In the latest annual report released by the Semiconductor Industry Association (SIA), there is a chart that easily triggers strong emotions.
From 2015 to 2024, China's R&D investment has been on a steady rise, with the curve catching up to and even slightly surpassing that of the US in 2024. Based on this, the SIA warns US policymakers that corporate and government R&D spending in China has already exceeded that of the US, particularly in later-stage development, urging the US to increase its investment again. The report also highlights that R&D expenditure by US semiconductor companies reached $76.8 billion in 2025, a year-on-year increase of 10.3%.
China's R&D investment catching up with the US is a shift that deserves serious attention. However, it merely indicates that China now has the capacity to deploy the world's largest-scale R&D resources; it does not mean that all technological fields have reached parity, nor can it directly lead to the conclusion that China's semiconductor industry has entered a stage of comprehensive leadership.
Money still matters. It is just that once investment reaches this magnitude, the competition begins to shift from "whether there is money" to "where the money is invested, who spends it, how it is translated into results, and whether the industry can endure long periods without tangible outcomes."
"Surpassing the US": First, Consider the Yardstick Being Used
The SIA cites data from the Organisation for Economic Co-operation and Development (OECD).
In statistics released in March 2026, the OECD stated that, calculated at Purchasing Power Parity (PPP), China's total R&D investment in 2024 reached approximately $860 billion. Measured in constant 2020 prices, it has caught up with and slightly surpassed that of the US. Converted to 2024 prices, the R&D investments of both China and the US have exceeded $1 trillion.
The logic behind PPP is not complex. With the same amount of money, the engineer salaries, laboratory rents, and certain research services that can be afforded in China are generally greater in quantity than what can be purchased in the US. Comparing via PPP more closely reflects the actual amount of R&D resources mobilized by the two countries.
However, the OECD also specifically cautions that existing PPPs are primarily designed for GDP comparison and are not specifically calculated for R&D activities. Using PPPs with different base years, China's R&D investment in 2024 could be equivalent to 90% to 95% of the US, or slightly higher; if converted directly at market exchange rates, China's R&D investment is only about half that of the US.
Therefore, stating that "China's R&D investment has surpassed the US" is not incorrect, but it is a statement that must be used with a specific statistical scope. A more prudent expression would be: measured by PPP, China's R&D investment has reached a scale comparable to that of the US; calculated by market exchange rates, the US still maintains a clear lead.
Moreover, these two calculation methods do not answer the same question.
PPP is more suitable for observing how much domestic R&D resources China has mobilized; market exchange rates more closely reflect the capacity to purchase international equipment, overseas intellectual property, global talent, and cross-border technical services. For an industry as highly internationalized and heavily reliant on expensive equipment and software as semiconductors, neither metric can be ignored. China's money is primarily spent on experimental development.
In 2024, China's total R&D expenditure reached 3.63268 trillion yuan, a year-on-year increase of 8.9%, accounting for 2.69% of GDP. Of this, enterprises invested 2.82116 trillion yuan, representing 77.7% of the national R&D funding. By 2025, China's R&D investment further increased to 3.9262 trillion yuan, with the proportion of GDP rising to 2.80%.
This structure easily creates the impression that China's R&D investment is primarily driven by enterprises and is therefore highly market-oriented. This statement is broadly true but not entirely complete. In 2024, China's R&D expenditure included 250.09 billion yuan for basic research, accounting for 6.88%; 430.55 billion yuan for applied research, accounting for 11.9%; and experimental development reached 2.95204 trillion yuan, accounting for 81.2%. Experimental development refers to utilizing existing scientific knowledge and engineering experience to develop new products and processes, or to improve existing ones.
By 2025, China's basic research funding increased to 277.8 billion yuan, reaching a proportion of 7.08%, an increase of 89.4% compared to 2020. Basic research investment is accelerating, but from an overall structural perspective, China's R&D resources still clearly lean towards engineering development and industrial application.
This is not an inherent flaw. The semiconductor industry is inherently not driven solely by academic papers. Process debugging, yield improvement, equipment enhancement, material validation, product iteration, and customer adaptation all belong to experimental development. For many technologies, the gap between "can be made in the lab" and "can run stably on the production line" may span several years or even longer.
In recent years, China's rapid formation of scaled capabilities in photovoltaic equipment, power batteries, communication systems, consumer electronics, and certain semiconductor segments is closely related to this investment structure that emphasizes engineering development and industrialization. However, it also means that China's advantages are more reflected in turning existing technologies into products, reducing costs, and expanding capacity. Long-term breakthroughs involving new materials, new device structures, underlying algorithms, and foundational tools still require a deeper accumulation of basic research.
The National Center for Science and Engineering Statistics (NCSES), under the US National Science Foundation, estimates that US R&D investment in 2024 was approximately $993 billion.
