How Much Is Your Business Data Worth in 2026? Real Deals, Prices and Examples

What has actually been reported

The clearest public numbers come from large platforms. Reuters reported in February 2024 that Reddit's AI content licensing contract with Google was worth about $60 million per year, according to one source. Reddit's own IPO filing disclosed data licensing contracts with an aggregate value of $203 million over roughly two to three years, across more than one buyer.

The Wall Street Journal reported in May 2024 that News Corp's five-year agreement with OpenAI could be worth more than $250 million, including compensation in cash and credits for OpenAI technology. Neither company officially disclosed the terms. Reuters reported in April 2024 that Shutterstock's initial AI deals with large tech firms ranged from $25 million to $50 million each, according to its CFO, and that Freepik licensed most of its 200 million image archive to two companies at 2 to 4 cents per image.

The table below summarises these reports. Each figure is attributed to its source and describes a specific deal, not a market average.

Published per-unit rates, and why they don't add up to one range

Reuters also quoted Defined.ai, a data marketplace, saying buyers were generally willing to pay around $1 to $2 per image, $2 to $4 per short video, $100 to $300 per hour of longer film and about $0.001 per word of text. Defined.ai said it splits those earnings with content providers, so a supplier receives less than the buyer pays. These are one company's indicative 2024 figures for particular content types, not AssetGate rates and not seller earnings.

Data programmes also publish their own marketing ranges. Scale AI describes illustrative partnership value of $10K to $1M+, and micro1 advertises tiers from $100K+ to $1M+ for qualifying datasets. These are advertised programme tiers, not transaction averages. Per-image, per-word and per-partnership figures measure different things, so combining them into one headline range would be misleading.

• Per-unit rates suit large, uniform media or text collections.
• Partnership tiers describe negotiated programmes with eligibility rules.
• Platform deals reflect continuous access to huge, growing corpora.

One-off, annual and multi-year deals are not comparable

A one-off licence pays once for a snapshot. An annual licence, like the reported Reddit arrangement, pays for ongoing access and usually includes updates or API delivery. A multi-year deal such as the reported News Corp agreement spreads value over several years and may mix cash with credits or product access. Always ask which structure a quoted number describes before comparing it with anything else.

Exclusivity, permitted uses and update obligations also change the number. A non-exclusive licence for internal evaluation is a very different product from an exclusive licence for commercial model training with quarterly refreshes.

Illustrative scenarios with explicit math

These scenarios only apply published rates to hypothetical volumes. They are not estimates for your data, and real offers can be lower, higher or zero.

Scenario A, specialist text: a company holds 20 million words of original, rights-cleared technical documentation. At the roughly $0.001 per word Defined.ai described, 20,000,000 x $0.001 = $20,000 paid by a buyer, before any marketplace split, legal review or preparation costs.

Scenario B, an image archive: 200,000 owned product photos at the 2 to 4 cents per image reported for Freepik's archive deals gives 200,000 x $0.02 = $4,000 to 200,000 x $0.04 = $8,000. At the $1 to $2 Defined.ai described for some images, the same volume would be $200,000 to $400,000, which shows how much content type and buyer demand move the price.

Scenario C, operational records: there is no reliable public per-record rate for workflow logs, support tickets or decision histories. Value here is negotiated against a specific use case, which is why we do not apply a number.

What drives value for a typical business

Large platform deals are not typical outcomes for small and mid-sized companies. What usually matters is whether your material is hard to find elsewhere and directly useful for a buyer's model or evaluation work.

• Uniqueness: data the public internet does not already contain.
• Depth: long historical coverage and consistent records.
• Context: decisions linked to outcomes, code linked to reviews and fixes.
• Quality: documented, deduplicated, consistently formatted material.
• Rights: clear permission to license, with personal data and third-party content handled lawfully.

Why there is no price per GB or per commit

Two gigabytes of duplicated logs can be worth nothing while a small, carefully labelled evaluation set can be valuable. Buyers pay for the capability the data adds and the risk it removes, not for storage size. Rights problems can reduce value to zero regardless of volume.

The practical next step is a qualitative read on fit and rights before anyone discusses money. AssetGate's free assessment does that from a description, with no data upload and no obligation. It does not produce a price.