Apple Just Raised R&D to Nearly 10% of Revenue. Traditional Businesses Should Raise Theirs Too.

Apple spent roughly $125 million a day on research and development in fiscal 2026. That comes to $45.77 billion, or 9.6% of net sales, up from 2.2% fifteen years ago. In the U.S., non-manufacturers outside tech, finance and professional services average 1.1%, and that counts only the ones reporting any R&D at all.

What did Apple just do?

Apple spent $45.77 billion on R&D in fiscal 2026, MacRumors reported on November 3. That is $11.22 billion more than the year before, a 32% jump, while sales grew 15%. Apple's financial statements put R&D at 9.6% of its $478.83 billion in net sales, up from 8.3% in fiscal 2025. Its 10-K rounds that to 10% and attributes the increase mainly to infrastructure costs, including AI, and to headcount.

Two details matter more than the headline.

• Apple wasn't always a heavy spender. In fiscal 2011 it spent $2.43 billion, about 2.2% of net sales. Since then, its sales have grown about 4.4 times and its R&D about 19 times.
• The money didn't come from distress. Apple's operating income rose 19% to $158 billion in the same year.

Apple's margins aren't yours, and 10% isn't a target for a distributor or a food manufacturer. We take a narrower lesson from it: one of the most profitable incumbents in the world looked at a dominant product line and chose to raise its R&D share anyway.

How much do companies spend on R&D as a percentage of revenue?

U.S. companies that reported performing or funding R&D spent an average of 5.1% of domestic sales on it in 2023, according to the National Science Foundation's business R&D survey. That average hides a wide spread, and traditional sectors sit near the bottom.

Company or sector	R&D as % of sales
Software publishers, 2023		14.6%
Apple, FY2026		9.6%
Deere equipment operations, FY2025		5.9%
All U.S. companies reporting R&D, 2023		5.1%
Machinery, 2023		4.2%
Motor vehicles and parts, 2023		3.2%
Apple, FY2011		2.2%
Finance and insurance, 2023		1.9%
Other manufacturing, 2023 Includes food, paper, plastics, metals		1.4%
Other non-manufacturing, 2023 Includes construction, retail, wholesale, transportation, health care		1.1%

Sources: NSF/NCSES, Table 3 (domestic R&D over domestic sales); Apple financial statements and 10-K filings; Deere investor presentation. Definitions differ between the company and survey figures, so read the comparison as directional.

Those averages flatter traditional industries. The NSF table counts only companies that performed or funded R&D, so a business with no R&D isn't in it, and the true figure across each whole sector is lower. The gap is also partly a labeling gap, since some process engineering and product testing never gets recorded as R&D.

Does spending more on R&D actually pay off?

On average, yes. Automatically, no. Four findings belong in any honest answer.

• R&D earns positive returns on average. A review of the economic literature by Hall, Mairesse and Mohnen concluded that private returns to R&D are "strongly positive and somewhat higher than those for ordinary capital."
• A bigger budget doesn't guarantee a better result. Booz Allen Hamilton's study of the 1,000 public companies that spent the most on R&D found spending levels "had no apparent impact on sales growth, gross profit, operating profit, enterprise profit, market capitalization, or total shareholder return." Booz pointed instead to how a company "generates, selects, develops, and commercializes ideas."
• Spending too little does hurt. In the same study, companies in the bottom 10% of indexed R&D-to-sales ratios underperformed the rest on gross margins, gross profit and shareholder returns.
• Low-tech manufacturers appear to get less per dollar. A 2011 study of U.S. and European R&D investors found high-tech manufacturers "far ahead" of low-tech ones on the productivity gained from research. It also found service firms showed a higher R&D impact than manufacturers.

These studies have limits. Booz's first study dates from 2005, and these samples consist of companies already investing in R&D. None of them measures a regional distributor going from zero to 1% of revenue, so applying them to that company is an inference, and we'd rather say so. Our reading is that the evidence argues against raising R&D on autopilot and against spending almost nothing.

