When Does the Money Stop? Modeling Royalty Stream Expiry in Pharmaceutical Transactions

The single most consequential variable in any royalty valuation is the terminal date, the point at which cash flows cease. Getting it wrong by even two years can shift an asset's net present value by 15 to 25 percent.

Yet the mechanics of royalty expiry remain among the least well understood aspects of pharmaceutical deal structuring, because the answer to "when does this royalty end?" is almost never as simple as "when the patent expires."

There is a second question that gets asked even less, and it is the mirror image of the first: when does the money start? A royalty is not a date. It is a window with two contractual edges, and both of them move.

Most modeling attention goes to the far edge. The near edge, when the clock actually starts ticking, is equally a contractual construct, equally jurisdiction by jurisdiction, and quietly value relevant in ways a headline term length hides.

This article works through both edges. It examines when a royalty begins, when it ends across five deal architectures, how supplementary protection certificates behave in Europe (the single most mismodeled input on the far edge), what legal constraints apply to post-patent royalties, how the institutional royalty funds approach duration, and where the analytical edge resides for investors and transaction structurers.

When the clock starts: commencement, not signing

Two clocks run in every royalty deal, and confusing them is the most common error at the front of a model.

The first is the contractual royalty-term clock. In a licence, the royalty term is almost always defined country by country, and in each country it begins on the first commercial sale of the licensed product in that country. It does not begin on the effective date of the agreement, on signing, or on regulatory approval. It begins when the first unit is sold.

This matters because the same trigger that starts the clock in a fixed-term deal (first commercial sale) is the trigger that, years later, ends it. A deal that pays for fifteen years from first sale has a start edge and an end edge welded to the same event, staggered fifteen years apart, and both edges move if the launch slips.

The second is the investor's economic clock, which starts at closing or funding. In a monetisation, the buyer wires cash on a closing date and begins accreting a return from that moment, whether or not a single royalty dollar has yet been received. The gap between the two clocks is where value quietly leaks.

The dead zone in pre-approval synthetic royalties

The clearest illustration is a synthetic royalty funded before launch. In the Royalty Pharma to Revolution Medicines transaction from June 2025, Royalty Pharma committed up to 1.25 billion dollars of synthetic royalty in five tranches of 250 million each, with the first two tranches payable before FDA approval.

The SEC filing states that royalties on US sales run for fifteen years after the first commercial sale of daraxonrasib in the United States, and ex-US royalties run for fifteen years after first commercial sale in the European Union.

Read those two facts together. The money goes out before approval, but the fifteen-year term does not start until first commercial sale, which is a post-approval event. Between funding and first sale there is a dead zone: capital deployed, clock not yet started, no royalty income.

The investor's economic clock has been running the whole time. A model that treats "fifteen years" as the duration without pricing the dead zone at the front overstates the return, because the true holding period is the dead zone plus fifteen years, and the discounting runs from funding, not from first sale.

Country-by-country commencement

Because each jurisdiction's term starts on its own first commercial sale, staggered launches stagger both edges of the window. A product that launches in the US in 2027, the EU in 2028, and Japan in 2029 has three different start dates and, under a fixed-term structure, three different end dates.

The "royalty duration" of the asset is therefore not one number but a set of overlapping windows, each opening and closing on its own schedule.

This is the same jurisdictional matrix that governs the far edge, applied to the near edge, and it is why launch-timing assumptions per country are a first-order input rather than a rounding detail.

The arrears lag, and the accrued stub

Royalties are reported and paid in arrears. The market convention is a quarterly report and payment within roughly 30 days of the calendar quarter end for a licensee's own sales, often 45 to 60 days where sublicensee royalties flow through an intermediary.

The economic effect is that the first royalty dollar reaches the holder not on the day of first sale but a quarter and a bit later, and every subsequent payment carries the same lag. Over the life of a stream this is a modest but real duration and cash-timing input, one quarter of receipts effectively deferred throughout.

The lag also creates an artefact at acquisition. When a buyer closes on an already-marketed royalty, there is usually an accrued-but-unpaid stub: sales have occurred, the royalty has been earned, but the payment has not yet been remitted.

Whether that stub transfers to the buyer or is retained by the seller is a negotiated term, and on a large stream it can be a meaningful slice of the first year's cash.

The clean question to ask of any acquisition is not only "when does the term end?" but "as of the closing date, who owns the receivables that have already accrued?"

Milestones bridge the gap, and mark the transition

Between the effective date and first commercial sale, the economics of most licences are carried by development and regulatory milestones, not royalties.

The transition milestone, typically a first-commercial-sale or first-sales-threshold payment, is the hinge between the milestone phase and the royalty phase, and it is the cleanest marker of the near edge in a model.

Where a monetisation includes milestone participation, as several recent structures do, the front of the cash-flow schedule is milestone-shaped and only later becomes royalty-shaped. Modeling the two phases with a single assumption blurs precisely the moment the clock starts.

The royalty window has two edges, and both move. The contractual term does not begin at signing or approval but at first commercial sale, while the investor's capital and economic clock start at funding. The gap between them, plus the quarterly arrears lag, defers cash that a headline term length does not show.

