- AI data is priced per unit, as a flat fee, as a revenue share, or as milestones tied to delivery and acceptance.
- Per-unit prices are common for media; flat fees are common for corporate archives and codebases.
- Intermediaries reported by Reuters paid rights owners 20% to 30% of deal value in some cases.
- Published benchmarks give ranges, not prices; every deal is negotiated.
Pricing models
- Per unit. Per image, minute, hour, word, document or task. Common for media and new collection.
- Flat fee. One price for a defined dataset. Common for archives, codebases and closed-company data.
- Milestones. Payments on signing, delivery and acceptance.
- Revenue share. A share of what an intermediary earns from reselling. A Defined.ai supplier in Brazil told Reuters it paid owners 20% to 30% of deal amounts (Reuters via The Star).
- Ongoing. Recurring fees for continuing data flows, such as Reddit's reported $60 million a year with Google (The Decoder).
Benchmarks
See our AI data prices page for every published figure with its source.
Costs to subtract
Preparation, legal review, de-identification, transfer and any broker fees. A deal is only worth doing if the net figure justifies the risk.
Questions
Should I take a flat fee or revenue share?
A flat fee gives certainty; revenue share can pay more if the intermediary resells widely, but you take on their sales risk and lose visibility of who holds copies.