AI Data Startup Micro1 Reaches $500M Gross Run Rate

Micro1 has reportedly reached a $500 million gross annual run rate, marking another sharp jump for the AI data startup as frontier model developers spend heavily on specialized human-generated training data.

The milestone, reported on August 20, puts Micro1 among the fastest-growing companies in the increasingly competitive AI training-data market.

The number is particularly striking given where the company stood little more than a year ago. Micro1 was generating roughly $7 million in annual recurring revenue at the beginning of 2025. By September, that figure had climbed to around $50 million, before crossing $100 million ARR in December.

From $7M to hundreds of millions

Micro1's financial trajectory has accelerated rapidly.

Reuters reported in July 2025 that the company had crossed $50 million in annualized revenue, up from approximately $10 million earlier that year. At the time, Micro1 was targeting $100 million in annualized revenue by the end of September.

By December, CEO Ali Ansari said Micro1 had already exceeded $100 million ARR. Sacra later estimated annualized revenue had reached $125 million by the end of 2025 and approximately $300 million by April 2026.

The newly reported $500 million gross run rate suggests the amount of business flowing through Micro1's platform has continued to expand sharply during 2026.

That does not necessarily mean Micro1 is generating $500 million in net annual revenue. As a marketplace-style business that pays experts and earns a margin on their work, gross run rate can include money ultimately paid to contractors, making it different from ARR or recognized net revenue.

AI labs are paying more for expertise

Micro1's growth reflects a broader change in how advanced AI systems are trained.

Early data-labeling businesses relied heavily on large pools of relatively inexpensive workers performing basic annotation. Frontier labs increasingly need specialists capable of producing and evaluating difficult material in areas such as software engineering, mathematics, medicine, law and physics.

Micro1 built its business around that shift. Its AI recruiting system screens applicants and connects AI developers with vetted domain experts for reinforcement learning, model evaluations and other post-training work.

Sacra says only about 1% of applicants make it through Micro1's screening process, while its infrastructure can process thousands of interviews. The company also handles payroll and compliance across more than 90 countries.

Micro1 has previously said its customers include major AI labs, Microsoft and Fortune 100 companies.

A $35M round preceded the surge

The startup raised $35 million in Series A funding in September 2025, led by 01 Advisors, at a $500 million valuation. Former Twitter COO Adam Bain joined its board as part of the financing.

At the time of that round, Micro1 was operating at only about a $50 million annual revenue run rate.

The contrast is notable: its latest reported $500 million gross run rate is 10 times the $50 million revenue run-rate figure disclosed alongside its Series A less than a year ago, although the different accounting measures mean the two figures are not directly comparable.

Micro1's last publicly announced valuation also happens to be $500 million, the same headline number as its newly reported gross run rate. That valuation, however, dates to September 2025 and should not be interpreted as the company's current market value.

Scale AI's shake-up opened the market

Micro1's expansion has also coincided with disruption at one of the industry's biggest players.

Meta invested roughly $14 billion in Scale AI in 2025 and brought Scale founder Alexandr Wang into Meta's AI operation. OpenAI and Google subsequently planned to reduce their reliance on Scale amid concerns about working with a supplier closely connected to a rival AI developer.

That created an opening for companies including Micro1, Mercor and Surge AI.

The scale of the market was already substantial. TechCrunch reported last year that Mercor was generating more than $450 million ARR, while Surge AI generated a reported $1.2 billion in revenue in 2024.

Competition is therefore no longer simply about who can label the largest volume of data. The increasingly valuable resource is access to people capable of generating new, expert-level data that cannot easily be scraped from the public internet.

Training data is becoming AI infrastructure

Micro1 is already moving beyond text-based human feedback.

The company has expanded into robotics data, including gathering real-world human activity that can be used to train embodied AI systems. Its own researchers argue that robotics development is creating demand for increasingly specialized post-training datasets as robots move into factories, homes, vehicles and other physical environments.

That could broaden the addressable market considerably.

The $500 million gross run-rate milestone therefore says something bigger than Micro1's growth alone. As frontier models become harder to improve simply by adding more publicly available internet data, high-quality proprietary human data is becoming an increasingly important layer of AI infrastructure.

If that trend continues, the companies controlling the pipelines connecting highly skilled humans, proprietary enterprise information and AI laboratories could capture an increasingly large share of the spending behind the next generation of AI models.