AI competition reshapes global semiconductor landscape: SMIC's cyclical positioning and structural opportunities.

The global semiconductor industry is experiencing a structural upward cycle driven by artificial intelligence. Unlike past cyclical recoveries fueled by consumer electronics or mobile internet, the core driver of this cycle comes from the exponential expansion of AI computing demand—from GPU clusters required for training large models to customized chips for large-scale deployment on the inference side. AI is reshaping the entire semiconductor value chain, from design to manufacturing.

According to the latest Gartner forecast, global semiconductor revenue will exceed $1.3 trillion in 2026, an increase of about 60% compared to 2025, marking the fastest growth rate in nearly 20 years. AI-related chips will account for 30% of the industry's revenue, becoming the absolute core of growth. Against this backdrop, wafer foundry, as a core segment of semiconductor manufacturing, is experiencing dual changes in demand structure and competitive landscape. As the largest wafer foundry in China, where does SMIC stand in this cycle? How does AI demand transmit from the design side to the manufacturing side? To what extent have geopolitical factors reshaped the company's growth logic? We analyze from three dimensions: industry data, financial performance, and supply chain structure.

Global Semiconductor Cycle: AI-Driven Structural Upturn

The semiconductor market in 2026 exhibits characteristics distinctly different from historical cycles. Gartner data shows that global semiconductor revenue in 2025 was $805.3 billion, is expected to surge to $1.32 trillion in 2026, and further climb to $1.55 trillion in 2027. This growth pace far exceeds previous industry expectations—in 2021, analysts generally believed that annual chip sales would not break $1 trillion until 2030.

The core factor driving this surge is the concentrated release of AI infrastructure investments. Hyperscalers are expanding AI data centers at an unprecedented pace. According to CreditSights and Fortune data, hyperscaler capital expenditures are expected to reach $602 billion to $700 billion in 2026. The combined annual capital expenditure plans of Alphabet, Amazon, Meta, and Microsoft exceed $700 billion. A significant portion of these funds flows into AI accelerator procurement, which in turn transmits to the wafer foundry segment.

From the perspective of the industry chain transmission path, AI demand's pull on semiconductors follows a clear hierarchical structure. At the highest level are AI chip design companies—NVIDIA achieved full-year revenue of $215.9 billion in fiscal 2026, a year-over-year increase of 65%, with the data center business contributing $193.7 billion, accounting for nearly 90% of total revenue. Entering fiscal 2027, NVIDIA's quarterly revenue has exceeded $81.6 billion, with data center business reaching $75.2 billion, a year-over-year increase of 92%. Chip shipment volumes of this magnitude directly drive wafer foundry capacity.

The second level is the wafer foundry segment. Counterpoint Research data shows that in the first quarter of 2026, the global wafer foundry 2.0 market revenue grew 23% year-over-year, reaching $86 billion. As the main beneficiary of the AI-driven upward cycle, TSMC's first-quarter revenue grew 41% year-over-year, with advanced process nodes (7nm and below) accounting for 74% of wafer sales revenue. DIGITIMES estimates that global wafer foundry revenue will exceed $230 billion in 2026, a further 17% increase over 2025.

The third level is the broader chip manufacturing ecosystem, including advanced packaging, semiconductor equipment, and materials. The increasing complexity of AI chips is driving the concept of "wafer foundry 2.0"—a deep integration of wafer manufacturing, advanced packaging, and testing capabilities. Competitive advantage in the industry no longer depends solely on process technology but also on the ability to deliver comprehensive solutions at scale.

SMIC's Performance and Capacity Status

Against the backdrop of overall industry growth, SMIC delivered a landmark report card in the first quarter of 2026.

According to the company's Hong Kong stock exchange announcement, first-quarter sales revenue reached $13k, breaking through the single-quarter $2.5 billion mark for the first time, a year-over-year increase of 11.5% and a sequential increase of 0.7%. Under the A-share accounting standard, operating revenue was RMB 13.2k, up 8.1% year-over-year; net profit attributable to shareholders was RMB 15.5k. Gross margin was 20.1%, up 0.9 percentage points sequentially, exceeding the previous guidance range of 18% to 20%.

In terms of capacity, monthly capacity at the end of the first quarter reached 1.078 million wafers (8-inch equivalent), a year-over-year increase of 10.8%; capacity utilization remained high at 93.1%. Wafer shipments were 2.51 million wafers, a year-over-year increase of 9.5%. Average selling price (ASP) rebounded 0.9% sequentially to $999 per wafer. 12-inch wafer revenue accounted for 76.4%, with the share of high-end capacity continuing to rise.

From a downstream application structure perspective, consumer electronics remained the largest revenue source, accounting for 46.2%, a year-over-year increase of 27%; the industrial and automotive segment rose significantly from 9.6% in the same period last year to 14%, a year-over-year increase of 63%, becoming the fastest-growing sub-segment. Smartphone and computer/tablet revenue accounted for 18.9% and 13.6%, respectively.

A noteworthy change is in regional revenue structure. The China region's revenue share further increased to 88.9%, significantly expanding from 84.3% in the same period last year; the US region's share narrowed from 12.6% to 9.3%. This change reflects the trend of domestic customers accelerating the shift of orders to local foundries to ensure supply chain security amid heightened geopolitical uncertainty. SMIC co-CEO Zhao Haijun stated during the earnings call that AI demand is crowding out global mature process foundry and memory capacity, driving a large number of orders back to China.

For the second quarter, the company provided positive guidance: revenue is expected to increase 14% to 16% sequentially, with gross margin between 20% and 22%. Management clearly stated that they are "more optimistic about the overall operation this year."