Looking at 2023, which currently has complete classification data, out of the $937 billion in US R&D expenditure, basic research accounted for about $138 billion, or 15%; applied research was about $174 billion, or 19%; and experimental development was about $625 billion, or 67%. NCSES notes that from 2021 to 2024, the proportion of experimental development in the US remained basically stable at around 67%. The US also spends the majority of its money on product and technology development; it is not just a group of university professors sitting in labs handling all national innovation. However, its proportion of basic research is about twice that of China.
The gap between 15% and around 7% reflects not only budgetary arrangements but also the result of the long-term division of labor in the scientific research systems of the two countries.
The US possesses a massive scale of research universities, national laboratories, federal research programs, and corporate research institutions. Universities and the public research system are responsible for exploring many directions that do not yield commercial returns in the short term, while enterprises translate these achievements into products. In fiscal year 2024, R&D expenditure by US higher education institutions reached $118 billion, of which the federal government provided 55% of the funding.
Enterprises in China, on the other hand, undertake a higher proportion of R&D activities. Enterprises are naturally closer to the market, but corporate R&D is inherently constrained by revenue, profit, and product cycles. It is unrealistic to demand that an enterprise in fierce competition invest in a basic project for ten years without seeing any orders.
This is also why basic research cannot be simply left to the market.
Massive Differences in R&D Investment Between China and the US in Semiconductors
Total national R&D investment covers medicine, agriculture, energy, aerospace, software, defense, and a large amount of social science research. It can only illustrate a country's overall scientific research resources and cannot be directly equated with semiconductor investment.
When the focus is truly narrowed down to the chip industry, the differences become even more apparent.
The SIA report shows that R&D expenditure by US-headquartered semiconductor companies reached $76.8 billion in 2025, an increase of 10.3% from 2024. According to the EU Industrial R&D Investment Scoreboard cited in the report, R&D expenditure by US semiconductor companies accounts for about 15% of sales, while Europe is at 14.6%, South Korea at 10.6%, China at 10%, Japan at 8.5%, and Taiwan, China at 8.1%.
Here, the statistical boundaries must also be noted. This proportion comes from a sample of enterprises and SIA analysis, not a complete census of all semiconductor companies in each economy. However, it at least indicates that US chip companies not only have a large absolute revenue scale but are also willing to continue investing a high proportion of their revenue into next-generation technologies.
The global sales of US semiconductor companies in 2025 were $425 billion, accounting for 53.4% of global sales. A large market share brings high revenue, high revenue supports high R&D investment, and R&D outcomes in turn help companies maintain product premiums.
This is the most difficult part of the US semiconductor industry to replicate.
It is not simply about investing more in research funding, but about possessing world-class enterprises such as NVIDIA, Broadcom, Qualcomm, AMD, Applied Materials, Lam Research, KLA, Synopsys, and Cadence, which have formed commercial closed loops in chip design, equipment, EDA (Electronic Design Automation), and system software, respectively. Corporate R&D does not end when the budget is spent. R&D outcomes go into products, products enter the global market, and profits are reinvested into next-generation R&D. Once this cycle is established, its efficiency is often higher than relying solely on external funding.
The problem with China's semiconductor industry is not just insufficient total investment.
During the same period, Chinese companies in chip design, manufacturing equipment, materials, and wafer fabrication have all been increasing their R&D expenditure, but the revenue scale of many enterprises remains limited. The proportion of R&D investment to revenue may not be low, but the absolute amount is hard to compare with international leaders. Some enterprises also need to simultaneously catch up in products, processes, software, and customer validation, spreading limited funds across multiple directions.
An enterprise with R&D investment accounting for 30% of its revenue is not necessarily richer than an international leader with a 15% investment ratio. The ratio represents the willingness to invest, while the absolute amount determines whether many projects can be sustained.
R&D Investment is Growing Rapidly, So Why Are Some Technologies Still Hard to Catch Up On?
The particularity of semiconductor technology lies in the fact that it is not a race where steadily moving forward is achieved by evenly distributing resources.
An advanced manufacturing chain involves materials, equipment, components, processes, EDA, IP (Intellectual Property), design, packaging, and testing. Achieving 90 points in the vast majority of segments cannot offset scoring only 50 points in a critical segment.
If a lithography machine lacks a critical optical component, the overall performance of the machine cannot be improved; completing laboratory validation for etching equipment does not mean it can run continuously in a wafer fab; design software being able to run a simple chip does not mean it can support complex designs with tens of billions of transistors; and a successful tape-out does not equal customers being willing to entrust their next-generation products to it. This makes it difficult to measure the conversion rate of semiconductor R&D using the number of papers, the number of patents, or even the R&D budget.
