In 30 seconds
- R&D expense is measured mainly as a percentage of revenue, and you only compare it within the same sector.
- Under US rules almost all R&D is expensed in the year it's spent; under IFRS, development can be capitalized.
- Kaplio's stock page expects a pharmaceutical company to spend at least 15% of revenue on R&D.
- Separate in-house R&D from acquired R&D: Merck hit 50.8% of sales in 2023 because of the Prometheus purchase.
- What matters is productivity: new sales and margins for every dollar invested, measured over several years.
Introduction
Intel spent $16,546 million on R&D in 2024, 31.2% of its revenue, and closed the year with a net loss of $18,756 million. Nvidia spent less, $12,914 million in its fiscal 2025 (ended January 26, 2025), and brought in $130,497 million.
Spending a lot on research guarantees nothing. What matters is what that money brings back, and you can measure it straight from the financial statements. That's what this lesson teaches you to do.
Explanation
R&D expense (research and development) is the money a company puts into inventing or improving products and processes. At a drugmaker it's mostly clinical trials and labs. At a tech company it's engineers' salaries, prototypes and computing power. It shows up on the income statement as its own line within operating expenses, next to selling, general and administrative expenses (SG&A).
Think of a farmer who plants an apple orchard. This year he pays for the saplings, the irrigation and the labor, and he doesn't pick a single apple. The harvest comes years later, and it may be good or bad. R&D works the same way: the cost is certain and paid today; the fruit is uncertain and arrives tomorrow.
The accounting problem: expense or investment?
Economically, a good part of R&D is an investment. In US accounting, it almost never is. US rules (US GAAP, topic ASC 730) require companies to expense R&D in the year they pay for it. There are exceptions: part of software development, and research projects that come with the purchase of a whole company, which are recorded as an intangible asset.
International rules (IFRS, specifically IAS 38) split the work into two phases. Research is always expensed. Development is capitalized (put on the balance sheet as an asset and amortized over several years) once the company can show, among other things, that the project is technically feasible, that it intends to finish it and has the resources to do so, and that it will generate future economic benefits. That's why a European carmaker can put part of its development spending on the balance sheet and a comparable US company can't.
What does that change for you? Under US rules, a company that does a lot of research looks less profitable than it really is, and one that cuts research looks more profitable than it really is. Aswath Damodaran, a professor at New York University, proposed in a 1999 paper treating R&D as an investment for valuation purposes: add it to assets and amortize it. Doing so changes the operating income and the ROIC of the companies that research the most.
The three questions to ask about R&D
How much it spends relative to what it sells (R&D to revenue). Whether that effort is normal for its sector. And what past R&D has delivered in new sales and margins. The first is arithmetic. The third is the one that separates companies that research well from companies that just spend.
Formula
R&D to revenue = R&D expense / Revenue × 100
R&D expense = the "Research and development" line on the 10-K income statement
Revenue = total sales for the same fiscal year
R&D productivity (approximate) = Revenue increase over N years / Cumulative R&D of the prior years
Example
The table below compares four large US drugmakers using Kaplio's latest data. Read it with one question in mind: which ones clear the 15% that Kaplio's stock page asks of a pharmaceutical company? Two warnings before you look. Johnson & Johnson also runs a medical technology business, which spends less on research than the drug business and pulls the company's average down. And if one of the four shoots up, first check whether it bought research projects from another company that period (you'll see why under "Pitfalls and limitations").
Step by step: Eli Lilly, fiscal 2025 (figures from its Form 10-K annual report, in millions of dollars):
1. Revenue: 65,179.
2. R&D expense, excluding R&D acquired from third parties: 13,337.
3. R&D to revenue: 13,337 / 65,179 = 20.5%. Above the 15% benchmark.
4. Productivity: Lilly spent $7,191 million on R&D in 2022, $9,313 million in 2023 and $10,991 million in 2024, about $27,495 million in total.
5. Its revenue went from $28,541 million in 2022 to $65,179 million in 2025: $36,638 million more.
6. Divide: 36,638 / 27,495 = $1.33 of new sales for every dollar of R&D in the three prior years. Lilly's own 2025 10-K credits the growth mainly to Mounjaro and Zepbound, its diabetes and obesity drugs.
Your turn. Fill in the blanks with the 2025 accounts (10-K, millions of dollars):
Pfizer: R&D of 10,437 on revenue of 62,579. R&D to revenue = ___%. Hint: it lands between 16% and 17%.
Merck: R&D of 15,789 on revenue of 65,011. R&D to revenue = ___%. Hint: it's above 24%.
Before you look: if Merck devotes a bigger share of its sales to R&D than Lilly does, do you think that makes it the one that will grow the most? Think it over and keep reading.