Which traditional companies already run on R&D?

These companies were picked because the bets worked, so treat them as illustrations rather than proof.

Deere has made farm equipment since 1837. It spent $2.31 billion on R&D in fiscal 2025, or 5.9% of equipment net sales. That year equipment sales fell 13% and net income fell from $7.1 billion to $5.03 billion, yet R&D edged up from $2.29 billion. Its See & Spray sprayers show the direction: they use computer vision to target weeds, and Deere says they can cut post-emergence herbicide use by more than 50% when used with a residual herbicide.

UPS put its R&D into a process. Its ORION routing system cost an estimated $250 million to deploy. UPS expected it to save 100 million miles and $300–400 million a year. A UPS executive put the logic in one line: eliminating one mile per driver per day is worth $50 million annually. Operations-heavy businesses repeat the same work millions of times, so a small improvement adds up fast.

Domino's sells pizza through software. The company generated more than 85% of its U.S. retail sales in 2025 through digital channels. Whatever accounting line funded that work, it changed how a restaurant chain reaches its customers.

Mid-sized companies do this too. Hermann Simon's research on "hidden champions," little-known mid-sized leaders of niche global markets, puts their R&D at 6.0% of revenue, against 3.0% for typical firms. These are Simon's own figures from a sample chosen for success, so they show an association rather than a cause.

Why can't a traditional company just buy the technology later?

You should buy most of it. Few companies need to build their own payroll system or train their own language model. R&D is the work that makes a purchased technology change your results, and that work is hard to outsource to a vendor.

• Doing R&D is how companies learn to use outside technology. Cohen and Levinthal's research on "absorptive capacity" argues that firms doing their own R&D are better able to use knowledge from outside.
• AI requires more than the AI. Brynjolfsson, Rock and Syverson write that technologies like AI "enable and require significant complementary investments, including co-invention of new processes, products, business models and human capital."
• Adoption alone rarely reaches the P&L. In McKinsey's November 2025 survey, only 39% of respondents reported enterprise-level EBIT impact from AI. High performers were nearly three times as likely to have fundamentally redesigned their workflows.
• The gap is already visible. As of May 2026, the Census Bureau found 39.7% of businesses in the information sector and 33.9% in finance and insurance using AI, against about 14% in retail and 19.8% nationally.
• Laggards don't catch up by default. OECD economists found a productivity gap between frontier firms and everyone else that has grown over time, even after accounting for investment levels. They suggest it may reflect a growing difference in "innovation performance in a broad sense." Their frontier examples include traditional names such as BMW, L'Oréal and Nestlé.

How much should your company budget for R&D?

No study gives a correct percentage. Booz found that even the largest spenders don't cluster around an industry norm. The guidance below is our judgment for businesses outside technology, and it's a starting point for your CFO to challenge.

Here is what those percentages mean in dollars:

If your margins are thin, size it against profit instead. At a 4% operating margin, 1% of revenue is a quarter of your operating profit, which is a serious commitment. In that case, we'd start at 10–15% of operating profit and fund it by moving money out of stalled IT projects and low-return marketing. Don't take it from maintenance, safety or quality.

If you are small, don't benchmark against a giant's percentage. Booz found larger companies could spend a smaller share of revenue with no discernible performance hit, because scale does some of the work for them.

What counts as R&D in a low-tech business?

The NSF, using the OECD's definition, describes experimental development as systematic work "directed to producing new products or processes or to improving existing products or processes." You don't need a laboratory: 79% of the $722 billion U.S. companies spent on R&D in 2023 went to development, and only 6% to basic research.

A practical test: if you already know it will work and you're just rolling it out, it isn't R&D. If the outcome is uncertain and you're running a structured test to find out, it probably is. Buying software licenses, routine IT maintenance and advertising don't count.

Here is what a first round of questions could look like. These are illustrative examples, not client results.

How do you start R&D in a company with no R&D function?

