Micro1, a competitor to Scale AI, raises funds at $500M valuation

Micro1, a three-year-old startup that helps AI companies find and negociate quality contractors for information labeling and training, has raised a $35 cardinal Series A backing circular that values the institution astatine $500 million. The circular was led by O1 Advisors, a task superior steadfast co-founded by Dick Costolo and Adam Bain, the erstwhile CEO and COO of Twitter.

The startup is 1 of galore companies looking to capable the spread successful the information marketplace created by caller changes involving Scale AI. After Meta invested $14 billion successful Scale AI and hired its CEO, AI labs including OpenAI and Google said they planned to cut ties with the startup, presumably implicit concerns that their probe could extremity up successful Meta’s hands. (Scale AI denies that it shares confidential accusation with Meta arsenic portion of its partnership).

However, AI labs inactive request these information services, and startups similar Micro1 purpose to prime up the slack.

Micro1 CEO Ali Ansari — who is conscionable 24 years aged — tells TechCrunch that his institution has been moving with starring AI labs, including Microsoft, arsenic good arsenic respective Fortune 100 companies. Ansari said Micro1 is present generating $50 cardinal successful yearly recurring gross (ARR), up from $7 cardinal astatine the commencement of 2025.

That’s inactive a acold outcry from larger competitors similar Mercor, which is generating more than $450 million successful ARR, and Surge, which reportedly brought successful $1.2 cardinal successful 2024. However, Micro1’s maturation and adoption among AI labs seems to beryllium increasing astatine a steadfast rate.

As portion of the caller funding, Micro1 is besides adding Bain to its committee of directors, alongside Joshua Browder, laminitis and CEO of the AI ineligible adjunct DoNotPay.

“Really the lone mode models are present learning is done nett caller quality data. Micro1 is astatine the halfway of providing that information to each frontier labs, portion moving astatine speeds I’ve ne'er seen before,” Bain said successful a connection to TechCrunch.

Reuters antecedently reported details of Micro1’s fundraising efforts.

All these companies — Micro1, Surge, Mercor, and Scale AI — proviso AI labs with entree to a ample basal of quality contractors who tin statement and make information for AI training. It’s go a important work that companies similar OpenAI, Anthropic, Meta, and Google request to physique cutting borderline AI models.

Scale AI was archetypal to predominate this space, with the archetypal penetration that it could wage comparatively small for low-skilled contractors astir the satellite to assistance statement information for AI exemplary training. However, Ansari says that the demands of AI labs person shifted successful caller years, and that companies present request high-quality information labeling from domain experts — specified arsenic elder bundle engineers, doctors, and nonrecreational writers — to amended their AI models. The hard portion became recruiting these types of folks.

This led Micro1 to physique its AI recruiter, Zara, which interviews and vets candidates who use to enactment arsenic 1 of company’s contractors, oregon arsenic Ansari calls them, experts. Micro1 says Zara has recruited thousands of experts — including professors from Stanford and Harvard — and that the institution plans to adhd hundreds much each week.

The marketplace for AI grooming information appears to beryllium changing yet again. Now, galore AI labs are funny successful moving with startups to make “environments” — virtual workspaces that tin beryllium utilized to bid AI agents connected simulated tasks. Ansari says Micro1 is gathering caller offerings successful the environments abstraction to conscionable this demand.

Luckily for startups similar Micro1, AI labs look to beryllium moving with aggregate grooming information providers. The quality of the concern is specified that it’s hard for immoderate 1 institution to grip each of 1 AI lab’s information needs. That means there’s plentifulness of concern to spell around, astatine least, for now.