Write an article about
Micro1 says its gross annual run rate has jumped from $100M to $500M in just eight months. The AI data startup has grown by connecting model developers with experts who create and evaluate the increasingly specialized data needed to improve their systems. That growth comes as spending on AI infrastructure continues to soar, but it also highlights another increasingly expensive part of building advanced models: the data itself.
The internet provided plenty of data to train early AI models. Finding the right data to make today’s more advanced models even better is becoming much harder.
Micro1 has positioned itself in the middle of that problem. It’s not just simply supplying large numbers of workers to label data, it actually recruits people with expertise in technical fields and connects them with AI companies looking to improve their models.
That value proposition is not exclusive to Micro1. There is plenty of competition for that business. Mercor and Handshake, and several other companies in this space, are also building large businesses around supplying AI developers with human expertise. The rapid growth across the sector suggests that better models have not reduced the need for people in the development process However, the human expertise needed is what is changing.
(Shutterstock/Robert-Way)
Micro1 was founded in 2022 by Ali Ansari, who has also been outspoken about where the company’s data should end up. Ansari said last month on X that Micro1 does not sell its data to Chinese model makers.
He wrote, “Some human data companies work with foreign adversaries. [A]nd the results show today in Kimi K3. We believe it’s shameful to claim American AI dominance desires while selling millions worth of data to countries that we are in adversarial competition with.”
It’s not just expertise that AI companies are willing to pay for. Mercor reportedly surpassed $2B in gross annualized revenue this summer. Handshake reached $1B earlier this year. Existing datasets are attracting serious interest too
Micro1 recently offered $12.5M for Spirit Airlines’ corporate data following the airline’s bankruptcy, topping a $10M bid from Google. The unusual bidding contest shows another side of the AI data boom. Companies are not only paying people to create new training material. They are also showing interest in proprietary data that can’t simply be pulled from the public internet.
The value of proprietary data comes largely from the fact that everyone else cannot easily get it. Major AI developers have access to much of the same public internet, and the biggest players can all spend heavily on the computing power needed to train and run their models. A company’s internal data is different. It can be the product of years of transactions, operations and interactions that cannot simply be recreated by another company.
That matters more as AI developers search for data that can still improve increasingly capable models. We know that there is an enormous amount of information available online, but much of it has already been used extensively for AI training. The data is accessible to competitors as well. That makes it harder to gain a competitive edge from public data alone.
(Shutterstock/KanawatTH)
Unique private datasets potentially offer something models have not encountered before. In that sense, the competition for better AI is increasingly becoming a competition for information that others do not have.
For businesses outside the AI industry, this creates a different question: what is the information they have accumulated actually worth?
Most enterprise data was created to run a business or satisfy regulatory requirements – not to train AI models. It appears that the emergence of buyers willing to pay for unusual datasets could give some of that information a second purpose.
However, turning corporate records into AI-ready data will be far from easy. Privacy restrictions, ownership questions and commercially sensitive information are just some of the challenges that limit what companies are able or willing to share. This means that the usefulness of any dataset will also depend heavily on its quality and what an AI developer is trying to accomplish.
Even with those limitations, data is beginning to look less like a byproduct of doing business and more like an asset with a market of its own. Micro1’s recent moves offer an early glimpse of what that market could become.
.Organize the content with appropriate headings and subheadings ( h2, h3, h4, h5, h6). Include conclusion section and FAQs section with Proper questions and answers at the end. do not include the title. it must return only article i dont want any extra information or introductory text with article e.g: ” Here is rewritten article:” or “Here is the rewritten content:”
