When a Royalty Is Not a Royalty

Not Every Royalty Glitters

A payment for a trademark is not automatically a royalty. What was actually transferred, how it was paid for, and to whom, decide the tax treatment. The label in the contract does not.

A recent decision of the Indian Income Tax Appellate Tribunal in Novartis AG (ITA 1027/Mum/2023, 12 October 2026) makes this concrete. Novartis transferred three Indian trademarks to USV, Cipla and Eris for a total of ₹433 crore. The tax officer, upheld by the Dispute Resolution Panel, saw a royalty taxable in India. The Tribunal saw a sale, taxable only in Switzerland under the treaty. Same facts, two readings. The difference lay in one question: what was actually sold?

This article looks at that question twice: first through the eyes of the company that transfers a trademark or patent, then through the eyes of the tax inspector who must judge what really happened.

The framework used here (what was sold, how it was paid for, and where the gain goes) follows Pramod Kumar, former Vice President of the Tribunal, in his LinkedIn post A Treaty That Chose to Speak and his article What Was Sold, and How It Was Paid For (October 2026). His reading of the treaty and its Protocol is my starting point, and some practical points overlap with the lessons he draws. The auditor's perspective, the checklists for the company and the inspector, and the sections on transfer pricing and on substance are entirely my own, as are any errors.

Before anything else: who is the buyer?

Reading the Novartis order, the first question I asked myself as a former inspector was not about the treaty at all. It was whether the buyer was related to the seller. If it is, the price becomes the first issue, before any question of classification.

• Related buyer. A sale of a trademark between associated enterprises falls under India's transfer-pricing rules (sections 92 to 92F of the Income Tax Act, 1961), and the seller's country applies the arm's length principle too. India will test whether the buyer overpaid, because a purchased trademark is depreciable and an inflated price means inflated deductions. The seller's country will test whether the seller undercharged. Both look at the same price from opposite ends, so the risk of double taxation is real. The OECD Transfer Pricing Guidelines (Chapter VI) approach the price through what each party could realistically have done instead: keep licensing, or take a licence.
• Independent buyer. A price agreed between unrelated parties is the best evidence that it is at arm's length. In Novartis the three buyers were described as independent purchasers and the order does not suggest otherwise, so the transfer-pricing question did not arise.

One related-party thread remains even in a sale to independent buyers. The order mentions earlier co-marketing arrangements with Indian group entities, without naming them. Kumar notes that where a sale ends such an arrangement, the position of the group entity concerned may deserve attention. Taking that a step further in my own terms: in transfer pricing this is a business restructuring (OECD Guidelines, Chapter IX): did the Indian group entity give up something of value, and should it have been compensated?

Who built the market? The medicines were already sold in India under these names, and the buyers had to keep the names. The price therefore pays not only for a word mark but for an established market position; the USV deed transfers the marks expressly with their "embedded goodwill". For the treaty that supports a sale, since goodwill is not a royalty-type right. Who built that market is a separate question, and two groups of Indian parties need to be kept apart. The first are the licensees: USV sold Jalra and Cipla sold Vysov as Novartis's licensees, and for USV the order records that the packaging showed Novartis AG as owner and USV as marketer. As independent buyers, they could weigh their own contribution in the negotiation, so their share could be priced in. The second are the Indian group entities with which, according to the order, there were earlier co-marketing arrangements; the order does not name them. Here my analysis goes beyond Kumar's remark. Under the OECD guidance on marketing intangibles (Chapter VI), a group entity that does more marketing than an independent distributor would may be entitled to part of the return on the brand. If that applied here, part of the price could arguably have been due to an Indian taxpayer, taxable in India regardless of the treaty. The order does not address this; it is a question for the next file, not a criticism of this one.

What was actually sold?

Three questions decide the classification: does ownership pass or only use, is the price fixed or dependent on sales or use, and what else is in the package?

Ownership or use. A licence grants a right to use; the owner remains the owner. In a transfer the seller loses the right, even if only for one country. The OECD Commentary on Article 12 accepts that a geographically limited right can be distinct, transferable property. Restrictions in the deed then do not by themselves show that the seller kept ownership, as long as they describe what was sold or protect what the seller kept elsewhere.

Fixed or contingent. The Protocol to the India–Switzerland treaty makes the price, in Kumar's word, the hinge. The deed decides whether there was an alienation. If there was, gains contingent on profits, productivity or use may still be taxed under Article 12, even after a genuine transfer, although the Protocol does not call them royalties. A fixed price goes to Article 7 or 13.

What is in the package. A pharmaceutical brand transfer rarely covers a bare word mark. Each component has its own classification:

Indian domestic law draws the line elsewhere. Explanation 2 to section 9(1)(vi) treats even the "transfer of all or any rights" in a trademark as royalty, unless the income is chargeable as capital gains. Read through domestic law alone, a royalty appears sooner than the treaty allows.