The default assumption, and why it is usually wrong

The instinctive assumption, that royalty term equals patent term, holds in a meaningful minority of deals, particularly older university out-licences and composition-of-matter-linked agreements. But in practice, pharmaceutical licence agreements define royalty obligations independently from patent life. The royalty term is a contractual construct. Patents may set a ceiling, a floor, or neither.

Royalty Pharma's 2025 10-K states this with precision: the duration of a royalty can be based on regulatory and marketing approval dates, patent expiration dates, the number of years from first commercial sale, the first date of manufacture of the patent-protected product, the entry of generics, or a contractual date arising from litigation, all impacted by the point in the product's life cycle at which the royalty is acquired.

Royalty Pharma further notes that where a royalty term is linked to the existence of valid patents, management must make judgments about the patent providing the strongest protection in order to align the period over which it forecasts cash flows to the royalty term. It is common for the latest-expiring patent in effect at acquisition to be extended, adjusted, or replaced with newer-dated patents in later periods.

The estimated weighted average duration of Royalty Pharma's portfolio is approximately 13 years based on projected cumulative cash royalty receipts, per its UK Annual Report. Several of its marketed royalties have unlimited durations and could provide cash flows for many years after key patents have expired.

This is not a theoretical concern. It is the central modeling challenge for every participant in the royalty financing market.

The five expiry architectures

Across the pharmaceutical royalty market, five distinct expiry architectures account for the vast majority of transactions. Each creates a fundamentally different modeling challenge and risk profile.

1. Patent-linked expiry

Royalties terminate on a country-by-country basis upon expiration of the last relevant patent in each jurisdiction. This is the most common structure in university out-licences, composition-of-matter deals, and older pharma-to-pharma agreements.

The Genmab to Janssen daratumumab licence illustrates the mechanics and the risks. The agreement specifies that Janssen's royalty obligation expires country by country on the later of 13 years after first commercial sale or expiration of the last-to-expire relevant Genmab-owned patent. The composition-of-matter patents did not begin to expire until March 2026, with US patents (including patent term extensions) expiring around 2029 and European patents around 2031.

This structure became the subject of binding arbitration in 2020 to 2022, when the parties disputed whether "relevant patent" included Janssen-owned formulation patents covering the subcutaneous formulation of daratumumab (DARZALEX FASPRO), which would not expire until the mid-2030s, or only Genmab-owned patents. The tribunal ruled in Janssen's favour: the royalty obligation extends only through the last-to-expire relevant Genmab-owned patent, not Janssen's own patents.

The financial significance is material. On a product generating over 10 billion dollars in annual global revenue, the difference between a royalty terminating in the late 2020s versus the mid-2030s is billions of dollars in present value. Genmab disclosed that it expects royalties to begin to decline materially in 2029 following expiration of US patent rights, a statement that effectively signals the binding constraint for valuation purposes.

Modeling implications. Patent-linked royalties require a granular, jurisdiction-by-jurisdiction patent schedule that accounts for patent term extensions (PTEs under Hatch-Waxman in the US, limited to 5 years), supplementary protection certificates (SPCs in Europe, up to 5 years plus a 6-month paediatric extension), and any paediatric exclusivity extensions.

The "patent expiry date" is not a single number, it is a matrix of dates, and the definition of "relevant patent" in the agreement determines which entries in that matrix control the royalty term. In Europe, as the next major section sets out, the entry that usually controls is not the base patent at all but the SPC, and the SPC is the input most likely to be modeled wrong.

2. Fixed term from first commercial sale

Royalties run for a defined number of years from first commercial sale in each country, regardless of patent status. This decouples the royalty term from IP uncertainty, making it the cleanest to model but the most exposed to commercial performance risk within the window.

The Royalty Pharma to Revolution Medicines transaction is the clearest recent example at scale, and the commencement mechanics discussed above are its front edge. Royalty Pharma committed up to 1.25 billion dollars for tiered royalties over a fixed term of 15 years on worldwide net sales of daraxonrasib, with royalties decreasing by sales tier and falling to zero above 8 billion dollars in sales.

The companion molecule, zoldonrasib (RMC-9805), has a first-to-trigger term: 15 years after the earlier of its approval in an overlapping indication with daraxonrasib or the first commercial sale of daraxonrasib, tying the companion's window to the lead product's timeline. Resulting duration is estimated at 2038 to 2041 depending on launch timing.

Modeling implications. The modeling advantage is a single key input per jurisdiction, the expected launch date, rather than a full patent landscape. The modeling risk shifts entirely to commercial performance, to generic or biosimilar erosion timing within the window, and to whether 15 years extends far enough beyond loss of exclusivity to capture tail value.

The tiered structure creates a concave payoff that must be modeled with explicit sales scenarios, not point estimates. And the near edge dominates: because the term starts at first sale, every year of launch delay pushes both edges out and lengthens the dead zone against the investor's clock.

3. Hybrid "later of" clauses

The most sophisticated and increasingly common structure defines the royalty term as the later of two or more triggers, typically the later of patent expiry or a fixed number of years from first commercial sale. This creates an asymmetric duration profile: a guaranteed minimum life (the fixed-term floor) and a contingent extended life (the patent ceiling, or vice versa).

