The Number That Has To Survive A Courtroom

Intangible assets made up 92 per cent of the market capitalisation of the S&P 500 in 2026 (Ocean Tomo). In 1975 the figure was 17 per cent. The accounting standard did not move at the same speed. There is no single accepted method for pricing a patent, and there is no published table of patent prices. So the number is chosen. It is chosen by a method, and the method is chosen by a person with an interest in the outcome. Set the number low and the balance sheet shrinks. Set it high and the auditor asks for the working papers. Set it in a licensing dispute and the other side asks for ten years of your files. Two valuations of the same patent can both be defensible and still differ by a factor of ten. The number is not the output. The number is the argument.

Cost, market and income. Three doors, six methods. Cost sums what it took to build the thing: salaries, materials, failed experiments, filing and prosecution fees, renewals. Replacement cost skips the dead ends. Reproduction cost includes them. Cost is a floor and nothing more. United States courts inside the Georgia-Pacific analysis reach for cost only as a minimum reference. German courts take cost as the lower bound in the Lizenzanalogie and then go looking for income. The gap between those two positions is not small. The mp3 patent family cost Fraunhofer a modest prosecution budget and returned licence fees in the high double-digit millions per year. A patent prosecuted for EUR 50,000 can carry EUR 50 million in licences. Cost sees the first number. Nobody buys the first number.

The market approach is comparison: licence deals, outright sales, auction results, each adjusted for scope, remaining life, legal strength and geography. Its workhorse is relief from royalty. You price the patent at the royalty you would have paid if you did not own it: rate times revenue times one minus tax, discounted over the remaining life. The published ranges: pharmaceuticals 5 to 15 per cent of net sales, biotechnology 3 to 10, medical devices 3 to 8, software 5 to 12, electronics 1 to 5, mechanical 2 to 5, chemical 2 to 5, cleantech and energy 2 to 6. The old heuristic was 25 per cent. The licensee keeps 75 per cent of operating profit and the owner takes 25. It has been criticised for two decades and it is still the opening position in a negotiation. ktMINE holds 141 million patents and more than 300,000 licence transactions. If your technology is not in the database, the comparison does not exist and the argument quietly moves back to income. Comparable licences are also the weakest sheet in the file. Two databases can show the same technology at different rates, and neither discloses the contractual terms that produced its number.

The income approach wins in practice. The big four firms, the IVSC, the AICPA and the courts all reach for it whenever data exists. The trap is the discount rate. A discounted cash flow run at a single 50 per cent rate bakes technical failure and cost of capital into one figure, and long-duration assets come out systematically cheap. Risk-adjusted net present value separates the two. Probability of success multiplies the cash flow itself. The discount rate is then free to reflect only the cost of capital. The published phase transitions are brutal. Preclinical to Phase I sits near 60 per cent. Phase I to Phase II near 60. Phase II to Phase III near 30. Cumulative probability of approval from the start of preclinical work lands near 12 per cent (IQVIA, 2025). Capitalised cost per new drug runs USD 2 to 4 billion (IQVIA, 2025). The probability that a valuation you cannot reconstruct in a deposition survives the other side is not zero. Most sellers are pricing it at zero.

The number is not the output. The number is the argument.

In the United States, 35 USC 284 sets damages at no less than a reasonable royalty, and wilful infringement can treble them. Georgia-Pacific (1970) lists fifteen factors. Apportionment has been mandatory since Lucent v. Gateway in 2009 and VirnetX v. Apple in 2015: damages attach to the patented feature, not to the whole product. Median awards moved from USD 5 to 15 million across 2010 to 2015, to 10 to 20 million across 2016 to 2020, to 12 to 25 million across 2021 to 2025. Germany and the Unified Patent Court sell something else: an injunction. An infringement finding produces one almost automatically, with no eBay-style balancing test. Three damages heads are on offer, lost profit, infringer profit and the licence analogy. That is why a German or UPC patent is worth 1.5 to 3 times a United States equivalent on a per-capita GDP-adjusted basis. The value is speed. Money can be modelled. An injunction that removes a product from a market inside months is harder to place in a spreadsheet, and it is the reason the European filing gets made.

Roughly half of litigated United States patents are invalidated (Lemer, 2008). A patent that has survived an inter partes review or a European opposition is not the same asset as one that has never been tested. The premium for surviving runs 2 to 5 times. Published stage values move with it. A granted patent in one or more jurisdictions sits at EUR 2 to 50 million. A licensed and litigation-enforceable patent sits at EUR 10 to 500 million. An enforced patent that survived a validity challenge sits at EUR 20 million to 1 billion. A global portfolio across several families sits at EUR 50 million to 5 billion. Every step up that ladder is bought with a document. The opposition decision. The claim construction. The licence that was actually paid. The price follows the paper trail.

Deloitte leads with relief from royalty for patents and multi-period excess earnings for the primary asset in a purchase price allocation. EY triangulates across all three approaches and is the busiest of the big four in standards-essential patent and FRAND work. KPMG sells an automated purchase price allocator plus an intangibles benchmark drawn from filings. PwC leans on its royalty rate and tax data. The standards underneath are IVS, ASC 805 in the United States, and the German IDW S 1 and S 5 (2023). One line in Deloitte guidance decides most arguments: the discount rate must reflect the risk of the asset, not the risk of the company. The boutiques do the testimony. Ocean Tomo runs IP auctions and expert work. FTI runs damages. Kroll runs valuation and litigation consulting. Where the valuation lands depends on who is asking. A private equity buyer, a tax authority, an impairment test and a district court will want four different numbers from the same patent. The method is chosen before the number is.

Build the method before the dispute rather than during it. Write the royalty rate down on the day the licence is signed. Keep the failed experiments in the cost file, because replacement cost is defensible and reproduction cost is a claim about your own spending. Document the apportionment before an expert does it for you. I restructured a industrial group. Entrade built 200 wood-gasification machines across the UK and Europe, and the first units never set the price while the tenth build did. I signed a personal guarantee on a company that folded on Christmas Day 2008. Those are cost files and cash-flow files. Neither one is a forecast. If the number has to survive a courtroom, it has to be reconstructible by someone who does not like you. A patent is not priced. It is argued at a number.