What company data is worth to AI labs in 2026
AI labs have largely run out of fresh public web text. The data they need now — how real teams actually work — lives inside companies. That has created a market, and some of its prices are public. Below are the figures we could verify, with sources, so you can sanity-check any number you're quoted. Everything here is publicly reported; none of it is a guarantee of what your company would be paid.
The headline deals (public figures)
| Deal | Reported value | Source |
|---|---|---|
| Reddit → Google (2024) | About $60M per year | Reuters |
| News Corp → OpenAI (2024) | More than $250M over five years (reported by WSJ) | Reuters |
| Dotdash Meredith → OpenAI (2024) | At least ~$16M per year fixed, plus variable payments | Adweek |
| Shutterstock AI licensing (2023 total) | $104M revenue from licensing to AI companies; anchor deals ~ $10M/yr each | Bloomberg via Yahoo Finance |
| Taylor & Francis → Microsoft (2024) | Almost £8M (~$10M) in the first year | as tracked by AI Watch.dog |
| Axel Springer → OpenAI (2023) | 'Tens of millions of euros' over three years (Bloomberg) | AI Watch.dog |
| Wiley (2024) | ~$44M expected from AI content-licensing deals | Books+Publishing |
What about ordinary companies?
Those deals involve huge public content libraries. Most businesses have something different and, for AI labs, arguably more useful: private operational data — Slack threads, SOPs, support tickets, project histories, CRM records, code. This is the material agents need to learn real work.
micro1, which runs a company data partnership programme, publishes its range: it says payments 'typically range from $150K to over $2M', paid within 30 days of acceptance, with larger multi-team or highly unique datasets at the top end (source: micro1 data partnerships). It looks for companies with 30+ full-time employees, mature operations, substantial historical records, mostly English, prioritising US then other Western markets. It also runs a separate track for companies that are no longer operating (source: micro1 company referral).
What moves the price
micro1 says compensation depends on: dataset size and volume; workflow complexity; data quality; domain expertise represented; and how unique the operational knowledge is (source: micro1 data partnerships). In practice that means: years of history beat months; connected systems (tickets linked to code linked to Slack) beat isolated exports; specialist domains (regulated industries, engineering, ops-heavy businesses) beat generic ones.
Privacy and what actually gets shared
micro1 states it commits not to resell or distribute data until personal and identifying information has been anonymised, and that data is transferred via secure APIs or approved methods (source: micro1 data partnerships). Before any deal, check what's covered by customer contracts, employee agreements and local law, and get your own legal advice. See our privacy and consent guide.
The honest caveats
- Big public deals (Reddit, News Corp) are not a benchmark for a 50-person company.
- Any estimate, including ours, is indicative. Final offers depend on data sources, size, quality, access terms and due diligence.
- Not every company qualifies, and an estimate does not mean you've been selected or are owed anything.
Get an indicative number for your company
Cadabra's 10-question valuation gives you an indicative range in about a minute, free and confidential.
GET YOUR FREE COMPANY DATA VALUATION
Cadabra is an independent referral partner of micro1. Figures above are as publicly reported at the time of writing; we haven't independently verified private deal terms.