Real-data example
| Company | Ticker | R&D as % of revenue |
|---|---|---|
| Merck & Co., Inc. | MRK | 42.7% |
| Pfizer Inc. | PFE | 20.1% |
| Eli Lilly and Company | LLY | 18.3% |
| Johnson & Johnson | JNJ | 15.4% |
How to read it
How to read it
With R&D, more isn't better: it depends on the sector and on what the spending returns. Pharma is the only sector where Kaplio's stock page sets its own bar, "≥15%", because "Serious pharma reinvests 15-20%". The other ranges in this table are rough guides, each with a real example from the sector calculated from its 10-K.
| Sector | Low | Typical | High | Example |
|---|---|---|---|---|
| Pharma (Kaplio threshold) | below 15% | 15-20% | above 20% | Lilly, 20.5% in 2025 |
| Software and internet | below 10% | 10-20% | above 20% | Alphabet, 15.2% in 2025; Microsoft, 10.7% in fiscal 2026 |
| Semiconductors | below 10% | 10-25% | above 25% | Nvidia, 9.9% in fiscal 2025; Intel, 31.2% in 2024 |
| Industrials | below 2% | 2-5% | above 5% | Caterpillar, 3.2% in 2025 |
| Consumer staples | below 1% | 1-3% | above 3% | Procter & Gamble, 2.4% in fiscal 2026 |
Look at the semiconductor row: the company that invested the most as a share of sales is the one that has fared worst. The percentage only tells you how big a bet the company is placing; productivity tells you whether it's winning the bet.
Pitfalls and limitations
1. Acquired R&D. Merck reported R&D expense of $30,531 million in 2023, 50.8% of its $60,115 million in revenue. Its labs didn't suddenly do twice the work: the 10-K explains charges of $10,200 million for the purchase of Prometheus Biosciences and $5,500 million for the collaboration with Daiichi Sankyo. When you buy a project (rather than a whole company), US rules make you expense it all at once. That year Merck earned just $365 million. Before you compare, separate in-house R&D from acquired R&D. This is also the answer to the pre-test: Merck's 24.3% in 2025 versus Lilly's 20.5% says how much each one is betting, not how much it will grow.
2. Capitalize or not. A company that reports under IFRS and capitalizes development shows more profit today and more amortization tomorrow. If you compare it with a US company, look in the notes for how much development it capitalized and how much it amortized, and add it all back to the expense.
3. The cut that flatters earnings. Cutting R&D lifts profit instantly, and the damage, if any, shows up years later. It's a red flag when operating margin rises just as R&D falls and the annual report offers no explanation.
4. The denominator. The ratio goes up when sales fall. Part of Intel's 31.2% in 2024 comes from its revenue dropping from $79,024 million in 2021 to $53,101 million in 2024.
5. Productivity is noisy. New sales also depend on pricing, manufacturing capacity and competition. Look at it over three to five years of data, never over a single year.
Case in point: Intel vs. Nvidia
Between 2021 and 2024 Intel spent $65,310 million on R&D (15,190, 17,528, 16,046 and 16,546 million). Over that stretch its revenue fell 32.8%, from 79,024 to 53,101 million, and operating income swung from a $19,456 million profit in 2021 to an $11,678 million loss in 2024. If you compare three equivalent years, Intel spent $50,120 million from 2022 to 2024, while Nvidia spent $28,928 million in its fiscal 2023, 2024 and 2025 (7,339, 8,675 and 12,914 million; its fiscal year ends in January), a little more than half. Over that time Nvidia's revenue went from 26,974 to 130,497 million.
Does that mean Intel should have spent less on research? Not necessarily. It means the amount says nothing without the result. An investor who looked only at R&D to revenue in 2021 would have seen Intel as the company most committed to the future. The better question was a different one: what sales and what margins is that money buying? That question leads you to ROIC and to margins, which is where good R&D eventually shows up.
Practice on Kaplio
Frequently asked questions
Is R&D an expense or an asset?
It depends on the accounting rules. Under US GAAP almost all R&D is expensed in the year it's paid. Under IFRS, research is expensed, but development can go on the balance sheet as an intangible asset if the project is feasible and will generate future benefits. Economically, much of R&D behaves like an investment.
Is R&D capitalized under IFRS?
Partly. IAS 38 splits R&D into two phases. Research costs always hit the income statement. Development costs are capitalized as an intangible asset once the company shows technical feasibility, the intent and resources to finish the project, the ability to use or sell it, and probable future benefits. They're then amortized over the asset's useful life.
What is considered R&D expense?
Salaries of scientists and engineers, clinical trials for new drugs, materials and prototypes, technology licenses, computing power to design chips or train models, and depreciation of the labs. In the US, buying research projects from other companies is also booked as R&D, as Merck did with Prometheus in 2023, when R&D reached 50.8% of its revenue.
Is R&D an operating expense?
Yes. On the income statement R&D sits on its own line within operating expenses, next to SG&A, so it reduces operating income in the year it's spent. That's why heavy research makes a company look less profitable under US rules. Investors measure it against revenue: Lilly spent 20.5% of its sales on R&D in 2025.