Micro1 Reports Exceeding $100 Million in Annual Recurring Revenue for AI Training Services

Micro1, a three-year-old startup specializing in human expertise for artificial intelligence (AI) training data, has announced it surpassed $100 million in annual recurring revenue (ARR), according to founder and CEO Ali Ansari. This figure marks a rapid escalation from approximately $7 million in ARR at the beginning of the year and more than doubles the revenue reported in September, when the company secured a $35 million Series A funding round at a $500 million valuation. The company's growth is attributed to increasing demand from leading AI laboratories, including Microsoft, and Fortune 100 enterprises seeking to refine large language models through post-training and reinforcement learning. Ansari projects the market for top-tier human data, currently estimated at $10-15 billion, could approach $100 billion within the next two years. Micro1's expansion occurs amidst a dynamic competitive landscape, which includes companies such as Mercor, with reported ARR exceeding $450 million, and Surge, which reported $1.2 billion in 2024. The sector has also seen shifts, notably after OpenAI and Google DeepMind reportedly adjusted their engagements with Scale AI, following Meta's $14 billion investment in the latter and the departure of Scale's CEO. Ansari states Micro1's ability to quickly recruit and evaluate domain experts has been a key driver. The company, which evolved from an AI recruiter, utilizes its internal tools to vet applicants for expert roles. Micro1 is also positioning for future growth by developing services for non-AI-native Fortune 1000 enterprises building internal AI agents for workflows and support operations, which require systematic human evaluation of AI performance. Additionally, Micro1 is entering the robotics pre-training sector, focusing on generating high-quality human demonstrations of physical tasks. Ansari indicated the company is constructing a large robotics pre-training dataset by collecting demonstrations from generalists. He stated, "We anticipate that a good portion of the product budgets at non-AI-native enterprises will go towards evals and human data, moving from 0% to at least 25% of product budgets." He added, "We're also helping robotics labs create robotics data; these two areas will account for a massive share of that $100 billion-a-year market." The company manages thousands of experts across various domains, some of whom earn nearly $100 per hour. Ansari noted, "There are Harvard professors and Stanford PhDs spending half their week training AI through Micro1." He further emphasized the expanding volume and range of expert roles beyond traditional technical fields.