Eight months ago, Micro1 was doing $100 million in annualized revenue. Now it's at $500 million. If that growth curve makes you a little dizzy, you're paying attention.
According to TechCrunch, citing a person familiar with the company, Micro1's gross annual run rate jumped from $100 million to $500 million over the past eight months. That's not typo-level growth, that's the kind of number that makes venture capitalists forget how to blink.
The catch: Micro1 doesn't keep all of that. Like its competitors, it hires domain experts (think doctors, lawyers, scientists) on contract to label and annotate data, and pays them out of that gross figure. The startup retains roughly 60% to 70% of it, putting net annual run rate somewhere between $150 million and $200 million. Still nothing to sneeze at for a four-year-old company.
Micro1 didn't start as a data company. Per Forbes, it began as an AI-powered recruiting service, and at the start of 2025 it was clearing about $7 million a year, which in this industry now counts as a rounding error. The pivot happened when a large data-labeling firm approached Micro1 for help with hiring. Founder Ali Ansari told Forbes the moment was a wake-up call: "We were like, why is this company hiring hundreds of engineers in two weeks? We said, holy shit, we should really focus on this market."
That's a familiar origin story in this corner of tech. Mercor, one of Micro1's biggest rivals, followed almost the identical path: AI recruiting startup notices the real money is in feeding data to AI labs, pivots hard. According to TechCrunch, Mercor has already hit $2 billion in gross annualized revenue this summer. Handshake, another player in the space, reached $1 billion earlier this year. Micro1 is still the smaller fish here, but it's growing fast enough that "smaller" is relative.
The plain-English version: AI models get smarter by ingesting huge amounts of high-quality data, and a lot of that data needs a human to add context, grade outputs, or correct mistakes before it's useful for training. That's the job. Highly educated contractors, per Forbes, get paid $60 to $170 an hour to evaluate AI outputs, with specialists in fields like medicine or finance pulling in as much as $500 an hour.
This was not always considered a good business to be in. Adam Bain, co-founder of 01A Ventures and a Micro1 investor, told Forbes that data labeling was "underloved, underappreciated" until recently, and that some investors stayed away because they worried the entire category would become obsolete once AI reached human-level cognitive ability. Others were put off by the unglamorous reality of managing armies of short-term contractors.
That skepticism has evaporated. Forbes reports the sector has minted four new billionaires in recent months: the Mercor founders and Surge founder Edwin Chen. Ansari, per Forbes, is on track to potentially join them if Micro1 locks in a valuation being discussed by investors, given his roughly 42% stake in the company.
Two things are driving Micro1's numbers up, according to TechCrunch: contract sizes are getting bigger, and the company is leaning more into synthetic data that doesn't require a human in the loop at all, like generating automated descriptions of video content. Some of that data can be resold to multiple customers, which TechCrunch reports pushes gross margins on this "off-the-shelf" data as high as 80% to 90%. That's a very different margin profile than paying a contractor by the hour, and it's a preview of where the whole industry wants to go.
Ansari is also betting on data collection that doesn't come from white-collar experts at all. He's building a robotics pre-training dataset by paying generalists to record themselves interacting with everyday objects at home, per TechCrunch, and told Forbes he envisions a future where "nearly everybody can be an AI trainer, not just white collar experts."
Selling the same dataset to multiple buyers is great for margins, but it's also created a real fight. TechCrunch reports that critics argue distributing off-the-shelf training data to Chinese AI developers helps those labs close the gap with top US models. Ansari addressed this directly on X last month, according to TechCrunch, saying Micro1 doesn't sell to Chinese model makers and calling out unnamed rivals: "Some human data companies work with foreign adversaries. And the results show today in Kimi K3. We believe it's shameful to claim American AI dominance desires while selling millions worth of data to countries that we are in adversarial competition with."
That's a pointed accusation aimed squarely at competitors, without naming names, in an industry where everyone is chasing the same handful of frontier lab customers.
Micro1 raised its Series A at a $500 million valuation last September, per TechCrunch, and the outlet understands the company may have recently raised another round at a significantly higher valuation. Forbes reported in December that investors had floated a $2.5 billion valuation, up from the $500 million round just months earlier. Whatever the exact current number is, the trajectory only moves one direction.
Nobody should be shocked that data labeling turned into a gold rush. If AI's performance scales with the quality and volume of training data, then the picture painted by Ansari, who estimated to Forbes that major labs currently spend $15 billion a year on training and could exceed $100 billion within two years, makes the math obvious. TechCrunch notes some researchers think data spending could eventually rival compute spending, and if that happens, none of Micro1's numbers today will look particularly large in hindsight.
What's genuinely interesting is the resale model. An 80-90% margin business hiding inside a low-margin contractor operation is the kind of thing that turns a decent startup into a great one, assuming the ethics keep pace with the economics. And that's exactly where this gets messy. The moment your product can be sold to more than one buyer, "who's buying it" becomes as important as "how much they're paying." Micro1 says it draws a line at Chinese model makers. Whether every player in this space draws that same line, and whether anyone's actually checking, is the story worth watching next.
Micro1 went from a recruiting side-project to a $500 million run-rate business in under a year, riding a wave that's lifting Mercor, Handshake, and others right along with it. The demand is real, the money is real, and the fight over who this data ends up helping is just getting started.
---
Disclosure: This post is based on reporting from TechCrunch and Forbes. No compensation was received from Micro1, Mercor, Handshake, or any party mentioned in this article.
---
SEO Title: Micro1 Hits $500M Run Rate in AI Training Data Boom
Meta Description: Micro1's revenue jumped from $100M to $500M in eight months as AI labs race for training data. Here's what's driving the boom, and the controversy behind it.
Slug: micro1-500-million-ai-training-data-boom
Primary Keyword: Micro1 AI training data
Tags: AI training data, Micro1, data labeling startups, AI industry trends, Ali Ansari
Disclosure: No compensation received. Sourced from T