The Dyax to Ares collaboration governing Bavencio (avelumab) is the textbook example, and one that generated significant litigation. Ares agreed to pay royalties on net sales of antibody products developed from Dyax-discovered fragments until the later of expiration of the licensed patents and ten years after first commercial sale. The licensed CAT patents expired in 2018, but first commercial sale occurred in 2017, extending the obligation to 2027, nine years beyond patent expiry.

The contract did not include a royalty step-down at patent expiry, a structural choice that became the basis for Ares's Brulotte challenge, discussed below.

A variant appears in the Genmab to Janssen agreement, where the term runs on the later of 13 years after first sale or expiration of the last-to-expire relevant patent, but also includes reduction events. Royalties are subject to reduction on patent expiration or invalidation in a country and on first commercial sale of a biosimilar there. This creates a three-layered model: the term is the later of two triggers, but the rate steps down on specified events within that term.

Modeling implications. "Later of" structures create a two-scenario framework in which both patent landscape and launch timeline must be modeled simultaneously. The binding constraint differs by country. Where the product launched early and patents were extended, the patent prong may control; where launch was delayed and no extension was granted, the fixed-term prong controls. Step-down provisions add a kinked cash-flow profile requiring separate pre- and post-step-down modeling.

4. Return-capped structures

In royalty-backed note financings and certain synthetic royalties, the stream does not expire on a calendar date. It expires when a cumulative payment cap is reached. Duration is a function of commercial velocity, not calendar time or patent status. The expiry date is an output of the model, not an input.

The Royalty Pharma to Zymeworks 250 million dollar note from March 2026 uses a time-dependent cap multiplier. Royalty Pharma receives 30 percent of worldwide tiered royalties on Ziihera (zanidatamab) until cumulative payments reach 1.65 times the note (412.5 million dollars) if achieved by 31 December 2033, or 1.925 times (481.25 million) if achieved thereafter. Full rights revert to Zymeworks once the cap is reached. Zymeworks described this as a longer-duration structure than a traditional royalty loan, with duration risk shared.

The NANOBIOTIX to HealthCare Royalty agreement from October 2025 adds a further layer: after the return cap (approximately 1.75 times MOIC if repaid by end 2030, rising to 2.50 times thereafter), a reduced royalty tail commences, not to exceed 14.9 million dollars per year, expiring 10 years after first commercial sale of JNJ-1900 (NBTXR3) in the US. Two sequential duration phases, a return-capped phase and a fixed-term tail.

The GENFIT to HealthCare Royalty agreement from January 2025 adds an annual sales cap overlay on Iqirvo (elafibranor), with GENFIT retaining royalties above the threshold and all royalties reverting after the cumulative cap and time limits.

Modeling implications. Return-capped structures require a revenue forecast that calculates cumulative payments year by year and determines when the cap is breached. The Zymeworks multiplier creates a step function: the investor's IRR is materially higher if the cap is reached before 2033 (1.65 times) than after (1.925 times), so the model must weight the probability of early versus late attainment. The tail in the NANOBIOTIX deal requires a second-phase model on top of the first. Sensitivity to commercial assumptions is far higher than in patent-linked or fixed-term structures.

5. Perpetual royalties

Some royalties have no contractual end date. They run in perpetuity, or more precisely until the underlying product generates no further revenue.

Royalty Pharma's interest in the Vertex cystic fibrosis franchise is the most significant perpetual royalty in the market. The royalties are perpetual and not tied to patent expirations, originating from the Cystic Fibrosis Foundation's funding of Vertex's CF programme. In its 10-K, Royalty Pharma notes that although the royalty is perpetual, it estimates Trikafta patent expiration in 2037 and potential generic entry thereafter. For Alyftrek, which incorporates deutivacaftor, it has applied an end date of 2039 to 2041 for accreting income.

The perpetual nature creates a distinctive challenge: the question is not "when does the royalty end?" but "what is the terminal value of an indefinite cash-flow stream subject to competitive erosion?" Royalty Pharma has argued that deutivacaftor is the same as ivacaftor and is therefore royalty-bearing, which would blend to roughly 8 percent for Alyftrek, extending the economic life to a next-generation product with its own exclusivity.

Modeling implications. Perpetual royalties require explicit terminal-value assumptions. The standard approach models cash flows explicitly through expected patent protection, then applies a post-LOE erosion curve to the tail, with a discount-rate premium reflecting the uncertainty of post-LOE cash flows. The successor-product question, whether next-generation formulations fall within the royalty-bearing definition, is a legal determination with multi-billion-dollar consequences and must be modeled as a probability-weighted scenario, not a binary.

The patent clock: where the far edge is anchored

For every architecture whose end is set by a patent, whether a patent-linked out-licence or a synthetic royalty written to run until patent expiry, the terminal date is only as reliable as the date the patent clock is anchored to. The most common far-edge error is not in the extensions. It is in the anchor.

