Licensing a Patent Inside a Standard: RAND, Royalty-Free and the Gap Between Them

No standards body can validate a technical proposal without the risk that the proposal infringes someone’s patent. The bodies therefore require participants to disclose relevant patents or patent applications and to commit to licensing them on stated terms. This is the entire content of a patent policy: a private contract offered by patent holders to unknown future implementers, brokered by a consortium that has no power to grant a licence itself.

The standard terms come under two labels, RAND (reasonable and non-discriminatory) and royalty-free, and both are less than they sound.

What a declaration promises and to whom

A patent declaration in a standards body is a promise, not a transfer. The holder commits to grant licences on stated terms to parties who implement the resulting standard. The commitment runs to a class of recipients defined by the standard, not to a named list, and it binds whoever later acquires the patent.

The policy structure that most bodies follow has three elements. First, disclosure: participants should identify known essential claims, where essential means the claim cannot be avoided by any conforming implementation. Second, a written assurance: the participant either commits to licence the claim on the policy’s terms or refuses to, in which case the working group must consider redesigning the technology. Third, a general disclaimer that the body has not verified essentiality, cannot determine validity, and takes no position on the scope of any claim.

The second element is where the committee’s procedure becomes commercially visible. A refusal to license forces a redesign during drafting, and a redesign is contested by everyone whose implementation of the technology is at stake, which is why such disputes reach the working group as procedural objections rather than technical ones. How a group resolves an objection of that kind is the subject of the account of voting rules and objections.

The third element is the load-bearing one. A patent holder’s statement that a claim is essential is an opinion, and the body does not test it. A body that did test it would be making a legal determination it is not constituted to make.

The RAND problem

“Reasonable and non-discriminatory” is three decisions in one phrase, and none of them is defined by any standards body.

Reasonable implies some ceiling on royalty, but the policy states no number and no method for deriving one. Whether the ceiling is a percentage of a chip price or a percentage of a finished product price can differ by orders of magnitude for the same claim; a royalty of a few percent on a finished device is a very different commitment from the same percentage on a component sold for a fraction of the device’s price. Bodies avoid specifying the base, so the base is argued per dispute.

Non-discriminatory implies that similarly situated implementers get similar terms, but it does not define how to establish that two implementers are similarly situated, nor whether comparison is possible when the terms are confidential. RAND commitments in practice are frequently discharged by non-disclosure, which makes the discriminatory question unanswerable by the party most interested in it.

The result is that a RAND commitment functions as an obligation to negotiate in good faith toward a licence whose price is not specified. That is not nothing: it is the basis for antitrust-adjacent arguments about abuse, and courts have treated the commitment as creating enforceable obligations. It is also not a licence.

Royalty-free and its conditions

Royalty-free policies remove the pricing dispute and replace it with a scope dispute. A commitment to license for free usually comes with conditions that determine whether a given implementation benefits.

Typical conditions include: reciprocal licensing, where the implementer forgoes asserting its own patents essential to the same standard against the licensor or other implementers; a field-of-use limit, excluding particular application classes; a defensive-suspension clause, where the licence terminates if the implementer sues over the standard; and a condition that the implementer conform to the specification, since a non-conforming implementation arguably falls outside the class to which the promise was made.

Reciprocity is the condition with the largest practical effect. A firm holding patents essential to a standard may be better served, financially, by refusing to participate in a royalty-free commitment and licensing separately, than by joining and giving up its own claims. Where enough firms choose the former, the royalty-free standard is free to implement but not free to practise, because a second pool of claims sits outside the commitment.

The pooling arithmetic

Patents essential to a single widely adopted standard can run into the thousands, held by hundreds of holders. Negotiating each individually costs more than most implementers can bear for a component of modest value, so the industry aggregates claims into pools: an entity holds licences from many patent holders and offers a single licence to the pool’s portfolio.

Pools convert the negotiation problem into an allocation problem. The pool’s total rate is capped by what the market will accept, and the allocation among members is decided by a formula that counts each holder’s essential claims, sometimes weighted by a judgement of their technical contribution. This makes claim-counting economically rational: a patent that reads on any mandatory clause is worth a share, and a claim family that is cited by three clauses is worth more than one cited by a single clause. It also makes the question of whether a clause is mandatory commercially significant, which connects the patent policy to the normative language discussed in the treatment of normative and informative text. Procurement rules take the same linkage from the other end, treating royalty-free terms as part of what makes a standard open, as the discussion of openness definitions sets out.

A design consequence follows that is frequently overlooked by working groups: making a feature optional reduces its value to patent holders and reduces licensing friction for implementers who skip it. Standards that load their mandatory path with encumbered technique and put alternatives in optional clauses have effectively guaranteed that the mandatory path is where the money is.

Injunctions and the order of operations

The dispute pattern that dominated the 2010s concerned whether the holder of a declared-essential patent may seek a court injunction to stop an implementer from shipping while licensing talks proceed. Once a commitment has been made, the argument against the injunction is that the holder has already agreed to license, so the only live question is price, and a court can set price. The argument for it is that the implementer may be unwilling to negotiate at all.

The Court of Justice of the European Union addressed this in Huawei v ZTE (C-170/13), reasoning that a holder who has given a RAND commitment may not obtain an injunction without first following defined steps: notifying the alleged infringer of the infringement and the specific claim, and making a written offer on the committed terms. The alleged infringer must then respond diligently, which may include accepting the offer or making a counter-offer with appropriate security.

The judgment did not set a royalty rate and explicitly left that to negotiation or to national courts. What it established is procedural: the commitment converts an injunction question into a negotiation sequence. For an implementer, that is the practical value of the policy — not a price, but an order in which the price can be argued.

Reading a policy before joining a working group

An implementer assessing a standard should read the body’s patent policy as a contract with holes and locate the holes. The disclosures filed against the standard are public in most bodies and list holders and claims without stating terms. The difference between a standard whose essential-claim list is short and one whose list runs to hundreds of holders is a difference in expected licensing cost, and that difference is knowable before any product work begins.

What is not knowable from the policy is whether a claim is valid, whether it is truly essential, or whether an undisclosed claim exists. The policy’s own disclaimer says as much. The commitment binds those who chose to make it, and the population of those who chose not to make it is the risk the policy cannot remove.