IT Capex and R&D: How Tech Companies Invest for Growth
How Do IT Companies Allocate Capital Between Capex and R&D?
Capital allocation decisions — how technology companies split their resources between capital expenditure and research and development — are among the most consequential choices management teams make, and among the most telling signals about a company's competitive strategy and long-term value creation potential. The IT sector is the most R&D-intensive in the economy: technology companies collectively spend more on research and development as a percentage of revenue than any other sector. Understanding what these investments are, what they produce, and how to evaluate their productivity is essential for investors seeking to assess whether technology companies are building or consuming competitive advantage.
Quick definition: In the IT sector, capital expenditure (capex) refers to investment in physical assets like data centers and manufacturing equipment, while research and development (R&D) spending covers engineers, scientists, and resources dedicated to creating new products and improving existing ones — both representing investment in future competitive position.
Key takeaways
- IT companies collectively spend 10–20% of revenue on R&D, versus 2–4% for the average S&P 500 company
- Semiconductor companies have the highest capex intensity (TSMC spending $30–36 billion annually)
- Software companies have minimal capex but high R&D intensity
- AI infrastructure investment is driving hyperscaler capex to record levels
- R&D productivity (revenue generated per dollar of R&D) varies enormously across IT companies
R&D intensity as a competitive moat signal
Research and development intensity — R&D spending as a percentage of revenue — is both a measure of competitive investment and an indicator of intellectual property density. Companies that consistently spend heavily on R&D are investing in staying ahead of the technology curve; those that reduce R&D to boost near-term margins may be harvesting competitive position.
The IT sector's most innovative and defensible companies tend to have sustained R&D intensity:
- Large pharmaceutical companies average roughly 15–20% of revenue in R&D
- Semiconductor companies spend approximately 15–25% of revenue
- Software companies spend approximately 15–30% of revenue, depending on growth stage
- Hardware companies spend approximately 5–12% of revenue
The challenge for investors is that R&D is immediately expensed under GAAP accounting rules — it does not create a balance sheet asset that can be valued. This means that high-R&D companies can appear to have lower earnings than low-R&D companies with similar underlying profitability, creating potential for undervaluation by investors who focus exclusively on reported earnings.
Semiconductor capex: the most capital-intensive in technology
Semiconductor manufacturing requires the highest capital investment of any business in the technology ecosystem. Building a leading-edge semiconductor fab capable of producing the most advanced chips (2nm, 3nm process nodes) costs $15–25 billion and requires several years from groundbreaking to commercial production. Equipment must be purchased from monopoly or near-monopoly suppliers (ASML for EUV lithography, Applied Materials and Lam Research for deposition and etch).
TSMC's capital expenditure trajectory illustrates the escalating capital requirements:
- 2018: approximately $10 billion in capex
- 2022: approximately $36 billion in capex
- 2024: approximately $30–32 billion in capex
This level of capital investment creates an enormous barrier to entry that protects TSMC's competitive position but also means that any period of revenue weakness creates margin pressure because the fixed costs associated with unused capacity are very high. The capital expenditure decisions made today determine manufacturing capacity available in 3–5 years, requiring management teams to forecast demand accurately years in advance.
The CHIPS Act of 2022 has altered capex economics for US-based semiconductor manufacturing by providing significant grants and tax credits that reduce the effective cost of building US fabs. Details of current program terms are maintained at the Department of Commerce at commerce.gov.
How it flows
Hyperscaler capex: the AI infrastructure buildout
Cloud hyperscalers (Amazon, Microsoft, Google, Meta) have entered a new, higher capex regime driven by artificial intelligence infrastructure requirements. Training large language models and serving AI inferences at scale requires data centers filled with GPU servers, high-bandwidth networking, and enormous power infrastructure.
Combined annual capital expenditure by the four largest hyperscalers was approximately $200+ billion in 2024, a step-change increase from the $100–120 billion range of 2022. This spending creates a direct revenue opportunity for:
- Nvidia (GPU servers)
- Broadcom and Marvell (custom AI ASICs and networking chips)
- Data center REITs (Digital Realty, Equinix)
- Utilities and power companies serving data center campuses
- Electrical equipment companies (transformers, power distribution)
The risk of this capex cycle for hyperscaler investors is near-term free cash flow compression. Microsoft, Google, and Amazon are all spending significantly more in capex than they did two years ago, reducing free cash flow per share in the near term even as the investments are expected to generate superior revenue and margins over a multi-year horizon.