The company: make the facts carry the label

For the seller the question is not only whether it is a sale, but whether it can still prove that five years from now. The classification is won when the deed is drafted, not at the appeal.

1. Let the deed speak of transfer. A clause stating that the seller gives up all ownership and all rights within the territory. In Novartis that was clause 1.2, and the Tribunal relied on it.
2. Write restrictions as boundaries. "Sold are the Indian registrations for these preparations" describes the property. "The buyer may use the mark only for…" reads as a licence condition. Same content, different reading.
3. Record the assignment. Have the buyer entered as proprietor in the trademark register. In Novartis the Tribunal treated the register entry as independent corroboration of the deed. The order records it only for USV; for the other two buyers it reached the same conclusion on the deeds alone.
4. Keep the price fixed. An earn-out or instalments linked to sales bring that part within Article 12. If a variable part is commercially needed, split it contractually and accept Indian tax on that part.
5. Allocate the price over the components. Trademark and goodwill, dossiers, know-how, services and inventory, each with its own amount, supported by a valuation. Without an allocation the inspector makes one, or pulls everything into Article 12.
6. Close the old relationship. Terminate and settle earlier licence or co-marketing arrangements separately, and document whether any group entity that loses its role is compensated. Otherwise the price looks like the continuation of the old fee, which is what the tax officer argued in Novartis.
7. Let go of what was sold. No buy-back option, no veto, no quality control of the kind a licensor keeps. A seller who keeps the strings has not sold in economic terms. That is where India's general anti-avoidance rule (GAAR) comes into view; as Kumar notes, the India–Switzerland treaty itself has no principal purpose test.
8. With a related buyer, document the price. A transfer-pricing analysis on both sides of the border, ideally with the same valuation, and an explanation of why the price matches what each party could otherwise have done.
9. Build the file. Tax residency certificate and Form 10F, board resolutions, the commercial reason (such as divesting mature brands), and confirmation that there is no Indian permanent establishment. Where the buyer must decide whether to withhold tax, obtain a nil or lower withholding certificate before payment.

Two arithmetic notes for the company. A sale is mainly cheaper than a licence where India's withholding tax would otherwise not be creditable at home, for example because of losses or a low home rate. And for groups with revenue of €750 million or more, the 15% minimum tax (Pillar Two) largely evens out the difference. The choice then mainly decides which country taxes, not how much. My companion article License or Sell? works this out in figures.

The inspector: start with the facts, not the outcome

In Novartis the Revenue's case rested on one proposition: that the restrictions in the deeds meant there was no sale. The Tribunal tested that proposition against the operative clauses one by one and rejected it. The lesson for an inspector is not that the Revenue had no case. Kumar gives its argument a fair hearing and notes that the obligations to keep the name and composition unchanged come closest to a licensor's control. The lesson is that an argument built on one proposition stands or falls with it. A stronger approach starts with what actually passed.

Form, substance, and what a deed can hide

The Tribunal began by noting that calling a deed an assignment is not conclusive, and it looked beyond the label: the register, the new packaging, the stamp duty, and the transfer of regulatory duties such as recalls and adverse-event reporting. Most of that evidence, however, comes from documents the parties drew up themselves. Independent evidence, the register entry and the packaging, is on record for one of the three buyers only. A deed can be drafted to tell the story the parties need. An inspector therefore tests it against what third parties and the market show.

In Novartis the public facts point towards substance. All three marks were brands of vildagliptin, the diabetes medicine Novartis itself sells as Galvus. The Indian patent expired on 9 December 2019; the deeds are dated 22 and 30 November 2019. After expiry, prices fell by more than 70% according to the trade press. Novartis sold the brands just before generic competition eroded their value, which is a clear commercial reason. Two of the buyers, USV and Cipla, had marketed the same brands as Novartis's licensees; the third mark, Zomelis, had been marketed by Abbott and went to Eris. That is the conversion of a licence into ownership. Kumar observes, in general terms, that converting an existing commercial relationship into a sale is where substance is most often questioned. He does not direct that remark at the Novartis buyers, and neither do I.

Where a deed could still steer the outcome, these are the questions I would ask:

• Supply. Does the seller continue to supply the active ingredient or the finished product? Then it keeps an economic hold, and part of the "price" may be an advance on the supply relationship.
• Settlement of the old licence. Are outstanding royalties or a licence buy-out folded into the price? That part is royalty, not sale proceeds.
• The regulator. Who holds the marketing authorisation after the transfer? Like the register, that is third-party evidence.
• Retained controls in practice. Who enforces the obligations to keep the name and composition unchanged, and has that ever happened?
• Price against value. Does the price match the expected sales after patent expiry? A price well above the present value of future royalties suggests something else is in the package.

