Revenue share vs. one-time licence fees

There are two broad ways to get paid for licensing data. One pays a fixed fee when the deal closes. The other promises a share of whatever revenue the data helps generate later. For company data licensed to AI labs for training and evaluation, the one-time fee is usually the better fit, and the reasons are more practical than philosophical.

This post compares the two models in plain terms. It is general information, not legal, tax or financial advice, and any company weighing deal terms should involve its own counsel and accountants.

A one-time licence fee is simple. The parties agree on a price for a defined dataset, under defined terms, and the buyer pays it. In our process, payment goes to the company entity, typically in 2–4 weeks, and the typical range is $20K to $400K+ depending on what the data contains.

A revenue share ties the seller's payment to something that happens later, such as the buyer's sales of a product, usage of a model or resale of the data. The seller might receive a small upfront amount and a percentage of a defined revenue stream for some period.

Both are legitimate structures in other contexts. Music, software and content licensing all use royalties. The question is which one suits a dataset made of a company's Slack history, email, tickets, code and CRM.

A revenue share only works if there is a clear revenue line to share. With company data used for AI training, there usually is not.

Imagine a 75-person B2B software company whose anonymized records become one of many sources in a training mix. The lab trains a model on that mix, post-trains it, builds evaluations and eventually ships products. Which dollar of the lab's revenue came from that one company's data? There is no honest way to answer, and any formula the parties invent will be arbitrary.

That leaves the seller with a payment stream based on a number it cannot verify, from a buyer whose finances it cannot see. Audit rights help in theory, but auditing a frontier lab's revenue attribution is not a realistic project for a small company, let alone one that is closing.

Many companies that license their data are winding down or already closed. For them, a promise of future payments creates real problems.

• A liquidator or assignee typically needs to realize value and distribute it within the life of the process, not wait years for royalty statements.
• A closed company may not exist in a form that can receive, account for and distribute payments in the future.
• Creditors and shareholders tend to value a known amount now over an uncertain amount later.
• Directors winding down informally want to close the books, not leave an open receivable that depends on someone else's business.

Operating companies have more flexibility, but similar logic applies. A fixed fee can be booked, explained to a board and planned around. A revenue share introduces a long tail of reporting and follow-up for a payment that may never materialize. Our post on five questions your board will ask about licensing data covers how boards tend to think about this.

A fixed fee still needs a basis, and that basis is the data itself. Value is driven by connection depth (the same project traceable across several tools), years of history, headcount and the number of systems involved. Our post on how company data is valued explains each factor.

That is also why nobody can give a firm number before counting. The read-only evaluation, which takes under an hour of the seller's time, measures what is actually in scope after exclusions. The price follows from that measurement rather than from a forecast of the buyer's future business. For a rough sense before any call, the value estimator gives a starting point.

A fixed fee also fits the shape of the licence. The licence is limited to training and evaluation; the buyer may never market to, profile, contact or re-identify anyone in the data, and the data is never resold for marketing or people search. Because onward sale is ruled out, there is no resale revenue for a seller to share in the first place.

Sellers sometimes ask about revenue share because they want to participate in the upside of AI. That is a reasonable instinct, and it is worth naming what it would cost.

A revenue share usually means accepting less up front in exchange for a speculative tail. It means keeping a company entity, a bank account and someone responsible for monitoring statements. And it means relying on a counterparty's reporting for years. For many sellers, especially those closing down, those costs outweigh the upside.

There may be unusual situations where a different structure makes sense, and every seller should take advice on its own circumstances. But a clean, fixed fee paid to the company entity, under a narrow licence with deletion of the raw archive within 30 days of sign-off, is a straightforward way to recover value without creating new obligations.

A one-time fee is measurable, timely and simple to account for. A revenue share for training data is hard to measure, slow to pay and difficult to verify. For a company that wants to recover value from its records and move on, the first is usually the more practical choice.

If you want to understand what a fixed fee might look like for your company's data, book a 15-minute call and we will walk through what drives the number.