Micro1 Hits $4B Valuation, Up Eightfold in a Year, as AI Labs Pay Up for Expert Data

Twelve months ago, Micro1 was valued at $500 million. Today, the San Francisco-based AI training data startup run by 25-year-old Ali Ansari is worth $4 billion, after raising a fresh round of more than $100 million, Forbes reported on September 22. The eightfold jump is one of the steepest valuation climbs in a sector that has quietly become one of the most lucrative layers of the AI economy: selling the human expertise that frontier models still cannot generate on their own.

The investor list says as much as the price tag. According to two people familiar with the deal cited by Forbes, participants include two frontier AI labs and two of xAI's cofounders. Micro1 counts frontier labs, Microsoft, Amazon and humanoid robotics company 1X among its customers, and several of those customers are now also shareholders. Micro1 declined to comment on the round.

From recruiting tool to data lab

Micro1 did not start as a data company. Ansari built Zara, an AI interviewer, as an undergraduate at UC Berkeley in 2022, and the business initially used it to vet engineers for clients. The pivot came when a data-labeling customer asked Micro1 to hire 600 engineers in three weeks, according to Inc. Ansari concluded that rather than recruiting experts for other data vendors, Micro1 could supply them to AI labs directly.

The results have been dramatic. Micro1 was generating roughly $7 million in annual recurring revenue at the start of 2025. By December 2025, eight months into the pivot, it had crossed $100 million in annualized revenue, and in August TechCrunch reported it had reached a $500 million gross annual run rate. That headline figure deserves a caveat: like peers that pay contract experts, Micro1 keeps roughly 60 to 70 percent of gross billings, which TechCrunch said puts its net run rate between $150 million and $200 million.

Zara remains the engine. A few thousand candidates pass through the AI interviewer each day and about 20 percent are accepted, Ansari told Inc., with more than five million signups since launch. Experts, from tax preparers to international law specialists, are typically paid $50 to $200 an hour. Ansari has framed the opportunity in expansive terms. "What we sell at Micro1 is improved intelligence, and there's going to be effectively infinite demand for that in the long run," he told Inc. "We're just getting started."

The company is also pushing beyond expert annotation. Forbes reported that Micro1 is building reinforcement learning gyms, simulated workplaces built from real enterprise data where AI agents can practice, and that it submitted a late $12.5 million bid for a portion of bankrupt Spirit Airlines' operational data, challenging Google's winning $10 million bid. The bankruptcy court has not yet ruled. Micro1 is also paying people $50 to $90 an hour to annotate videos of robots performing tasks, and TechCrunch reported it is increasingly producing synthetic and off-the-shelf datasets that can carry gross margins as high as 80 to 90 percent.

A crowded, well-funded field

Micro1's round landed on the same day Snorkel AI announced it had raised $350 million at a $3.5 billion valuation, nearly triple its May 2025 mark, with annualized revenue above $350 million. The two deals underline how much capital is chasing the same thesis. Mercor, which also began as an AI recruiting startup, was valued at $10 billion in October 2025 and hit $2 billion in gross annualized revenue this summer, per TechCrunch. Handshake reached $1 billion earlier this year. Surge AI, bootstrapped and past $1 billion in revenue, is reportedly seeking a $25 billion valuation in its first raise in five years.

The market was reshaped in June 2025, when Meta took a 49 percent stake in Scale AI for $14.3 billion and several labs paused or ended contracts with Scale. That opened the door for challengers pitching higher-skilled, domain-expert data. "We're in a phase where model gains increasingly come from better targets, not just more data," Rumman Chowdhury, founder and CEO of Humane Intelligence, told Inc., describing companies like Micro1 and Mercor as building a "human intelligence layer" for AI.

Why It Matters

The Micro1 round is a clear signal that frontier labs view specialized human data as a strategic input on par with compute, not a commodity cost. When labs themselves invest in their data suppliers, they are buying both capacity and a degree of supply security in a market where the best experts are scarce and contested.

It also shows how fast valuations are moving on revenue that is partly pass-through. Investors are paying $4 billion for a business whose net revenue is a fraction of its gross billings, betting that higher-margin synthetic data, RL environments and evaluation products will widen margins over time. There is a geopolitical dimension too. Ansari has publicly criticized rivals that sell data to Chinese model makers, writing on X that "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." As scrutiny of data exports grows, where a vendor sells may matter as much as what it sells.

What to Watch

The first test is whether Micro1 can shift its mix toward higher-margin products. Ansari has said its agent evaluation product, Cortex, could surpass the lab business within two to four years, a bet that enterprises deploying AI agents will pay for continuous evaluation. The Spirit Airlines data ruling will show whether startups can outbid Big Tech for proprietary corporate records. And with Snorkel, Mercor and Surge all flush with capital, expect intensifying competition for expert contractors and customer contracts, along with the possibility that labs build more of this capability in-house. For now, the frontier labs writing checks to Micro1 are signaling that they would rather own a piece of the supply chain than compete for it.