The race to build the most advanced artificial-intelligence tools has spawned a new market for real-world data. It comes with both opportunities and pitfalls for investors.
Google won an auction last Friday to purchase Spirit Airlines' software code, internal messages, and financial and operational data for $10 million, according to filings from Spirit's bankruptcy proceedings. Google hopes to use the data to improve its products and AI models, a spokesperson said.
That $10 million outlay for a busted airline's data is just the tip of the iceberg. AI companies are vacuuming up corporate information of all sorts-a process that has taken on greater importance now that scraping the internet has become more legally fraught.
The demand for data is linked to how AI large-language models are created and refined. The computer systems ingest huge volumes of human information and use it to mimic human processes, such as writing reports, interpreting and responding to messages, and making recommendations. More data means more examples the model can pull from.
In response, public companies are weighing whether to license their data. And start-ups are opening marketplaces through which smaller enterprises can sell everything from contracts to email threads.
"Companies are sitting on decades of records that show how real work gets done, and that data is now some of the most valuable material for training and evaluating AI," said a spokesperson from Mercor, an AI data start-up that made a $7.5 million bid in the Spirit auction.
The economics make perfect sense. The five largest cloud providers are making a combined $800 billion in capital expenditures in 2026 to stand up data centers, and AI labs are paying hundreds of billions to use the facilities. Millions of dollars-with an "M"-mean very little to them when creating new AI tools, but that money can be a game-changer for data sellers.
The market appears to be picking up. In January, Cloudflare acquired Human Native, a marketplace connecting content creators with AI developers. Two months later, News Corp, the owner of Barron's publisher Dow Jones, struck a licensing deal with Meta Platforms, where the publisher sold access to its content for $50 million a year.
For smaller companies and even individuals looking to sell their data, an array of start-ups are emerging to broker deals.
In the last 11 days, micro1, an AI data venture, has committed more than $20 million to license operational data, CEO Ali Ansari said in a social media post Tuesday.
Investors can get in on the action, too. Barron's made two stock picks last month based on the growing value of AI training data.
Reddit struck agreements with OpenAI and Google in 2024 to license conversations from its forums, bringing in around $60 million a year from each lab. The social-media platform is now in negotiations with Google, and some investors and analysts expect a much larger deal.
John Wiley & Sons, a publisher of academic and technical literature, has also racked up big-name customers. The company's licensing agreements with Amazon Web Services, Microsoft, and Anthropic have generated more than $110 million since 2024.
The key for investors, and indeed companies, is pinpointing when data sales may do more harm than good.
Privacy is one challenge, particularly for consumer-facing businesses. Opera, the Norwegian web-browser company, has resisted selling data on those grounds.
"If we lose the users' trust, if they think, 'Oh, if we use the Opera browser, then they're actually going to sell our data,' then it's probably not a wise move," Frode Jacobsen, chief financial officer at Opera, said at an investor conference in January.
Disruption by AI is the other threat. If the labs use a company's proprietary information to build something better, or at least more popular, they could siphon away demand.
"Our data is really the crown jewels of the future," Gary Morrison, CEO of hostel booking platform Hostelworld Group, said in a conference call last month. "It is everything that we know about profiles, about bookings, about events, about people you're meeting, what you're talking about."
But companies like Hostelworld, and indeed Reddit and John Wiley, will have to survive in an AI-dominated world whether they like it or not. That means many more executives will have to decide how much "the crown jewels" are worth.
Comments