R&D Accounting Under US GAAP: Why It Is Expensed

Introduction

Merck earned $365 million of net income in 2023, down from $14.5 billion the year before, even though its sales grew. The difference was research and development (R&D). Merck's R&D expense more than doubled, to $30.5 billion, largely because of a single $10.2 billion charge for buying Prometheus Biosciences, a biotech company whose main asset was an experimental drug for inflammatory bowel disease. Merck paid for a pipeline and, under US GAAP, recorded almost the entire price as an expense in the quarter the deal closed.

The treatment follows from one of the most consequential rules in US accounting: research and development is expensed as incurred. A pharmaceutical company spending billions on drugs it expects to sell for decades carries none of that spending on its balance sheet, while a company that buys a finished product through a business combination records it as an asset. The rule shapes the earnings, balance sheets and returns of every research-intensive company, and the exceptions to it explain some of the largest swings in reported profits.

Accounting Standards Codification (ASC) 730 requires research and development costs to be charged to expense when incurred. The rule dates from 1974, when the Financial Accounting Standards Board (FASB) concluded that the future benefits of most research are too uncertain to measure reliably, and that a single, simple rule would produce more comparable statements than case-by-case judgments about which projects might succeed. It was one of the FASB's first standards, and the board's role in setting US rules is described in who sets accounting rules.

Research and development

Research is planned investigation aimed at discovering new knowledge for a new or significantly improved product, process or service; development is the translation of that knowledge into a plan or design. Under US GAAP, both are generally expensed as incurred under ASC 730.

R&D expense includes the salaries of scientists and engineers, materials and supplies used in research, payments to contract research organizations, and an allocation of overhead. Equipment and facilities used in research that have alternative future uses, such as a laboratory building, are capitalized and depreciated, with the depreciation reported in R&D expense; equipment bought for a single project with no other use is expensed immediately. The broader logic of when costs become assets is covered in capitalizing vs expensing costs.

Three groups of costs need a closer look, and each matters in a different industry: software has its own rules, and research bought or licensed from others can end up either as an expense or as an asset.

Software

Software developed to be sold is governed by ASC 985-20: costs are expensed as R&D until the product reaches technological feasibility, and capitalized afterwards until it is released. Because most software companies reach technological feasibility only shortly before release, little is usually capitalized. Software developed for internal use, including the platforms behind many online services, follows ASC 350-40 and is capitalized once management has committed to funding the project and its completion is probable, which allows considerably more capitalization.

R&D Acquired From Others

When a company buys R&D rather than performing it, the accounting depends on what it buys. In a business combination, the acquisition of a business under ASC 805, in-process research and development (IPR&D) is recorded as an indefinite-lived intangible asset at fair value, tested for impairment and amortized once the project is completed. Pfizer's acquisition of Seagen put $20.8 billion of IPR&D on its balance sheet this way, a provisional figure that its 2024 Form 10-K finalized at $19.9 billion. The whole deal is traced in how an acquisition hits the three statements.

In an asset acquisition, such as a deal in which substantially all of the value lies in a single asset or group of similar assets, acquired IPR&D with no alternative future use is expensed immediately, just as internal R&D would be. That is what happened with Prometheus. Merck's 2024 annual report explains that the deal was treated as an asset acquisition because the lead drug candidate accounted for substantially all of the fair value of the gross assets acquired, leading to the $10.2 billion charge to R&D. The test that separates the two kinds of deal is covered in business combination vs asset acquisition.

In-process research and development (IPR&D)

Research and development projects that an acquirer obtains but that are not yet complete, such as drug candidates still in clinical trials. In a business combination they are capitalized at fair value as indefinite-lived intangible assets; in an asset acquisition, if they have no alternative future use, they are expensed.

Collaboration and Licensing Payments

Upfront and milestone payments to license another company's experimental drug before it is approved are also generally expensed as R&D, because the rights acquired have no alternative future use. Payments after approval, for a product that can be sold, are capitalized as intangible assets. In 2023, Merck also expensed $5.5 billion related to a collaboration with Daiichi Sankyo on cancer drugs, alongside the Prometheus charge.

The three years show how business development flows straight into earnings under the expense rule. R&D was 23% of sales in 2022, 51% in 2023 and 28% in 2024. Net income fell by more than $14 billion and then recovered by more than $16 billion, while the underlying research organization changed far less than those figures suggest.

Expensing R&D creates three distortions that bankers must allow for.

The Balance Sheet Understates Research-Driven Companies

A drugmaker or software company with valuable products built in-house shows few intangible assets, while a competitor that acquired similar products carries them at their purchase price. Book value and asset-based ratios are therefore not comparable between organic builders and serial acquirers.

Returns on Capital Are Overstated

Because the investment in R&D never appears in invested capital, return on invested capital for a research-heavy company can look exceptionally high. Analysts sometimes capitalize R&D for their own analysis, treating each year's spending as an asset amortized over an estimated life, which raises invested capital and changes operating income by the difference between the year's spending and the amortization of past spending. How the ratios are built is covered in return ratios: ROE, ROA and ROIC.

Earnings Depend on How R&D Is Acquired

Merck's 2023 results would have looked entirely different had Prometheus been structured, or qualified, as a business combination, with about $10 billion of IPR&D and goodwill on the balance sheet instead of in expenses. The cost would not have disappeared, only moved in time: the IPR&D would have sat on the balance sheet until approval and then turned into amortization over the drug's commercial life, or into an impairment charge if the drug failed. Two buyers of similar pipelines can therefore report very different earnings in the year of purchase and in the years after it.

In valuation, bankers treat R&D as an operating cost but separate recurring research spending from one-off acquired IPR&D and licensing charges. Some companies remove such charges from non-GAAP earnings; others, including Merck, include them: Merck's 2023 non-GAAP earnings per share of $1.51 included $6.21 per share of charges for upfront payments and asset acquisitions. Comparisons across pharmaceutical companies therefore require checking each company's policy. For drug developers, value often sits largely in the pipeline, which is valued directly, as covered in the Valuation guide's article on pipeline DCF and probability-weighted NPV.

Tax treatment adds a further layer. For tax purposes, US companies had to capitalize and amortize domestic research costs from 2022 through 2024, which created deferred tax assets; immediate expensing returned from 2025, and those assets now unwind, as explained in bonus depreciation and R&D expensing. For financial reporting, the rule remains the same: research is an expense, whether it succeeds or fails, and software aside, research reaches the balance sheet mainly when it is bought in a business combination.