India applies substance over form narrowly. Since Azadi Bachao Andolan and Vodafone, the courts look at a transaction as a whole but set it aside only if it is a sham or artificial; the statutory route to substance is GAAR, in force since 2017. With independent buyers and a patent expiry as the trigger, there was little ground for that in Novartis. The questions above remain the right test for the next case.

What to ask for

• The complete deed with all schedules, side letters and transitional agreements, not just the main text.
• Earlier licence, distribution or co-marketing contracts with the same buyer or its group, and how they ended.
• The trademark register extract for each mark, for each buyer.
• The valuation or fairness opinion behind the price, and its allocation over the components.
• With a related buyer: the transfer-pricing documentation on both sides.
• The buyer's accounting: an intangible asset with amortisation, or an expense.

What to test

1. Is the buyer independent? With a related buyer, transfer pricing comes first: is the price what independent parties would have agreed? Then GAAR. With an independent buyer and a business reason, an abuse case is very hard to make, and in Novartis none was made.
2. Did ownership really pass? Who enforces against infringement, who pays renewals, who may sub-license, is there a reversion right? The more of that stays with the seller, the more it resembles a licence.
3. Is the price really fixed? Look for milestone instalments, later price adjustments and set-off against future supplies. A disguised variable part falls within Article 12 under the Protocol.
4. What else is inside? Dossiers, product information, know-how and transitional services may fall within Article 12 under this treaty. Splitting the price over those components is harder to rebut than an all-or-nothing position.

What does not work

• Territorial and product restrictions as proof of retained ownership. They define the property, or protect what the seller keeps elsewhere.
• The absence of a cost in the books (see the next section).
• Setting the domestic royalty definition against the treaty. The treaty prevails.

The mirror test. The tax authority cannot have it both ways. If the seller's receipt is a royalty, the buyer should expense it. If the buyer amortises the trademark as an acquired asset, it confirms the sale. A consistent reading of both sides of the transaction is the strongest file an inspector can build, and the fairest.

The balance sheet is silent; the register speaks

That Novartis had no cost for the trademarks in its books proves nothing about their existence. It follows directly from the accounting rules.

Self-created: never on the balance sheet. IAS 38.63 prohibits recognising internally generated brands, mastheads, publishing titles and customer lists. The spending cannot be separated from building the business as a whole. India's Ind AS 38 follows the same line, as does Dutch GAAP (RJ 210).

Patents are more nuanced. Research is always expensed (IAS 38.54). Development costs must be capitalised once six criteria are met (IAS 38.57), including technical feasibility and demonstrable future benefit. In pharmaceuticals that threshold usually lies around marketing approval, so little is capitalised.

Acquired: on the balance sheet. A purchased trademark is recognised at cost (IAS 38.24–25), or at fair value in a business combination (IFRS 3). That is why group balance sheets are full of acquired brands and almost never show home-grown ones.

For Novartis this means:

• At the seller the trademark stands at nil while it is worth ₹433 crore. The hidden reserve is realised in one go on sale. For tax, section 55(2)(a) follows the same logic: nil cost, so the entire proceeds are gain.
• At the buyer the trademark suddenly appears, at its purchase price. In the accounts it may have an indefinite useful life: no amortisation, only an annual impairment test. For tax the buyer meanwhile writes it down at 25% declining balance. The same trademark leads two lives, and the gap shows up as deferred tax.

For the inspector, "no cost, so not proven" confuses two things. Non-recognition is a rule about measurability, not a finding about existence. The existence of the right is shown by the trademark register, not by the balance sheet. The Tribunal said the same: the cost of the mark for computing a gain is a separate question from whether it was alienated. What the nil cost does show: there is no cost base to soften the tax, which makes the classification an all-or-nothing question.

Where the two views meet

Company and inspector look at the same transaction and search in opposite directions. The company leans on the deed and the register; the inspector on economic continuity: the same brand, the same market, the same buyers, and a price equal to the present value of future royalties. Which view decides the outcome depends on the treaty.

Under the India–Switzerland treaty both matter, in sequence. The deed and the register, read as a whole, decide whether there was a sale. Once there was, the Protocol lets the payment terms decide where it is taxed. Under the US and UK treaties the tax officer would have had an easier case, since India keeps the right to tax the capital gain there, subject to the question, left open in Novartis, of whether an Indian-registered mark is a capital asset situated in India.

For large groups, the 15% minimum tax has applied since 2024. It largely neutralises the choice between licence and sale in the residence country. The contest over classification then becomes a contest between States, not between taxpayer and tax authority. That does not make the facts less important: they still decide which country collects.

The same questions, from both sides

Conclusion. A royalty is not always a royalty, and a sale is not always a sale. The label in the contract glitters, but tax authorities and courts look at what lies beneath: who the buyer is, who holds the rights after the transaction, how it was paid, and what else went with it. For the company that means: build the facts so they carry the label. For the inspector: test the facts before choosing the outcome. In Novartis the Revenue's case rested on the restrictions alone, and the deeds and the register outweighed them.