Interview with Sabrina Halper and Ali Ansari, micro1 Annual Revenue Exceeds $1 Billion
Sabrina Halper interviewed Ali Ansari, a 25-year-old independent founder and CEO, whose company micro1 has just surpassed $1 billion in annual recurring revenue.
During the interview, Ali shared that he moved to the U.S. from Iran at the age of nine without knowing English, starting by reselling textbooks and eventually building one of the fastest-growing AI data companies today.
Topics discussed included: whether experts are training their replacements; how to teach AI subjective judgment; the need for new data to align AI since knowledge is not static; bottlenecks between robots and real-world data; the high risks of selling training data to China; and whether the AI economy can support infrastructure investments.
The company has transformed from an AI recruitment tool to providing expert human training data, RLHF feedback, and evaluation for cutting-edge AI labs; it previously faced near bankruptcy due to client losses but has since rebuilt and maintained profitability, with a core team of about 180 people, fully remote, covering over 14 countries.
Before surpassing $1 billion in annual revenue, the company completed a Series B funding round in the summer, exceeding $500 million, with a valuation of $4 billion, led by Dream Ventures, with participation from Microsoft and a $25 million employee stock buyback.
The interview also discussed the limitations of internet data as training data and the need for continuous new data to maintain alignment between models and human judgment.
Source: Public Information
ABAB AI Insight
Ali Ansari, as an independent founder, started from an Iranian immigrant background, initially gaining experience through eBay and textbook reselling, later studying computer science at Berkeley and Stanford. He first developed AI recruitment tools and then shifted the business focus to expert human data around 2025, providing training, RLHF, and evaluation services for cutting-edge labs; the company rebuilt after losing major revenue and announced in October 2026 that it surpassed $1 billion in annual recurring revenue while remaining profitable.
On the capital path, the company charges AI labs through an expert network and structured data environment, with Series B led by Dream Ventures and participation from Microsoft, achieving a valuation of $4 billion and providing employee liquidity; resources are concentrated on recruiting field experts and generating real-world and subjective judgment data from non-internet sources, rather than relying on publicly available web data.
This is similar to the competition among human data providers like Scale AI, Mercor, and Surge in the post-training phase of cutting-edge models, or the path of early data labeling companies transitioning to expert-level RLHF; currently, micro1 is in the phase of migrating AI training data from general labeling to domain experts and real-world environments, with robot data and aligning subjective judgment becoming the next bottleneck.
Essentially, this represents a restructuring of the industrial chain: training data is shifting from internet scraping and general labeling to expert-driven structured environments and continuous new data supply, making human judgment a core input for model alignment and capability enhancement; infrastructure investment relies on the continuous expansion of data quality and expert networks, while selling data to specific countries involves strategic and compliance constraints, making data supply a key link in AI capability competition.
ABAB News · Cognitive Law
Expert data becomes a continuous input for model alignment
When annual revenue crosses a billion, data quality determines infrastructure returns
After internet data becomes ineffective, real-world and subjective judgment become bottlenecks