Global Semiconductor Industry 2025: US Dominance, China's Catch-up, and a New Round of Supply Chain Restructuring

Global Semiconductor Industry 2025: US Dominance, China's Catch-up, and a New Round of Supply Chain Restructuring

In November 2025, Yole Group released two major industry reports — *Global Semiconductor Device Industry Overview (Second Half of 2025)* and *China's Semiconductor Industry 2025*. The two reports present complementary perspectives on the structural changes taking place in the global semiconductor industry: a strong market rebound, continued dominance of the value chain by U.S. design giants, and China's unprecedented push toward vertical integration from equipment to devices.

According to Yole's forecasts, global semiconductor device revenue will reach $743 billion in 2025, up 14% from 2024. This growth extends the recovery momentum seen since 2024, driven by the enormous demand for logic and memory chips from AI, as well as the completion of inventory adjustments across end markets.

The geographic concentration of the global market remains striking. In 2024, U.S. companies — led by fabless leaders such as Nvidia and Broadcom — captured 56% of the global semiconductor device market. Mainland China, meanwhile, continues to play a critical role in electronics manufacturing and consumer markets, accounting for roughly one-third of global semiconductor usage in locally assembled electronic systems. Yet on a demand basis, China has already been surpassed by U.S. spending and is no longer the world's largest semiconductor device demand region. Behind these figures lies a deeper power shift in the global semiconductor value chain: design leadership, manufacturing division of labor, and national strategic competition are being reinterwoven.

Structural Features of Global Growth: Logic, Memory, and AI Demand in Tandem

Looking at device categories, Yole expects the structure of the global semiconductor market in 2025 to remain relatively stable: logic and processors account for 40%–50%, memory for 20%–30%, power, analog, and discrete devices for 17%–23%, and optoelectronics and sensors for 12%–14%. Among them, growth in logic and memory is driven mainly by continued ramp-up of data center AI accelerators and HBM memory, while power devices benefit from growth in electric vehicles and industrial power.

One notable trend is that although China was no longer the largest demand country in 2024, it still attracts about one-third of semiconductor devices for local assembly as a global hub for consumer electronics manufacturing. Chinese electronics manufacturers' willingness to procure locally made devices is rising, creating incremental room for domestic Chinese design houses and foundries.

China's Strategic Layout: Rapid Capacity Expansion, but Gaps in Technological Depth

One of the core conclusions of the Yole report is that China is building a "broad and complete" vertical semiconductor ecosystem. Over the past five years, China's domestic OSAT revenue has grown by 57%; foundry revenue reached $16.4 billion; and mainland wafer foundry capacity currently equals 71% of local electronic assembly demand, reaching 112% of local electronics demand. In other words, mainland manufacturing capacity has already exceeded its internal reference demand and is approaching a "sufficient" level, yet it still does not cover all assembly and manufacturing requirements.The equipment segment remains the biggest weakness. Although localization of equipment in China is advancing, overall coverage remains limited. Yole expects that by 2030, revenue of local equipment companies is expected to reach $33 billion, implying that penetration in some segments could rise to 52%. On the other hand, export controls imposed by the United States, the Netherlands, and Japan on advanced equipment still choke China's breakthrough into advanced process nodes.

In the device segment, Yole expects that the global market share of China-mainland local devices could increase to about 10% by 2030, with the combined annual revenue of local manufacturers approaching $100 billion. At present, Chinese design companies have gained some position in areas such as MCUs, power devices, analog chips, mature logic, and certain AI ASICs, but overall they remain focused on the mid-to-low-end market and domestic demand.

Industrial chain analysis: structural competition among upstream, midstream, and downstream

Upstream: the "bottleneck" segment of equipment and materials

Semiconductor equipment is the most closely watched link in China's supply chain. Yole's statement that "equipment localization is progressing slowly but has huge potential" essentially reflects that domestic equipment has been able to replace some imports in mature process nodes, but there are still significant technical bottlenecks in high-end lithography, metrology, deposition/etch, and advanced packaging equipment. If local equipment revenue reaches $33 billion by 2030, it would mean China becomes a buyer-supplier that cannot be ignored in the global equipment market, changing the current landscape dominated by a few companies from the United States, Japan, and the Netherlands in equipment supply.

Materials and EDA/IP are also hidden barriers. The localization rates of photoresists, specialty gases, large silicon wafers, and high-purity chemicals remain relatively low. As export controls expand, China's self-built capacity in the materials segment will also become a long-term goal.

Midstream: domestic substitution in manufacturing faces an imbalance between "quality" and "quantity"

On the chip manufacturing side, the self-sufficiency rate of mature process capacity at China-mainland foundries has been improving, but Yole points out the imbalance with two striking figures: foundry capacity has reached 112% of local electronics demand, yet it only covers 71% of local electronics assembly. This set of data shows that China's capacity expansion has been far faster than the effective alignment with local design demand. In fact, a large amount of capacity depends on overseas customers or export processing demand, while high-end chips used in domestic electronics assembly still rely on imports.

