Upfront license fee vs running royalty: which suits a mid-sized company?

The verdict: certainty usually beats upside for a mid-sized licensor

A paid-up license, where the licensee pays a fixed fee once, suits a company that wants a known amount, a single payment and no ongoing administration. A running royalty, where payments follow the licensee's sales or usage, suits a licensor that can measure that usage, audit the reports and wait years for the money.

For operational data licensed for AI training, the first profile is far more common. A mid-sized company rarely has licensing staff, and the use of records inside a trained model is hard to observe or count. SourceX deals follow the paid-up pattern: one all-in price, paid once, typically within about 60 days of invoicing once the buyer selects the data.

One-time fee vs running royalty, side by side

Why royalties are awkward for AI training data

Royalties work when there is a countable event: a unit shipped, a seat sold, a stream played. Records licensed for AI training are absorbed into model development, and the resulting product's revenue has no clean line back to any one dataset. Any royalty base would have to be invented and then verified through reports the licensor cannot easily check.

That leaves a mid-sized licensor with three poor choices: accept the licensee's numbers on trust, pay for audits that may cost more than they recover, or negotiate a proxy metric both sides will argue about. A one-time fee for a defined snapshot of records, delivered once and used under agreed terms, avoids all three.

When a running royalty still makes sense

Royalties are the right tool in other licensing settings, and a CFO should recognize them:

• A patent or design licensed into a physical product with countable unit sales.
• A brand licensed for merchandise, where retail sales are reported anyway.
• Software components or content embedded in a product sold per seat.
• A licensor with in-house licensing staff or IP counsel who can negotiate definitions and run audits.

If the company's situation looks like one of these, a royalty, or an upfront fee plus a royalty, may be worth modeling. Data licensed for AI training rarely fits that pattern.

The accounting question is separate from the pricing question

How the licensor recognizes revenue depends on what the license grants, not only on when cash arrives. Under ASC 606, an entity assesses whether a license gives the customer a right to use its intellectual property as it exists when granted, recognized at a point in time, or a right to access it throughout the license period, recognized over time. Deloitte's ASC 606 licensing guidance walks through the functional versus symbolic IP distinction behind that test, and FASB clarified the licensing guidance in ASU 2016-10 without changing the standard's core principle.

Do not assume a one-time payment means one-time recognition, or that royalties mean recognition spread over the term. The contract decides, so take the draft license to the company's auditor before signing. Tax timing is a separate question again; ask the company's tax adviser how the payment will be treated in the year it lands, a natural agenda item for the year-end planning meeting.

This is general information, not legal, tax or financial advice. Confirm with your own counsel, tax adviser or professional body before acting.

A CFO checklist before choosing a structure

• Can the licensee's use be counted reliably by someone other than the licensee?
• Does the company have the staff or budget to review royalty reports and run audits?
• How long can the company wait for cash, and how much does certainty matter this year?
• What is the counterparty's credit risk over a multi-year term?
• How would a future buyer of the business value each structure in diligence?
• What does the auditor say about revenue recognition under the draft terms?
• What does exclusivity cost over the term? Work it through with exclusive data license opportunity cost.
• What will preparing the data cost internally? See what it costs a company to prepare data for licensing.

If most answers point to certainty and simplicity, a one-time fee is the better structure. Whether to license at all is a separate decision, covered in is licensing company data worth it for a mid-sized company.

How the SourceX structure works

The company receives a single all-in price, with SourceX's fee already included and no separate charges, paid as a one-time payment. It keeps ownership of its data because it grants a license rather than transferring title; license vs assignment explains the difference. Licenses are typically exclusive for AI training for an agreed term, and nothing binds the company until it agrees price and terms and signs.

Pricing comes after qualification. The company must be US-based, have reached 50+ full-time employees at peak (contractors excluded), show several years of documented operations, hold the rights to license its records and have an authorized sponsor. The company fit checker gives a preliminary, non-binding read.

For fractional CFOs advising on the structure

If you are the adviser who spots the opportunity, you can introduce the company as a SourceX partner. Partners earn 25% of the eligible platform fees SourceX actually collects from the referred company's licensing deals, capped at $100,000 per referred company. Like the company's own payment, the reward depends on cash actually changing hands: it is payable only after the buyer pays and SourceX receives its fee, and rewards are not guaranteed. It comes from SourceX's fee and is never deducted from what the company receives. The fractional CFO partner page covers the role.

Next step

Work through the checklist with the client's CEO or owner. If the one-time structure fits and the company qualifies, register as a partner to make the introduction, or have the company apply at sourcex.si/apply.