Financial Reporting
Your Code Isn't on the Balance Sheet: How Startup IP Actually Gets Valued
US GAAP expenses the R&D behind your most valuable asset, so it barely hits the balance sheet. How appraisers really value startup IP, with worked numbers.
By 409.AI Team - 2026-09-07
# Your Code Isn't on the Balance Sheet: How Startup IP Actually Gets Valued
Picture a company that just closed a $40 million round on the strength of two things: a model the team spent four years training, and a patent family covering how it gets deployed. Then the new controller circulates the balance sheet. Total assets: cash, a receivable, some laptops, and prepaid cloud credits.
The thing investors just paid $40 million for does not appear anywhere on it.
That isn't a bookkeeping mistake. It's how US GAAP works, and it's why almost every founder is caught off guard the first time somebody asks what their intellectual property is worth. The answer is never the amount you spent building it.
Why your best asset has almost no book value
ASC 730-10-25-1 requires that research and development costs be [charged to expense when incurred](https://kpmg.com/kpmg-us/content/dam/kpmg/frv/pdf/2023/handbook-research-and-development.pdf). The future benefit is too uncertain to sit on a balance sheet, so the accounting answer is to run it through the income statement and move on. Software gets a narrow exception: [ASC 350-40](https://www.eisneramper.com/capitalizing-internal-use-software-0123/) lets you capitalize certain internal-use software costs, but only in the application development stage, and only once management has authorized and committed funding to a project that is probable of completion. Preliminary-stage costs are expensed as incurred, and most of what an engineering team does falls outside the capitalization window entirely.
The tax code takes a similar position from a different direction. [IRC Section 197](https://www.law.cornell.edu/uscode/text/26/197) gives you a 15-year amortization deduction for patents, know-how, trademarks, customer lists, and goodwill, but the deduction attaches to *acquired* intangibles. Self-created ones are generally excluded unless they come along with the acquisition of a trade or business.
So the accounting system and the tax system both agree: what you build yourself is invisible until somebody buys it.
Meanwhile the economy has gone entirely the other way. Ocean Tomo's [2025 Intangible Asset Market Value Study](https://www.prnewswire.com/news-releases/ocean-tomo-releases-2025-intangible-asset-market-value-study-results-302686446.html), released in February 2026, put intangibles at roughly 92% of S&P 500 market capitalization, against 17% in 1975. Tangible assets are down to about 8% of what public markets pay for. Private companies are no different. The value is real. It's just that nobody writes it down until a transaction forces the question.
The moments that force the question
You rarely commission an IP valuation because you're curious. Something triggers it:
Somebody acquires you. The buyer has to split the price across identifiable assets under ASC 805, and every intangible that meets the separability or contractual-legal test comes out of goodwill and onto the buyer's books at fair value. That exercise is the [purchase price allocation](https://409.ai/articles/asc-805-purchase-price-allocation-startup-acquisition), and it's where most founders meet an intangible appraiser for the first time. If part of the price is contingent, the [earnout gets valued too](https://409.ai/articles/earnout-contingent-consideration-valuation-asc-805).
The acquisition doesn't work out. Those intangibles now have carrying values that have to be tested. A technology asset that stops generating cash becomes [an impairment charge](https://409.ai/articles/goodwill-impairment-private-company-asc-350), and the write-down amount is a valuation conclusion.
You move IP into a holding entity or a trust. Gift and estate transfers need a supportable fair market value on the transfer date, which is the same discipline as [gifting startup equity](https://409.ai/articles/gifting-startup-equity-2026-estate-tax-exemption) and gets the same IRS scrutiny.
A big company wants your model but not your company. The [licensing-plus-hiring structures](https://409.ai/articles/reverse-acqui-hire-cap-table-409a-valuation) that became common over the last two years turn on one number: what the license itself is worth, separate from the team.
A lender takes IP as collateral, or you end up in litigation. Both need a number somebody can defend line by line.
Three approaches, and what each is honestly good for
There is no single method. An appraiser picks per asset, and a real report usually applies two or three across the intangibles it identifies. ASC 805-20-55-13 sorts identifiable intangibles into five families: marketing-related, customer-related, artistic-related, contract-based, and technology-based. Each one has to be [separable or arise from contractual or legal rights](https://www.gaapdynamics.com/intangible-assets-asc-350-and-business-combinations-asc-805) to come out of goodwill at all, and different families call for different math.
Relief from royalty: trademarks, trade names, patented technology
The logic is a thought experiment. If you didn't own this patent, you'd have to license it. What would that cost, and what are you therefore saving?