The anchor is the filing date, not the grant date, and not the priority date. A US utility patent expires 20 years from the earliest non-provisional US or PCT filing date to which benefit is claimed. Three things do not move that clock, and each is a live trap in a valuation:

• The grant date. Term runs from filing, so anchoring on the issue date overstates duration by the whole of prosecution, routinely two to four years on a pharmaceutical patent. Because the term now depends on the filing date rather than issuance, a slow grant does not buy a longer life.
• Provisional applications. A US provisional does not count toward the 20-year term; it secures priority for patentability only. A March 2025 provisional followed by a March 2026 non-provisional expires in March 2046, twenty years from the non-provisional, not twenty-one from the provisional.
• Foreign priority. A Paris Convention claim to an earlier foreign application does not shift the US term. The clock still runs from the US or PCT filing.

The twenty-year term runs from the earliest non-provisional or PCT filing date. Anchoring on the grant date overstates duration by the whole of prosecution; anchoring on a provisional or a foreign priority understates it. PTA, any terminal disclaimer, and then PTE or the SPC layer on from that anchor.

On top of the filing-anchored expiry sit adjustments that pull in opposite directions and are both easy to miss. Patent term adjustment (PTA under 35 U.S.C. 154(b)) adds days for USPTO prosecution delay, sometimes years, extending expiry beyond the nominal 20. A terminal disclaimer, filed to overcome a double-patenting rejection, caps a patent's term at an earlier relative's expiry and can silently delete that PTA.

Regulatory patent term extension under Section 156, covered in the SPC section, is layered on after that. Europe is simpler at the anchor: an EPC patent runs 20 years from the actual filing date with no general PTA equivalent, and the SPC does the extending, computed, as set out below, from that same filing date.

Which patent, and therefore which filing date. The terminal date is set by the last-to-expire relevant patent, and different patents in a family carry different filing dates. The composition-of-matter patent is usually filed earliest and expires first, setting the floor. Later-filed secondary patents (formulation, salt, polymorph, method of use, device) expire later and can set the ceiling, but only if the licence's definition of "relevant patent" reaches them.

This is the Genmab-Janssen question in another guise: the arbitration was, at bottom, a fight over which patent family's filing date anchored the European end of a royalty on a ten-billion-dollar product.

Synthetic deals written to the patent

The anchor matters most where the royalty is written to end at patent expiry rather than at a fixed term, which is the case in a meaningful share of synthetic royalties. Here the investor is not buying a contractual duration at all. They are buying the patent's remaining life, and the terminal date is whatever the filing-anchored, PTA-adjusted, extension-boosted, invalidation-exposed expiry turns out to be.

That exposure is asymmetric, and it is why patent-tied synthetics are modeled differently from the fixed-term kind. The upside is capped: the royalty cannot run past the relevant patent's extended expiry, however well the product sells. The downside is open: an inter partes review at the PTAB, an EPO opposition, a successful Paragraph IV challenge, a terminal-disclaimer or double-patenting problem on the very patent the model leaned on, or generic entry by settlement can each cut the stream short of nominal expiry.

Revolution Medicines' fixed term of fifteen years from first commercial sale removes exactly this risk by decoupling the end from the patent. A patent-tied synthetic keeps it, and the investor bears it with no operating control over the product or its defence. Some deals split the difference with a "later of patent expiry or a fixed number of years from first sale" floor, which converts the pure patent bet into a hybrid with a guaranteed minimum.

The practical consequence is that in a patent-tied synthetic the terminal date is not a constant to be read off the contract. It is a distribution: the nominal filing-anchored expiry as the right-hand bound, shifted earlier by the probability-weighted risk of invalidation and early generic entry.

That is the machinery of Steps 2 through 4 below, but here it determines the entire duration of the asset rather than only its tail.

The SPC problem: Europe's most mismodeled terminal date

For any patent-linked or hybrid royalty with European exposure, the far edge is almost never set by the base patent. It is set by the supplementary protection certificate, and the SPC is the input a duration model is most likely to get wrong, because it is treated as "patent plus five years" when it is nothing of the sort.

An SPC is a sui generis right that extends patent-like protection for an authorised medicinal product after the basic patent expires. It is not the patent, it does not share the patent's scope, and it is not automatic. Three features make it dangerous to model casually.

It is computed, capped, and conditional

The term of an SPC is not a flat five years. It equals the time between the filing date of the basic patent and the date of the first marketing authorisation in the EEA, minus five years, capped at five years. If fewer than five years elapsed between patent filing and first authorisation, no SPC term is available at all.

A paediatric extension of six months is available where an agreed paediatric investigation plan has been completed and the product is authorised across the relevant Member States, taking the maximum SPC term to five and a half years.

Worked through: a patent filed in 2001 with first EEA authorisation in 2014 yields the full five years (thirteen years elapsed, minus five, capped at five), so protection runs to roughly 2026 plus any paediatric extension.

A patent filed in 2001 with authorisation in 2008 yields only three years. The same molecule, the same base patent, different authorisation timing, materially different terminal date. A model that assumes "patent plus five" for every European asset is systematically wrong in both directions.

Critically, the SPC covers only the product authorised, not the full scope of the basic patent's claims. So in a licence whose royalty term runs to the last-to-expire relevant patent, the relevant right in Europe is usually the product-specific SPC, and the Genmab-Janssen style question, which patent counts, becomes a question about which SPC counts.

