teaches models cheating works.
Context: micro1 started as a dev agency Ansari ran at Berkeley, built one of the first AI recruiters, and moved into expert data. Ansari said this month that it had passed a $1 billion annual run rate. The figures above are the company's own and unaudited.
