In 2024, China became the world’s leading R&D executor according to an adjusted international comparison, surpassing the United States for the first time. The United States retains comparative advantages in several categories of highly-cited patents and in venture capital-financed innovation, but these advantages are contested and have eroded in certain critical domains. Scientific competition is won less through announcements than through continuity.

The Essentials

  • China surpassed the United States in R&D volume in 2024, according to the NSF’s 2026 report (National Center for Science and Engineering Statistics), but the American advantage in highly-cited patents and private innovation remains intact.
  • This distinction between volume and quality of innovation corresponds precisely to what economists of Schumpeterian growth call the “technological frontier”: spending more does not guarantee pushing it forward.
  • Current American budgetary trade-offs threaten the infrastructure of fundamental research and training pipelines whose effects on competitiveness only appear with a decade-long lag.
  • The institutional stakes are central: no country has maintained a durable technological advantage without a mechanism for constant renewal of laboratories, talent, and transfer to the private economy.
  • The open question is that of balance between defending the technological frontier and the speed of China’s advancement in high-citation-intensity segments.

R&D Volume Does Not Define the Technological Frontier

When the NSF publishes its figures and China appears at the top, the immediate reaction is to count dollars and sound the alarm. This reaction is understandable. It is also insufficient.

The 2026 report from the National Center for Science and Engineering Statistics specifies that China’s surpassing is established on the basis of an adjusted international comparison, which corrects for purchasing power parity gaps. In nominal terms, the United States remains far ahead. But even the adjusted indicator raises the question of what is actually being measured: total R&D expenditures aggregate fundamental research in public universities, engineering projects in Chinese state enterprises, software development in large American private platforms, and military programs on both sides. These activities do not have the same effects on the scientific frontier.

Economist Philippe Aghion, 2025 Nobel laureate, built part of his career around a distinction that this debate demands. The Schumpeterian tradition insists on innovation, entry, competition, and creative destruction; it does not imply that the volume of R&D investment is unimportant. A country that accumulates expenditures without a mechanism of creative destruction in science can produce volumes without necessarily pushing the frontier forward. It can converge toward it without necessarily moving it.

The data from the 2026 NSF report illustrate exactly this tension. Patents assigned to American inventors have, in several critical technologies, a relatively higher probability of belonging to the first percentile of cited patents; this does not prove they are more numerous in absolute value. The intensity of innovation in the American private sector, measured by corporate R&D spending as a proportion of their value added, constitutes an important indicator. Highly-cited patents and private innovation are useful indicators of the impact and economic translation of R&D, to be interpreted jointly with expenditures, publications, talent, and output.

China Is Moving Upmarket, and This Movement Is Real

This does not mean that Chinese progress is a statistical illusion. The opposite error consists of minimizing what is happening.

China has experienced very rapid growth in its R&D expenditures since 2000; the exact factor must specify the unit, the final year, the treatment of inflation, and the conversion rate. It now trains more engineers and PhDs in natural sciences than any other country. Chinese scientific output and its share of highly-cited articles have advanced rapidly; available data do not justify a universal comparison based on journal impact factors. In batteries, electric vehicles, and photovoltaics, China dominates or weighs heavily in production, and it is very present in artificial intelligence patents; the assertion that it defines global standards requires additional sectoral evidence.

Carl Benedikt Frey, who works on technological transitions and their geography, emphasizes that waves of innovation are not linear. A country can accumulate significant sectoral advances without holding the frontier overall, then, at a moment difficult to anticipate, shift toward a dominant position on a segment that becomes structuring for the global economy. The question posed by China’s upmarket shift is therefore not “is it already at the global frontier” but “in which segments will it be at the frontier in ten years, and will these segments be the ones that matter.”

That is precisely where current American budgetary trade-offs become concerning. As an article already published in this journal notes, building labs does not retain researchers: the continuity of careers, the security of multi-year funding, and international attractiveness count among the conditions for maintaining research capacity. Budget cuts can affect these dimensions, with effects that may only appear in the medium term.

Scientific Infrastructure Degrades in Silence

There is a profound asymmetry between the pace of building scientific capacity and that of its degradation. A country can lose in a few years assets that took a decade to build.

The United States built its scientific advantage on a specific model: research universities financed in part by federal funds, capable of training PhDs who then join industry or remain in the academic system, with mechanisms for technology transfer, patents, licenses, spin-offs, that convert public research into private innovation. This model works. It is associated with the development of technology hubs such as Silicon Valley, Boston, and Seattle. The FY2026 budget request, not the strategic plan itself, documents proposed reductions in certain lines; the strategic plan does not formulate this diagnosis.

Cuts or freezes can reduce grants and delay investments; some effects may appear long-term, but their magnitude varies and is not always quantifiable. PhD students who do not find funding leave the system. Teams that lose their funding dissolve. Expensive experimental platforms degrade due to lack of operating budget. None of these phenomena are immediately visible in patent or publication statistics, whose production timelines are several years.

When the effect appears in the data, it is already difficult to correct.

China, conversely, maintains continuity of investment that results from a central political decision: science and technology are explicitly in the strategy of power. The continuity of Chinese investment ensures important volumes and multi-year visibility, which can contribute to building solid scientific institutions without guaranteeing them.