It may never issue: the Article 3(a) case law

The most expensive SPC modeling error is counting a certificate that never issues or does not survive challenge. Whether an SPC can be granted turns on Article 3(a) of the SPC Regulation, that the product is "protected by a basic patent in force," and the Court of Justice has spent a decade narrowing what that means.

In Teva v Gilead (C-121/17), the Court set a two-part cumulative test for combination products: the combination must necessarily, in light of the description and drawings of the patent, fall under the invention covered by that patent, and each active ingredient must be specifically identifiable in light of all the information disclosed by the patent.

Gilead's SPC for Truvada, a combination of tenofovir disoproxil and emtricitabine, was invalidated across Europe because the patent described the second ingredient only in functional terms ("optionally other therapeutic ingredients"), which was not specific enough. The practical lesson for a royalty modeler: an assumed European SPC on a fixed-dose combination, where the patent does not name each component as part of the invention, is a coin toss at best, and its loss can strip years off the European terminal date.

The Court also closed a duration-extension theory. In Santen (C-673/18), the Grand Chamber overruled its earlier Neurim decision and held that a new therapeutic application of an already-authorised active ingredient cannot ground a fresh SPC.

The first marketing authorisation for the product is the reference, regardless of the new use. So a valuation that assumes a repurposed old molecule earns a new SPC on the strength of a second-use patent is assuming a certificate the law will not grant.

The manufacturing waiver erodes the tail even when the SPC stands

Even a valid, granted SPC no longer buys a clean monopoly through its full term. Regulation (EU) 2019/933, the SPC manufacturing waiver, operational since July 2022 for SPCs applied for on or after 1 July 2019, carves two exceptions into the certificate holder's rights.

Generic and biosimilar makers may manufacture in the EU during the SPC term for export to third countries where protection has expired, and may manufacture and stockpile in the EU during the final six months of the term for a day-one launch the moment the SPC lapses.

The contractual expiry of the SPC does not move. Its economic value does. The last six months of protection now sit against a competitor that has already built EU inventory and is positioned to launch at volume on day one, and against a competitor that has been serving ex-EU markets from EU plants throughout the term. Post-LOE erosion in Europe is therefore faster and better pre-positioned than the pre-waiver historical analogs suggest.

A model that applies a pre-2022 European erosion curve to a post-waiver launch understates the speed of the decline. This is the contractual-versus-economic gap, appearing inside the SPC term itself.

It is fragmented, and reform is close but not here

SPCs are national rights. They are applied for, granted, refused, and can expire on different dates country by country, and divergent national outcomes on the same patent are common. Even the start date of an SPC can differ by a day between Member States because of small differences in how patent term is counted. The European "SPC expiry" for an asset is a matrix, not a date, in the same way the underlying patent landscape is.

The EU is reforming this, but the reform is not yet law. The Commission's April 2023 proposals would create a centralised examination procedure at the EUIPO in Alicante and a unitary SPC riding on unitary patents, with the national route retained for national patents.

The European Parliament adopted its first-reading position in February 2024. As of mid-2026 the files remain in the Council, where the debate has narrowed to the legal basis for the unitary SPC, with the Presidency circulating a Solution A built on Article 118 TFEU and a Solution B built on Article 114. The substantive Article 3 conditions and the CJEU case law above remain unchanged under all versions, and the reform adds an anti-"SPC squatting" consent requirement.

For modeling, the reform changes the process, not the arithmetic. Centralised examination should reduce divergent national outcomes and make the terminal-date matrix more predictable, but the term formula, the paediatric extension, the manufacturing waiver, and the will-it-issue tests all survive. A unitary SPC also inherits the exposure of the underlying unitary patent, so a single central revocation could collapse protection across all participating states at once, which is a concentration of terminal-date risk that the current fragmented system diffuses.

The non-patent clocks: data, market and orphan exclusivity

Patents and SPCs are not the only rights that gate generic and biosimilar entry in Europe, and for some products they are not the binding one. Regulatory data protection runs on its own clock, independent of the patent estate. Under the current regime a reference product receives eight years of data exclusivity plus two years of market protection, with a further year available in defined cases, the familiar "8+2+1."

During data exclusivity a generic or biosimilar cannot even reference the originator's dossier; during market protection it may be approved but not sold. Orphan medicines carry a separate ten years of market exclusivity, and the reform adds a standalone data-protection period for repurposed products.

For a royalty modeler this matters in two ways. First, where these periods run longer than the patent or SPC, they, not the IP, set the date of first possible competition, and a model anchored only on the SPC will place generic entry too early. Second, the regime is changing.

The EU pharmaceutical legislation overhaul, the largest in two decades, is close to final and expected to be adopted around the autumn of 2026; the compromise text reshapes the baseline data-protection period, attaches conditions tied to launching across member states, and moves market protection toward a one-plus-one-plus-one structure.

A European duration model built today should flag which assets depend on the current 8+2+1 and test them against the reform's transition. This is the European counterpart to the US data-exclusivity and IRA compression discussed below: a non-patent revenue clock that can move the effective far edge without any patent expiring.

One further European change bears on the near edge rather than the far. Since 12 January 2025 the EU Joint Clinical Assessment, under the HTA Regulation (EU) 2021/2282, has centralised the clinical assessment of new oncology active substances and advanced therapy medicinal products at EU level, with orphan medicines following in 2028 and all remaining medicines in 2030.