R&D productivity: measuring return on innovation
R&D productivity — the revenue or profit generated per dollar of R&D spending — is notoriously difficult to measure but critically important for assessing long-run value creation. Several proxies help investors evaluate whether R&D spending is productive:
Revenue growth relative to R&D spending: Is the company's revenue growing faster than its R&D expense? A company increasing R&D 20% annually while growing revenue only 10% annually is seeing declining R&D productivity.
New product revenue as percentage of total: Companies that disclose new product contribution to revenue provide insight into whether R&D is generating commercially successful innovations or primarily maintaining existing products.
Patent filings and citations: While a crude proxy, patent filing volume and the citation count of those patents (how often other companies reference them as prior art) provides some signal about the technical productivity of R&D programs.
Management commentary: The most sophisticated investors read R&D commentary in quarterly earnings calls and annual reports carefully. Management teams that articulate specific R&D investment rationale, expected products, and timelines are easier to evaluate than those who describe R&D vaguely.
Real-world examples
Alphabet's R&D spending illustrates both the scale and the optionality of technology R&D. Google spent approximately $45 billion on R&D in 2023, nearly 15% of its revenues. Much of this goes to maintaining and improving Google Search, YouTube, and advertising technology — core businesses with clear revenue connections. A portion goes to Google DeepMind and Google Brain (AI research), Google Cloud (infrastructure and services), and moonshot projects through Alphabet's Other Bets (Waymo autonomous driving, Verily life sciences). The Other Bets historically generated minimal revenue relative to their investment, but Waymo's autonomous ride-hailing service is approaching commercial scale.
Microsoft's capex trajectory demonstrates how quickly the AI investment cycle changed hyperscaler spending priorities. Microsoft's annual capex was approximately $20 billion in fiscal 2022. By fiscal 2024, it had risen to approximately $44 billion, driven primarily by Azure AI infrastructure expansion. This doubling of capex created concern among some investors about free cash flow impact but was broadly viewed as necessary investment to compete with Amazon and Google in the AI cloud market.
Common mistakes
Treating R&D reductions as margin improvement. A company that cuts R&D from 20% to 15% of revenue shows near-term margin improvement, but if that R&D was funding future product generations, the cut may be harvesting long-run competitive position for near-term profits. R&D-intensive industries require sustained investment to maintain product competitiveness.
Ignoring the capex cycle in semiconductor investments. Semiconductor companies that commit to high capex during boom cycles (forecasting continued demand growth) often find themselves with excess capacity when the cycle turns. The combination of high fixed costs from new facilities and declining revenues during inventory corrections creates severe margin compression. Monitoring capex guidance versus revenue expectations is critical for semiconductor investors.
FAQ
Should I prefer low-capex or high-capex technology companies?
Neither is universally better. Low-capex software companies can generate exceptional returns on capital by investing primarily in people (R&D). High-capex semiconductor and hardware companies can generate strong returns if their capital generates competitive advantages that translate into pricing power. The test is return on invested capital — whether the company earns more than its cost of capital on incremental investments.
How does the IRS treat technology company R&D expenses?
Under the Tax Cuts and Jobs Act of 2017, changes to Section 174 required companies to amortize domestic R&D expenses over 5 years and foreign R&D over 15 years starting in 2022, rather than deducting them immediately. This change significantly impacted technology company reported earnings and cash taxes for companies with large R&D programs. Congress has periodically discussed reversing this change; confirm current treatment at irs.gov before drawing conclusions about technology company after-tax cash flows.
Related concepts
- IT Sector Overview
- Semiconductors Explained
- Cloud Computing in IT
- IT Sector Valuation Multiples
- AI Impact on IT Sector
Summary
Capital allocation between capex and R&D is a defining characteristic of IT sector companies, with different subsectors exhibiting dramatically different investment profiles. Semiconductor manufacturers carry the highest capex intensity in technology, spending tens of billions annually on leading-edge fabrication capacity. Hyperscalers have elevated their capex to record levels to build AI infrastructure. Software companies invest primarily in R&D — people and programs — generating intellectual property with minimal physical asset intensity. Evaluating whether these investments are productive — through revenue growth, new product success, and return on invested capital — is a core competency for IT sector investors who want to distinguish companies building durable competitive advantage from those spending capital without proportional value creation.