Economist Daron Acemoglu formulated a caveat that tempers Schumpeterian optimism: institutions do not automatically produce the progress they could produce. They can be captured, underfunded, diverted from their mission. The American research system is a remarkable institution, but an institution is not immortal. It requires active maintenance, deliberate renewal, political choices that protect it from short-term logic.

Transfer to the Private Economy Remains the American Structural Advantage

The United States possesses important technology transfer mechanisms and strong commercialization capacity; their exact comparative advantage in transfer depends on the indicators retained.

Major American technology companies—Microsoft, Google, Amazon, Apple—invest themselves massively in R&D. But beyond these names, the network of deeptech startups, university spin-offs, and venture capital funds specialized in science constitute an important ecosystem. The United States maintains very strong corporate R&D and marked concentration in information services, but the 2026 NSF data do not allow one to assert that they are very far ahead worldwide in private R&D in absolute value and at PPP.

This transfer model depends on conditions that are not guaranteed. It assumes universities capable of producing knowledge sufficiently advanced to be valuable. It assumes intellectual property mechanisms that incentivize investment without blocking knowledge diffusion. It assumes trained talent capable of bridging fundamental science and commercial applications. The American model of technology transfer depends heavily on public policy decisions, but also on private capacities for financing, entrepreneurship, and commercialization.

AI amplifies these inequalities between innovation systems: countries capable of coupling quality fundamental research with an industrial fabric capable of absorbing and deploying advances can benefit from multiplier effects. The United States is in this position. They will not remain there without maintaining both legs of the system.

Today’s Decisions Condition Capacities in 2040

The prospective dimension of this competition is the most difficult to integrate into budgetary debates. The effects of fundamental research funding choices are often measured over the long term rather than the short term.

A materials physics laboratory closed today will not immediately produce its effects in patents. Its consequences may appear later in companies that will not be founded, materials that will not be developed, and industrial applications that will remain with competitors. The decision and its effect may be separated by several years. This lag makes budget cuts particularly insidious: they are politically invisible at the moment they do the most damage.

The plan indicates that the NSF finances both fundamental research and solution-oriented research in order to keep the United States at the forefront of discovery; it does not formulate an explicit objective at twenty or thirty years. The effects of a decline in fundamental research on innovation capacity depend on institutional and economic contexts.

For the United States, a continuation of current budgetary trade-offs could affect doctoral programs, attractiveness to foreign talent, and renewal of experimental platforms. This scenario is not inevitable. It is conditional. The signals to monitor are precise: the number of researcher visas granted, the renewal rate of multi-year NSF grants, the proportion of PhDs trained in the United States who remain there after their training. These indicators are all measurable, and the NSF tracks them.

The alternative is a deliberate reinvestment in fundamental research infrastructure, coupled with a talent policy that maintains the United States as the destination of choice for the world’s best researchers. Attractiveness policy is not reduced to salaries: it also includes research freedom, funding stability, and the quality of international collaborations.

This tension, as suggested by Aghion’s work on the relationship between competition policy and growth, may depend on institutions capable of maintaining selection based on quality. A research funding system that rewards risk-taking, tolerates failure, and renews the leadership teams of scientific programs can foster frontier innovation. This lesson applies to public funding agencies as much as to companies.

Competition Also Plays Out in Standards and Alliances

An angle that crude R&D expenditure comparisons tend to erase: the technological frontier is not defined only by discoveries, it is also defined by standards, protocols, and alliances that determine which discoveries become global norms.

The role of the United States in defining international technology standards varies across sectors, standardization bodies, and retained indicators, such as committee presidencies, adopted technical contributions, patents essential to standards, or market shares. This role is distinct from R&D expenditures, but may depend on them partially. Highly-cited patents measure an influence or relative impact on subsequent inventions; they do not allow, by themselves, to measure the structuring of a scientific field.

China has understood the stakes and is deploying an explicit strategy for increasing power in international standardization bodies, particularly in the sectors of telecommunications, energy, and artificial intelligence. This strategy is not new, but it is intensifying as its technical capacities advance. This strategy is part of the development of its technical capacities.

Alliances with Europe, Japan, and South Korea strengthen the American position in international standards, alongside American internal capacities. Restrictions on scientific mobility or international cooperation can weaken certain American assets in standards, but the effect depends on specific policies and sectors concerned. Technological competition includes budgetary, industrial, scientific, diplomatic, and geopolitical dimensions, and its instruments are often institutional.

Over the horizon of the next decade, the United States will need to determine whether it maintains the institutional conditions that have made its scientific system one of the most productive in recent history, or whether it allows the foundations of important but variable advantages to degrade according to indicators and technologies, in a context of intensified competition in several critical domains.

Sources

  1. National Center for Science and Engineering Statistics, NSF, National Science Board: Science and Engineering Policy Indicators 2026
  2. NSF, Strategic Plan FY 2026-2030 (no link: document available on nsf.gov)
  3. OECD, Main Science and Technology Indicators, longitudinal data on R&D expenditures (no link: database available on stats.oecd.org)