The JCA does not set price or reimbursement, which stay with the member states, but it reshapes the launch-to-reimbursement timeline that governs how quickly European royalty revenue ramps after first sale. For an oncology or advanced-therapy asset it is now a first-order input into the near edge and the shape of the European revenue curve, alongside the launch-date assumptions above.

The US analogue is different, and carries its own trap

The US equivalent, patent term extension under 35 U.S.C. 156, is often modeled as interchangeable with an SPC. It is not. A US PTE extends only one patent per approved product, chosen by the patentee, by half of the testing phase plus all of the approval phase, capped at five years and subject to a ceiling of fourteen years of effective patent life from approval.

It is distinct from patent term adjustment under Section 154, which compensates for USPTO delay. An SPC, by contrast, is product-specific rather than patent-specific and is computed from a different formula.

There is also a live trap on the US far edge. In the wake of In re Cellect (2023) and Allergan v MSN (Fed. Cir. 2024), obviousness-type double patenting is measured by expiration date, including adjustment, so a terminal US patent that a valuation leans on for its late expiry can in some family configurations be vulnerable to invalidation by an earlier-expiring relative.

Allergan limited the damage, holding that a first-filed, first-issued, later-expiring parent cannot be invalidated by a later-filed, later-issued, earlier-expiring child sharing a priority date, and that the original patent sets the maximum period of exclusivity.

But the practical instruction stands: before a model rests the US terminal date on the latest-expiring, extension-boosted patent in a family, confirm that patent is not the one exposed to a double-patenting challenge.

In Europe the far edge is a computed SPC whose final months are degraded by the manufacturing waiver, not the base patent; in the US it is a single chosen patent extended under Section 156, capped at five years and at fourteen from approval. The two are built differently and rarely land on the same date.

Contractual expiry versus economic expiry: where the valuation risk lives

The critical analytical distinction is between the contractual right to receive royalties and the economic value of that right given competitive dynamics.

A royalty that survives patent expiry but sits on a product facing generic or biosimilar competition is worth far less than its pre-LOE headline rate implies. The SPC manufacturing waiver, above, is one instance of this gap opening inside the protected term. Generic and biosimilar erosion is the larger one.

Generic erosion dynamics

The post-LOE erosion profile is determined primarily by molecule type and competitive entry dynamics.

The contractual right to a royalty can persist long after the revenue that backs it has eroded. Small molecules lose most of their value within two years of loss of exclusivity; biologics decline over five to seven. Economic expiry, not contractual expiry, sets the value of the tail.

Small molecules. Revenue erosion of 70 to 90 percent typically occurs within the first 12 to 18 months of generic entry in the US. Depth and speed correlate strongly with the number of entrants: a first generic (often with 180-day exclusivity under Hatch-Waxman) may capture modest share at a limited discount, but a second and third entrant trigger a competitive collapse in pricing.

In markets with ten or more generic competitors, prices can fall by 90 percent or more. The curve is further shaped by formulation complexity (oral solids erode faster than complex injectables), payer substitution policy, and brand loyalty. Lipitor (atorvastatin) declined near-vertically; Advair (fluticasone with salmeterol), with its complex inhaler, followed a much shallower path.

Biologics. Biosimilar erosion follows a different dynamic. Typical year-one erosion is 20 to 40 percent, with the full curve extending over three to seven years. The slower pace reflects manufacturing complexity limiting entrants, the absence of automatic substitution in most jurisdictions, switching inertia, and reference-product rebate strategies.

The Humira (adalimumab) biosimilar wave, the largest biologic LOE to date, showed that even with multiple entrants the reference product can hold meaningful share for years through formulary management and patient programmes.

Erosion modeling approaches. Three are standard.

First, empirically derived price-erosion curves matched by therapeutic area, formulation, and competitive landscape. A 2025 analysis of 140 originator drugs found that post-LOE decline follows a predictable but heterogeneous trajectory best modeled with a three-parameter exponential decay, with subgroup variation by year of entry, class, and product features.

Second, Bass diffusion models, taking the generic entrant's perspective and modeling uptake through innovation and imitation coefficients (a p of 0.03 and q of 0.38 baseline, calibrated to analogs).

Third, Monte Carlo simulation of generic entry timing, layering distributions over the date of first entry, the number of entrants, and price and volume impact, which is particularly valuable for biologics where entry timing is less certain.

The IRA as a non-patent LOE event

The Inflation Reduction Act of 2022 introduced a non-patent mechanism for revenue reduction that any forward-looking duration model must incorporate. Medicare price negotiation applies to small molecules selected 9 years post-approval and biologics 13 years post-approval, with the negotiated price applying to Part D and, from 2028, Part B. Where Medicare is a significant share of volume, negotiation creates a revenue inflection independent of patent status that may precede, coincide with, or follow patent expiry.

The revenue base on which the royalty is calculated can therefore decline before the royalty term expires, even absent generic entry, a form of economic compression that reduces the present value of post-negotiation cash flows.