More practical indicators include: whether equipment can enter the customer's production line and what the mean time between failures is; whether materials can pass long-term validation and how good the batch consistency is; whether process yield can be continuously improved; whether EDA tools can support real large-scale projects; and whether chips can be mass-produced and continue to secure customer orders in next-generation products. These indicators are not very suitable for the first page of project application materials, yet they determine whether R&D investment can ultimately be transformed into industrial capability. China's powerful manufacturing system and massive market provide unique advantages in this process.
New equipment and materials can find trial scenarios more quickly, and chip companies also have the opportunity to iterate repeatedly in communications, automotive, industrial, consumer electronics, and data centers. However, a large market does not automatically bring high-quality validation. If customers are only willing to try it once under policy requirements and switch back to the original supplier for the next-generation product, the R&D closed loop is still not established.
Truly effective domestic substitution is not about a product being procured once, but about its performance, cost, and service being repeatedly purchased by customers. When discussing China's R&D structure, basic research and engineering should not be placed on opposite sides of a scale. The easiest trap to fall into is thinking that because the proportion of basic research is low, more money should be taken from engineering development to shift towards basic research.
Things are not that simple. China's semiconductor industry currently still has a large number of engineering problems to solve. Equipment stability, material purity, process windows, packaging thermal management, chip software ecosystems, and customer service all require continuous investment. Weakening engineering investment will not automatically generate basic innovation; instead, it may cause already achieved technological results to stall at the prototype stage. A more reasonable direction is not to choose one over the other, but to connect the two ends.
Basic research requires more stable, long-term funding, especially in directions related to the underlying capabilities of semiconductors, such as materials, physics, chemistry, mathematics, and computer architecture. For such projects, industrial revenue cannot be evaluated every three years, nor can every paper be required to immediately correspond to a product line.
Engineering, on the other hand, requires better pilot-scale platforms, public testing conditions, and customer validation mechanisms. Technologies developed by universities and research institutions, without engineering teams taking over and without the participation of wafer fabs and equipment companies in validation, will find it hard to bridge the gap from the lab to mass production. In the field of semiconductor research in China, there is no lack of project approvals, nor is there a lack of published results.
What is truly scarce is a mechanism that can continuously work on a technology for ten years, endure multiple failures, and ultimately refine it to the point where customers are willing to use it.
As R&D Investment Enters the First Tier, Efficiency Issues Will Become Increasingly Prominent
When a country's annual R&D investment is only hundreds of billions of yuan, increasing the budget itself can solve many problems. When the investment approaches 4 trillion yuan, the situation changes. New funds are still useful, but the marginal improvement brought by each additional yuan of investment begins to depend more on resource allocation. Duplicate construction of laboratories and industrial parks among local regions consumes funds; enterprises choosing to roll out too many product lines simultaneously to secure project support consumes funds; and the lack of shared equipment and pilot-scale platforms, leading to duplicate procurement by each team, also consumes funds.
The semiconductor industry especially needs concentrated resources. Certain directions require parallel exploration of multiple technical routes and cannot just bet on a single answer; however, multi-route competition does not mean dozens of regions building similar projects. For equipment and materials fields that require large investments, have long cycles, and few customers, excessive dispersion will instead leave every team short of money, short of talent, and lacking real customers.
The more difficult task for R&D policy in the next stage is not to continue proving "we are also investing," but to allow projects to be eliminated, let funds exit ineffective directions, and simultaneously provide long-term support to teams that are making real progress. This is much harder than announcing a new investment figure.
Chip competition certainly involves comparing financial investment, but it also compares scientific planning and organizational capabilities. China's R&D investment catching up with the US first illustrates that the continuous increase in scientific and technological investment over the past decade has generated a scale effect. China is no longer relying on low-cost manufacturing to absorb technology diffusion, but has become one of the economies with the largest investment in global R&D resources.
This shift should not be downplayed. A massive engineering workforce, a complete manufacturing chain, and continuously increasing corporate R&D expenditure have indeed improved China's ability to solve complex industrial problems. The progress China has made in certain semiconductor equipment, materials, packaging, memory, and mature process nodes is not the result of a single policy, but the accumulation of long-term investment. However, surpassing a certain country in R&D expenditure does not come with a button that automatically cashes in technological leadership.
Semiconductor competition compares money, but also compares basic science accumulation, talent density, corporate revenue scale, customer validation, global markets, and industrial synergy. More importantly, it compares who can endure uncertainty for a longer period.
Engineering development can set quarterly targets, but basic research often has no clear endpoint. Equipment can pass acceptance tests, but yield and stability still require years of polishing. An enterprise can secure its first customer through policy, but the second customer must be won by the product itself. China has already solved the problem of "whether it has the capacity for large-scale investment."
What will truly determine the gap going forward is whether this money can pass through papers, projects, and prototypes, and ultimately remain in products, production lines, and customer orders.