The near-bottomless demand for unique AI training data from top labs and companies is driving a massive boom for data labeling startups.
One of those fast-growing companies, Micro1, is a four-year-old startup that has grown its annual gross run rate from $100 million to $500 million in the past eight months, according to people familiar with the company. Similar to its peers, which employ professionals such as doctors, lawyers, and scientists on a contract basis, Micro1 holds about 60% to 70% of that figure, with a net annual operating rate of between $150 million and $200 million.
While Micro1 still lags behind competitors such as Mercor (which reached $2 billion in annual revenue this summer) and Handshake (which reached $1 billion earlier this year), the startup’s revenue growth shows there is more than enough demand to support multiple players providing AI training data.
Some researchers have hypothesized that future AI data spending could rival computing spending.
This outlook bodes well for Micro1, which expects contract size to grow at an accelerating rate and profits to grow over time. The startup increasingly generates synthetic data without human interaction, such as creating automatic descriptions for video content. Additionally, because some of the data the company generates can be sold to multiple customers, gross margins on this “off-the-shelf” data can be as high as 80% to 90%, people familiar with the startup’s finances told TechCrunch.
Selling the same dataset to multiple customers has sparked recent controversy, with critics arguing that distributing ready-made data to Chinese AI developers could make Chinese models as powerful as top American models.
Micro1 founder Ali Ansari told X last month that the startup doesn’t sell data to Chinese model makers, unlike some of its competitors. “Some human data companies are collaborating with foreign adversaries. (A) The results of this were on display today at KimiK3. We believe it is shameful to assert America’s desire to dominate AI while selling millions of worth of data to countries with which we are adversaries.”
Like Mercor, Micro1 started as an AI recruitment startup. But when Ansari realized that clients creating data labels were using his AI platform to vet and hire engineers for annotation, he decided to pivot and branch out into the data label business as well.
Ansari previously told TechCrunch that in addition to having experts evaluate model outputs (a concept known as reinforcement learning gyms), the company builds pre-training datasets for robotics by having hundreds of generalists record everyday object interactions at home.
Micro1 raised a Series A last September at a $500 million valuation, but TechCrunch understands that the startup may have recently raised another round at a significantly higher valuation.
Micro1 did not respond to a request for comment.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