The Brulotte constraint: post-patent royalties under US law

For royalties that extend beyond patent expiry, whether through "later of" clauses, fixed-term structures, or perpetual terms, US enforceability has been an active area of litigation, and the September 2024 Third Circuit decision in Ares Trading S.A. v. Dyax Corp. is the most significant clarification in a decade.

The Brulotte and Kimble framework

The Supreme Court's 1964 decision in Brulotte v. Thys Co. held that royalties calculated on post-expiration use of a patented invention are unenforceable per se, an impermissible extension of the patent monopoly.

The 2015 Kimble v. Marvel Entertainment decision affirmed and clarified this, identifying permissible structures: deferred payments for pre-expiration use amortized over a post-expiration period; royalties running until the latest-running patent in a portfolio expires; royalties tied to non-patent rights with appropriate structuring; and arrangements involving equity or joint ventures that allocate commercialisation risk.

The Ares v. Dyax decision (3d Cir. 2024)

Dyax had a phage-display antibody-fragment platform. In 2006 Ares entered a collaboration under which Dyax used the platform to identify fragments targeting PD-L1, and licensed Ares the platform (CAT) patents.

Ares agreed to pay royalties on products developed from Dyax-discovered antibodies until the later of patent expiry and ten years after first sale. The CAT patents expired in 2018; Bavencio launched in 2017; the obligation therefore ran to 2027, nine years past patent expiry, with no step-down.

The Third Circuit held Brulotte inapplicable, reasoning that the royalty was not calculated on activity requiring post-expiration use of the CAT inventions. Ares never practiced the licensed patents to make Bavencio; manufacture and sale did not practice the phage-display inventions. The royalties were effectively reach-through royalties, compensation for a product whose discovery was enabled by the patented technology but whose manufacture does not itself practice it.

Implications for duration modeling

Two practical consequences. First, for reach-through royalties and platform licences (antibody discovery, mRNA, gene therapy, AI-driven discovery), the Brulotte constraint does not apply where the royalty-bearing product does not itself practice the licensed patents, so parties may agree terms extending well beyond patent expiry without a step-down.

This expands the available duration of royalty streams in platform deals.

Second, for direct product-patent licences, the standard solution remains the post-patent step-down. As Freshfields noted, in deals involving both patent and know-how it is customary to reduce royalties at patent expiry, with the residual rate attributed to know-how, typically a step-down of 30 to 50 percent.

Modeling these structures requires a two-rate framework: the full rate through patent expiry, the stepped-down rate to the end of the contractual term.

The modeling framework

Step 0: Fix the near edge

Before modeling when the royalty ends, fix when it starts. Establish, per jurisdiction, the expected first-commercial-sale date that triggers the contractual term. Separately establish the investor's economic clock, the closing or funding date, and price the dead zone between funding and first sale where they differ.

Apply the quarterly arrears lag to every receipt, and resolve who owns any accrued-but-unpaid stub as of closing. For structures with a milestone phase ahead of royalties, model the two phases separately rather than with one blended assumption.

Step 1: Identify the contractual trigger

Read the agreement. Classify the expiry architecture as patent-linked, fixed-term, hybrid, return-capped, or perpetual. If patent-linked, determine precisely how "relevant patent" is defined. The Genmab-Janssen arbitration shows that this single definitional question can shift duration by five or more years on a multi-billion-dollar product.

Step 2: Build country-by-country expiry schedules

For patent-linked and hybrid structures, construct a jurisdiction matrix (US, EU5, Japan, and rest of world at minimum) mapping each right.

Key sources are the FDA Orange Book (US patents and exclusivity), the EPO and national SPC registers (EU patents and SPCs), and the Japan Patent Office. Anchor each patent's nominal expiry on its earliest non-provisional or PCT filing date, not the grant date and not a provisional or foreign priority date, then layer the adjustments.

In the US, add any patent term adjustment for prosecution delay and confirm no terminal disclaimer caps it, then apply the Hatch-Waxman PTE (capped at five years, and unable to push effective patent life beyond fourteen years from approval) and confirm the extended patent is not exposed to an obviousness-type double-patenting challenge. In Europe, do not assume "patent plus five."

Compute each SPC from the patent filing date and first EEA authorisation, add the paediatric extension only where the conditions are met, confirm the SPC will actually issue under the Article 3(a) case law (especially for combination products and second-use claims), and apply an accelerated erosion assumption to the final stretch to reflect the manufacturing waiver.

Step 3: Determine the binding constraint per country

Apply the "later of" or other contractual logic to each jurisdiction. The binding constraint may differ by country. In a typical schedule, the US constraint might be the PTE-extended patent plus paediatric exclusivity, while Japan's is the 13-year fixed term from first sale because the local patent expires earlier, and the EU's is the SPC as computed and validated above.

Step 4: Model generic or biosimilar entry timing

Construct a probabilistic model of first generic or biosimilar entry per major jurisdiction. Inputs include the number of ANDA or biosimilar filings expected, historical Paragraph IV success rates in the therapeutic area, the strength of secondary patents, settlement patterns (including agreed early-entry dates, as with the Merck-Januvia settlement permitting generic launch in May 2026), and at-risk launch probabilities. Run Monte Carlo simulation to generate a probability-weighted distribution of entry dates. The divergence between nominal patent expiry and probable LOE, which can be three to seven years for a strong patent thicket or negative two years for a portfolio vulnerable to inter partes review, is a critical source of model sensitivity.