Say your patented process supports $12 million of revenue this year, growing to about $22.8 million by year five. A market royalty rate of 5% produces savings of $600,000 in year one, rising to roughly $1.14 million. Tax those savings at 21%, discount them at 16% on a mid-year convention, and the first five years are worth about $2.4 million in present value.
Two things then adjust that. The model should run to the end of the asset's economic life, not stop at five years, so a patent with real runway is worth considerably more than the five-year slice. And because a buyer amortizes the acquired asset over 15 years under Section 197, there's a tax amortization benefit worth about 9% at these inputs. Round numbers, roughly $2.6 million for the five-year window, more once you extend the life.
The whole method lives or dies on the royalty rate. Pulling 5% out of the air is not a valuation. Defensible rates come from observed licensing agreements for comparable technology in comparable industries, and the appraiser has to explain why the comparison holds.
Multi-period excess earnings: customer relationships and core technology
Use this when the asset generates identifiable cash flow but needs other assets to do it. This is where founders get surprised.
Take a customer base producing $18 million of revenue in year one, attriting at 20% annually, at a 25% EBITDA margin. Year one contributes $4.5 million of earnings. But those customers don't earn anything without working capital, servers, the assembled workforce, and the trade name. So you charge the asset rent for using them. At a contributory asset charge of 10% of revenue, year one drops to $2.7 million of excess earnings, $2.13 million after tax. Run that out five years, discount at 17%, and you land near $5.3 million.
Founders consistently expect the $4.5 million number and get the $2.7 million one. The contributory asset charges are not a haircut somebody applied to be conservative. They are the recognition that a customer list, on its own, does not deliver a product.
Cost to recreate: early-stage software and assembled workforce
The last resort, and the one most often misused. You estimate what a market participant would spend to rebuild the asset. Twenty-two engineer-years at a fully loaded $240,000 each is $5.3 million. Add a developer's profit of 12% and you're at $5.9 million. Subtract obsolescence for the parts already stale, call it 25%, and you're around $4.5 million.
Notice what this method never asks: whether the asset earns anything. A company can spend $30 million building a platform worth $2 million to any buyer. Cost is a floor and a sanity check, not an answer, and an appraiser who leans on it for a revenue-generating technology asset is going to have a difficult conversation with your auditor.
Where founders' numbers go wrong
The most common error is treating R&D spend as value. Your burn rate tells you what building it cost, which is a fact about the past. Value is a claim about future cash flow, and the two are only loosely related.
The second is using headline licensing deals as comps. A public announcement almost never includes the royalty structure, the field-of-use limits, the exclusivity terms, or the equity that changed hands alongside it. Two deals with the same press-release number can be worth very different amounts.
The third is assuming IP value tracks your post-money valuation. Your last round priced preferred stock with liquidation preferences and control rights attached, which is exactly why [409A valuations land below post-money](https://409.ai/articles/income-approach-409a-valuation) and why an IP appraisal isn't a slice of your cap table either. Different asset, different question.
And a quiet one: unenforced patents. A patent you've never asserted, never licensed, and can't afford to litigate still has value, but a market participant discounts it heavily. Appraisers ask about enforcement history for good reason.
What you'll be asked for
Expect a document request covering the patent and trademark schedule with filing dates and jurisdictions, every inbound and outbound license, revenue broken out by product line and by the IP that drives it, a five-year forecast with the assumptions written down, R&D headcount and spend by project, customer contracts with renewal terms and churn history, and any prior appraisals or offers received.
Most of the delay in these engagements is the revenue attribution. Founders know total revenue cold and have rarely traced which product line depends on which patent. Doing that work before the appraiser asks will save you two weeks.
One last point worth internalizing: these valuations are almost entirely built on unobservable inputs, which puts them squarely in [Level 3 of the fair value hierarchy](https://409.ai/articles/asc-820-level-3-classification-significant-unobservable-input). That isn't a weakness in the work. It's a signal about how the conclusion gets reviewed, because Level 3 measurements draw the most audit attention and carry the heaviest disclosure requirements. The report has to survive somebody reading every assumption and asking where it came from.
If you're heading into an acquisition, a trust transfer, or a licensing negotiation, the number you'll need doesn't come from your accounting system. Start the attribution work now, while the answer is still yours to document rather than defend. 409.ai's [IP and intangible asset valuation](https://409.ai/products/ip-valuation) is built for exactly that point in the process.