1. Create a budget line and name an owner. The owner should report to the CEO or COO. Keep the budget separate from IT's run costs, or maintenance will absorb it.
2. Pick three questions worth real money. Start where a small improvement repeats thousands of times, such as a quote, a route, an inspection or a claim.
3. Write the stop rule before you spend. Decide what result would end the project while nobody is attached to it yet.
4. Time-box the work. Allow four to six weeks for a working prototype on your real data, then 90 days for a pilot with the people who do the job today.
5. Report it to the board like any other investment. Each quarter, show R&D as a percentage of revenue, tests run, tests stopped and the dollar impact of what shipped. If nothing ever gets stopped, you're only testing safe bets.
6. Bring in your tax advisor before the first dollar. Credits depend on records you can't easily recreate later.

Does the U.S. tax code help?

Yes, and it recently got better. For tax years beginning after December 31, 2024, Section 174A lets companies deduct domestic research costs immediately instead of amortizing them, and software development still counts as a research expense. Foreign research must still be capitalized and amortized over 15 years.

The Section 41 research credit may also apply. Its definition of a business component includes "any product, process, computer software, technique, formula, or invention," so improving an internal process can qualify. The exclusions matter: market research, efficiency surveys and routine data collection are out, and software built primarily for internal use is excluded unless regulations provide otherwise. This is general information rather than tax advice, and state rules differ.

When should a company not raise R&D?

Our answer applies to almost every business, with four exceptions:

• You're fighting for liquidity. Fix that first, because R&D pays back over years.
• You plan to sell or wind down within about two years. The next owner would collect most of the benefit, unless the work would raise your sale price.
• You can't name the questions. Without a specific question and a measure, the budget will turn into a collection of demos.
• Nobody senior can own it. Without an owner with authority over the workflow, successful pilots won't reach production.

Frequently asked questions

What percentage of revenue should a company spend on R&D?

No study has proven a universal figure. U.S. companies that report R&D averaged 5.1% of domestic sales in 2023, according to the National Science Foundation. For traditional businesses spending under 1%, R&D Teams recommends committing 1% of revenue for two years and moving to 2–3% once experiments start reaching the P&L. That recommendation is our judgment, not a research finding.

How much do traditional industries spend on R&D?

In 2023 NSF data, U.S. companies reporting R&D averaged 1.1% of sales in the non-manufacturing group that includes construction, retail, wholesale, transportation and health care. The manufacturing group that includes food, paper, plastics and metals averaged 1.4%. Finance and insurance averaged 1.9%, motor vehicles and parts 3.2%, and machinery 4.2%. Companies with no R&D are not counted, so true sector averages are lower.

Does higher R&D spending lead to better financial results?

On average R&D earns positive returns, but a bigger budget does not guarantee a better result. A review by Hall, Mairesse and Mohnen found private returns strongly positive and somewhat higher than those for ordinary capital. Booz Allen Hamilton found no apparent link between spending level and business results among the 1,000 largest R&D spenders, although the bottom 10% of spenders did underperform on gross margins, gross profit and shareholder returns.

Is buying AI software the same as doing R&D?

No. Buying licenses is procurement, because you already know what you are getting. R&D starts when there is real uncertainty and you run a structured test to resolve it, such as whether an AI agent can quote routine orders as accurately as your sales team using your price book and order history.

Can a company without engineers do R&D?

Yes. Development accounts for most business R&D: 79% of the $722 billion U.S. companies spent in 2023, according to the NSF. A company can hire outside engineers to build and test, but it needs a senior internal owner who controls the budget, picks the questions and enforces stop rules.

Is R&D tax deductible in the United States?

Generally yes for domestic work. Under Section 174A, domestic research costs can be deducted immediately for tax years beginning after December 31, 2024, while foreign research costs must still be capitalized and amortized over 15 years. The Section 41 research credit may also apply, and its definition of a business component includes processes and software, subject to exclusions. Confirm your position with a tax advisor.