Step 5: Apply post-LOE erosion curves

Layer the erosion model onto the revenue forecast from the LOE date forward, calibrated to molecule type, formulation, therapeutic area, and competitive dynamics.

For perpetual royalties, the curve does not terminate the stream; it asymptotically approaches zero while the contractual right persists. The tail value may be minimal in absolute terms but can carry non-trivial present value if the tail discount rate is set too low.

Step 6: Construct the probability-weighted terminal cash-flow schedule

Combine Steps 3 to 5 into a single probability-weighted schedule. Pre-LOE cash flows are discounted at the base royalty rate (product-level commercial risk, counterparty credit, deal structure). Post-LOE cash flows should carry a discount premium of 200 to 400 basis points, or be modeled as probability-weighted base, upside, and downside scenarios with explicit entry timing and erosion depth per scenario. Royalty Pharma's 10-K discloses that the fair value of its financial royalty assets is calculated from projected royalties based on sell-side consensus (or management judgment where unavailable), discounted at individual per-asset rates, and classified as Level 3 because the inputs are significant and unobservable.

How the institutional royalty funds model duration

Royalty Pharma

Royalty Pharma's accounting distinguishes two classifications that embed different duration treatments.

Financial royalty assets (the majority of the portfolio) are accounted for under ASC 310 using the effective-interest method, with income accreted over the estimated royalty term. Management applies an estimated end date per asset, which may be a patent expiry, a contractual date, or, for perpetual royalties, an applied end date based on estimated LOE and competitive dynamics. These are periodically reviewed. Royalties received on fully amortized assets are recognised as "Other royalty income."

Intangible royalty assets (currently limited to the Januvia and Janumet DPP-IV patents) are accounted for under ASC 360 using straight-line amortization over expected patent lives, with an impairment test against undiscounted future cash flows. Critical estimates include demand, marketing strategy, pricing and reimbursement, and country-specific patent expiry dates.

The Q1 2025 10-Q provides asset-level estimates. The Vertex CF franchise is perpetual with an applied accretion end date around 2035 that is periodically reviewed; the Alyftrek royalty is perpetual with an estimated 2039 to 2041 duration; the Erleada royalty is perpetual with an applied end date.

HealthCare Royalty

HCRx's structures show a preference for capped-return architectures with explicit duration boundaries. The REGENXBIO agreement (May 2025, up to 250 million dollars) uses a bond structure paying quarterly interest from royalty and milestone revenue, with the principal representing the cap. The NANOBIOTIX deal adds a royalty tail post-cap; the GENFIT deal adds annual sales caps. In each case duration is bounded by the cumulative cap, a time limit, or both, creating a defined maximum that limits downside but also caps upside relative to a perpetual or patent-linked structure.

Duration variability: a comparative summary

Deal architecture determines where duration risk sits. Fixed-term synthetics buy near-certain duration with little exposure to sales velocity; return-capped notes trade certainty for a bet on how fast the product sells; perpetual foundation royalties have the least defined end of all.

Conclusion

A royalty is a window with two contractual edges, and the value lives in the width and the position of the window, not in any single headline number. The near edge is a launch date and an arrears lag, both of which quietly defer cash against the investor's clock and neither of which appears in a stated "fifteen-year term."

The far edge depends on which of the five architectures the deal employs, on which country you are modeling, on what the patent and SPC landscape looks like in that country, on how generic or biosimilar competition develops, and, in several landmark cases, on what an arbitration tribunal or federal court decides "relevant patent" means.

In Europe, the far edge is usually an SPC, and the SPC is the input most often modeled wrong. It is computed rather than fixed, capped, conditional on the timing of authorisation, degraded in its final stretch by the manufacturing waiver, and, for combination and second-use products, at real risk of never issuing under the Article 3(a) case law.

A model that writes "patent plus five" for European assets is not conservative, it is simply inaccurate.

The analytical edge in royalty valuation lives in the spread between contractual expiry and economic expiry, and in the honest treatment of the near edge that most models skip.

A perpetual royalty on a small molecule facing imminent LOE has a contractual duration of infinity and an economic duration of perhaps two to three years. A return-capped note on a pre-launch biologic has no defined contractual duration and a dead zone at the front that a fifteen-year headline conceals.

The institutional funds have built serious infrastructure around all of this: probabilistic patent models, Monte Carlo generic-entry simulations, jurisdiction-by-jurisdiction expiry matrices, and periodic re-estimation embedded in their reporting.

For everyone else, whether a biotech CFO structuring a first royalty transaction, a university tech transfer office negotiating licence terms, or an investor evaluating an acquisition, the starting point is the agreement itself, not any assumption about what "royalty term" typically means.

Because in pharmaceutical royalty financing, there is no typical.

All information in this article was accurate as of the research date and is derived from publicly available sources including SEC filings, court opinions, EU and US regulatory and legislative materials, company press releases, and professional-services and academic literature. Information may have changed since publication. This content is for informational purposes only and does not constitute investment, legal, or financial advice. The author is not a lawyer or financial adviser.