Tracking business R&D in real time: the largest investors raised R&D spending by 6% in 2025

Timely information on business research and development (R&D) investment is key for understanding how innovation systems respond to economic shocks, structural change and policy intervention. Official business R&D statistics provide robust and internationally comparable measures, but they are published with a significant delay, one to two years after the end of the reference period. How can this gap be addressed through meaningful, up-to-date signals on the direction of global R&D? 

The OECD Short‑term Financial Tracker of Business R&D (SwiFTBeRD) dashboard was first launched in 2021 to help address this gap, building on work initiated in 2020 when OECD analysis of quarterly company reports provided key insights into the impact of the COVID-19 crisis as it unfolded. SwiFTBeRD tracks a panel of the world’s largest R&D firms to provide a timely view of R&D trends. Its use of publicly-disclosed data in annually and quarterly company reports is the main differentiating factor with long-standing official R&D statistics also published by OECD. This and other key features are summed up in the comparison table below. Public disclosure of R&D expenditures is only required in some jurisdictions and remains optional for most firms, especially SMEs, whereas statistical reporting can be mandated by national legislation to all firms. Companies report more information on R&D under statistical secrecy than in public reports, which follow accounting rules serving other purposes. Interpreting R&D spending disclosures requires adjustments for example to add capital expenditures for R&D to R&D expenses, or exclude R&D expenses that reflect for example the acquisition of R&D assets developed in past years by other firms.  

The apparently stark trade-off between the comprehensiveness (and its avoidance of bias) in official statistics and the timeliness of public reports is however mitigated by the fact that R&D spending is a highly concentrated activity, with few companies accounting for a large share of the total within and across countries. SwiFTBeRD is based on the premise that adequately curated data from a relatively small group of leading R&D companies can provide not just timely but also relevant and sufficiently reliable initial insights on the overall direction of business R&D investment. 

SwiFTBeRD comprises a panel of 60 companies representing close to 46% of R&D reported by the top 2000 global R&D investors in the EU Industrial R&D Investment Scoreboard.  The panel composition is reviewed periodically to reflect changes at the top of this group. Firm-level adjustments using additional data within reports help not only bring reported values closer to the concept of in-year R&D expenditure used in official statistics, but also improve consistency across companies and over time, in view of systematic M&A activity, spin-offs, restructuring and changes in reporting practices. SwiFTBeRD connects time series from different company reports with overlapping years to provide meaningful trends at firm and aggregate levels. 

Companies in Software & computer services and IT hardware and electronics account together for over 60% of total R&D spending covered in SwiFTBeRD in 2025, followed by life sciences and auto and aerospace manufacturing. This marks a major structural shift in R&D investment. Since 2018, the weight of the combined “digital” industry groups has substantially increased, driven by the share of software and computer services going from 20 to 35%, while the share of automotive and aerospace has fallen from 27 to 17%. Other industries are not represented among the very top R&D investors. 

Companies active in the semiconductor supply chain, including lithography equipment (ASML) to chip design and manufacturing (in addition to NVIDIA, AMD, BROADCOM, INTEL, QUALCOMM and TSMC, as well as Samsung) play a prominent role in the panel alongside the “GAFAM” group comprising Meta, Amazon, Microsoft, Alphabet, alongside China’s Tencent. The June 2026 release of SwiFTBeRD includes for the first time estimates of R&D expenditure for Amazon, which while it does not report R&D expenses but a broader category of Technology spending, publishes information on R&D tax credits which allow for plausible estimates explicitly based on R&D. 

NVIDIA is the company in the SwiFTBeRD panel with the largest nominal R&D growth in 2025 at 43%, followed by TESLA at 41%, and META at over 30%. Semiconductor firms Broadcom and AMD join the top 5.

Previous OECD qualitative analysis of company reports showed that the initial wave of AI R&D investments was not visible in SwiFTBeRD estimates because AI R&D was highly concentrated in AI start-ups working in partnership with major established companies. The latter’s investment positions in this area were not necessarily disclosed as R&D. The surge of AI R&D investment has now fully reached the large R&D investors providing both IT equipment and services. Among the transport equipment manufacturers in automotive, recent differences in R&D growth may also be related to patterns in AI adoption as well as investments in electric vehicle technologies. 