In the packaging and testing segment, the rapid growth of Chinese OSAT companies (five-year revenue +57%) demonstrates competitiveness in mature packaging. However, in the most critical advanced packaging technologies for AI chips (such as CoWoS, 2.5D/3D packaging), a few players such as TSMC still hold a monopoly, and mainland Chinese companies find it difficult to enter the highest-end market.

Downstream: shifting from export-driven to domestic-demand-driven, but high-end demand remains constrainedChina’s electronics manufacturing industry was once highly dependent on exports, but the share of domestic demand is now rising. Yole notes that the scale of electronics manufacturing is equivalent to 158% of domestic electronics demand, which means China’s industry still has a huge export volume. However, as domestic consumption upgrades and locally made chips penetrate into communications, home appliances, automotive and other fields, downstream system OEMs are reserving an increasing share for local semiconductor companies. This structural adjustment on the demand side will provide China’s semiconductor materials companies and equipment makers with a long-term application testing ground.

Competitive Landscape: The US Dominates Design, Asia Manufactures, and China’s Challenge Unfolds in Layers

According to Yole’s data, US companies capture 56% of global device revenue share, thanks mainly to their undisputed dominance in the design segment. NVIDIA and Broadcom are the biggest winners of the AI era, and behind them are TSMC’s and UMC’s advanced process and advanced packaging capacity. Therefore, the real global semiconductor power structure is shifting from the traditional pattern of “US design + Taiwan manufacturing + Korean memory + European equipment/automotive chips + Chinese assembly” to a multi-track pattern of “US design + Taiwan advanced manufacturing + China’s mature manufacturing and local ecosystem.”

Along this chain, the challenges China faces are clear:

• Design: IP barriers are obvious in AI GPUs, high-end CPUs and other fields, but accelerated breakthroughs are possible in AI ASICs, edge AI and automotive-grade chips.
• Manufacturing: Advanced process nodes cannot obtain EUV lithography machines, leaving process technology at 7nm and its improved versions. While large-scale expansion of mature processes can generate revenue scale, it is difficult to support generational iteration of equipment and EDA.
• Equipment and materials: Certification cycles with B2B customers are long, but driven by national security needs and the “internal circulation” incentive, the localization penetration rate will continue to rise.

Meanwhile, Southeast Asia, Japan and Europe are also trying to benefit from the supply chain restructuring. Malaysia, Vietnam and other countries continue to take on packaging and testing capacity. Japan has clear advantages in equipment and materials, and is strengthening its role in the advanced manufacturing landscape through the Kumamoto fab built with TSMC. Europe, meanwhile, is consolidating its position in power semiconductors and automotive chips, and is attracting wafer fabs through the European Chips Act.

Investment Perspective: From Cyclical Boom to Structural Narrative

The semiconductor market entered a new upcycle in 2024–2025, but investors have already begun to distinguish which growth is a cyclical rebound and which is lasting structural change. AI chips and high-bandwidth memory are the most certain earnings drivers at present. China’s semiconductor investment, in turn, will mainly revolve around the main thread of localization, with equipment, materials, EDA and mature-process foundry being the areas of highest policy priority.Yole has identified two targets for mainland China by 2030—a 10% share of the global device market and roughly $100 billion in annual device revenue, along with $33 billion in equipment revenue—which implies that the medium-to-long-term revenue growth of mainland Chinese listed semiconductor companies will likely outpace the global average. However, they must strike a balance among massive capital expenditures, technological backwardness, and export controls, so profit margins may remain under pressure for a prolonged period.

Long-Term Outlook: A Dual-Track World for the Next Five to Ten Years

Extending the horizon to the next three to five years, global semiconductor companies must adapt to two parallel and mutually compatible logics. One is the AI-driven expansion of high-performance computing, which continuously tightens supply of advanced logic, HBM, and advanced packaging capacity. The other is the boundary drawn by geopolitics, as countries launch trillion-scale investments in semiconductor supply-chain autonomy, creating splits within the same industry between “advanced and mature,” “authorized and self-developed,” and “global standards and local standards.”

By 2030, the most likely scenario is that the United States continues to dominate global high-end design, Taiwan retains the core position in advanced foundry, Korea holds on to memory, and mainland China makes significant headway in mature process nodes and domestic equipment/materials, while still trailing the world frontier in advanced process by more than two generations. The semiconductor industry will shift from “hyper-globalization” to “managed cooperation,” with a complexity exceeding any period in the past.

For investors and industry strategy setters, the most important takeaway from the Yole report is not the $743 billion in revenue itself, but the fact that global semiconductor growth drivers are diversifying and regionalizing. Value creation no longer comes solely from lower power consumption and higher performance of chips, but also from supply-chain security and resilience. “In the future, value growth will increasingly depend on understanding geo-economic dividing lines and on the strategic choices countries make under constraints of capital, technology, and resources.” That, perhaps, is exactly the true keyword of the global semiconductor industry in 2025.