Firm-level data is aggregated by SwiFTBeRD into indicators at the level of industry groupings and headline figure for the entire panel.  The latest data show that business R&D across larger investors continued to expand in 2025, with 6% inflation‑adjusted growth, up from 3% in 2024. 

Growth in 2025 is driven by digital‑related sectors, with R&D investment in software and computer services and ICT hardware expanding, respectively, at close to 15% and 7% in inflation-adjusted terms. Growth in R&D expenditure in the life sciences sector slowed down to zero, while R&D investment across the panel’s large companies in the automotive and aerospace sector continued its declining trend since 2024.

The SwiFTBeRD responds to the need for higher frequency R&D data in conditions of high economic uncertainty and volatility. For instance, this has helped assess how R&D in different companies and industry groups responded to the COVID-19 crisis, as well as how they emerged from this crisis. The data show the overall resilience of R&D in the IT industries over 2020, compared with the critical decline among automotive and aerospace companies in Q1 2020 from an initial weak position, alongside the initial decline within the life sciences industry which eventually recovered in Q3 2020 to exceed the performance of the IT industry throughout 2021 while new vaccines and treatments were developed. 

R&D expenditure growth has been accelerating since Q1 2025, principally driven by the digital industries, especially software and computer services, which saw a year-on-year growth rate of 20% in Q4 2025. The growth of IT hardware and electronics is in line with the entire panel while R&D growth came to a halt for the life sciences sector in 2025 after a brief respite in 2024 while the automotive and aerospace sector has been accumulating negative growth since Q4 2024, at levels not seen since the beginning of the decade.

The availability of SwiFTBeRD data since 2018 provides a good opportunity for assessing its overall performance nowcasting BERD. Despite the differences in scope and methodology, both series have common underpinnings in the records that companies keep about their R&D activities. The figure below provides a comparison of the aggregate SwiFTBeRD indicator with the OECD’s official estimate of BERD for the OECD area, notwithstanding the presence of non-OECD countries in the panel. The first point that stands out is the close similarity in the pattern of changes in the annual growth rates over the period. Furthermore, the effective comparison of growth rate values does not imply a clear bias although there can be some years in which growth rates may differ by up to 1% or 2%. Further investigation has revealed that the main gap stems from 2023 and the use of a common GDP deflator for the entire OECD area. High inflation in Türkiye had a disproportionate effect on the common OECD GDP deflator which depressed the real growth estimate for total SwiFTBeRD. In the case of OECD BERD this effect is mitigated by the smaller weight of this country in total BERD. The one-off divergence is caused by the choice of deflator, not by the nominal R&D estimates, which provide rather consistent signals. 

Deflating SwiFTBeRD values is not straightforward, since companies' R&D is not limited to the country where they are headquartered. The lack of a reliable mechanism to allocate business R&D to countries is the reason why OECD does not provide country-based nowcasts of BERD through SwiFTBeRD. However, a dedicated module allows users to construct purely illustrative synthetic R&D nowcasts by modulating industry trends with user‑defined industry weights, thus simulating what might be the benchmark R&D growth for an economy where different industry groups have a given relative importance.  This is a means to assess how industrial structure may influence observed R&D dynamics. 

SwiFTBeRD also sheds light on how business R&D moves with the economic cycle. Official data confirm R&D is procyclical, but its response to the cycle varies widely, and neither GDP nor gross fixed capital formation (GFCF) predicts BERD as well as SwiFTBeRD does - as in 2020, the first OECD-wide recession in decades in which BERD did not fall. This is highly relevant information for economic and R&D policy modellers and decision makers. 

SwiFTBeRD is designed to support the timely interpretation of short‑term developments in business R&D investment, complementing less timely but more comprehensive BERD statistics. SwiFTBeRD indicators should be interpreted as early signals of direction rather than precise measures of R&D levels or exact trends. For several reasons, including several blindspots discussed in this note, they are not intended to provide exhaustive or representative measures of business R&D across all firms let alone used to describe or compare internationally a country’s aggregate R&D performance. 

The SwiFTBeRD dashboard is designed as an interactive tool that allows users to explore business R&D investment dynamics from multiple perspectives. Its indicators and analytical components may be extended by the OECD to new groups of firms, additional industries or thematic applications, depending on future analytical needs. By combining a stable core set of indicators with flexible exploration tools, the SwiFTBeRD dashboard gives policymakers an early read on the developments that matter most today across a wide range of industries well before the official statistics confirm them.