How AI Demand Transmits to Wafer Foundry: SMIC's Supply Chain Position

The transmission of AI demand to the wafer foundry segment is not evenly distributed but exhibits significant structural characteristics. Understanding this transmission mechanism is key to assessing SMIC's degree of benefit in this cycle.

Scarcity premium of advanced process nodes. AI training chips (such as NVIDIA GPUs) have extremely high requirements for process technology, currently relying mainly on TSMC's 5nm, 4nm, and 3nm nodes. In the first quarter, TSMC's advanced process nodes accounted for 74% of wafer sales revenue, with 3nm at 25% and 5nm at 36%. This capacity is expected to remain fully loaded throughout the year, with related orders already booked into 2027. SMIC's most advanced N+3 process, according to SemiAnalysis teardown, is equivalent to TSMC's N6 level. This means that in the foundry segment for top-tier AI training chips, SMIC cannot directly replace TSMC in the short term. However, the capacity shortage of advanced process nodes is creating a "spillover effect"—when the world's most advanced capacity is fully occupied by AI training chips, other chips that previously used advanced nodes (such as high-end smartphone processors, AI inference chips, etc.) begin migrating to sub-advanced nodes, creating incremental demand for SMIC's N+2 and N+3 processes.

Rise of AI ASICs and diversified foundry demand. A notable structural change is the evolution of the AI chip market from a "GPU-only" paradigm to a "GPU + ASIC dual-track." JPMorgan expects the digital AI ASIC market to reach approximately $60 billion to $70 billion by 2026, maintaining a compound annual growth rate of 40% to 50% or more in the coming years. Institutions predict that ASIC chip shipments will be about 7.7 million units in 2026, with a market share of 45%, and will surpass GPU share to reach 58% in 2027. Unlike NVIDIA's general-purpose GPUs, ASICs are typically designed for specific customers and workloads, and their requirements for process nodes are more diverse—not all ASICs need the most advanced 3nm or 5nm; some inference chips can achieve competitive power efficiency at 7nm or even more mature nodes. This trend provides differentiated opportunities for wafer foundries with broad process coverage capabilities.

Supply-demand reversal in mature process nodes. The most surprising change in this cycle is occurring in the mature process segment. TrendForce points out that as AI-related capacity continues to crowd out wafer foundry resources, coupled with TSMC, Samsung, and other major players gradually reducing 8-inch and some 12-inch mature process capacity, the supply-demand structure for mature process nodes is rapidly tightening. Prices for some supply-constrained process nodes have shown a tendency to increase by 5% to 10% in the second to third quarters of 2026, and the price hike effect is expected to extend into 2027. SMIC has implemented a new round of price increases for mature process nodes starting July 2026, following earlier increases that drove a 5% to 6% sequential increase in ASP in the second quarter. The supply-demand reversal in mature process nodes has given SMIC improved pricing power in the area where it has the largest capacity share.

Structural push from localization substitution. Geopolitical factors are accelerating the "de-Americanization" of China's semiconductor supply chain. Continuing US export control measures—including designating SMIC as a "controlled facility" and proposing a comprehensive ban on DUV lithography machine exports to China—while constraining access to advanced process equipment, are also driving Chinese chip design companies to accelerate the transfer of orders to local foundries. SMIC's China region revenue share has risen from 84.3% in the same period of 2025 to 88.9% in the first quarter of 2026, a trend expected to continue. Reuters, citing Semicon China data, indicates that the global capacity share of Chinese wafer fabs at the 22-40nm mature nodes is expected to rise from 32% in 2025 to 41% in 2027. In 2026, mainland China's wafer foundry capacity is expected to exceed 4 million wafers per month (8-inch equivalent), accounting for 34.4% of the global total, with a compound growth rate of 13.4%, the highest growth rate globally.

Structural Opportunities and Constraints

Based on the above analysis, SMIC's structural opportunities in this AI-driven semiconductor cycle can be understood from three dimensions.

First, the upturn in mature process nodes is improving the profit base. SMIC's capacity structure is dominated by mature process nodes. The global supply-demand for mature process nodes shifting from loose to tight directly drives the company's capacity utilization to remain high (93.1%) and ASP to continue recovering. TrendForce predicts the price hike effect will extend into 2027, meaning SMIC is expected to benefit from the price upcycle in the coming quarters.

Second, the explosion in AI inference demand is creating new incremental markets. As the focus of the AI industry gradually shifts from "model training" to "inference applications," the demand structure for inference chips is changing. Inference chips have lower requirements for process advancement than training chips but place higher demands on cost, energy efficiency, and shipment scale. This market characteristic aligns more closely with SMIC's capacity and cost structure. According to market sources, Huawei has confirmed that the DeepSeek V4 model is powered by its Ascend 950 PR AI chip, and SMIC is considered the only Chinese company with the capability and capacity to support this chip using its N+3 process.

Third, foundry demand from domestic AI chip design companies is undergoing a structural shift. Under geopolitical constraints, Chinese AI chip design companies (including HiSilicon, Cambricon, Horizon Robotics, etc.) cannot rely on advanced capacity from overseas foundries like TSMC and must seek local alternatives. As a scarce provider of advanced process foundry capacity domestically, SMIC holds an irreplaceable position in this shift. CMB International, in a recent initial coverage report on SMIC's H-shares, noted that the company is "China's core wafer manufacturing platform, combining domestic scale, extensive mature and specialty process coverage, advanced logic capability, and deep involvement in the domestic fabless ecosystem," giving it a "Buy" rating with a target price of HKD 110.

However, the realization of these opportunities faces clear constraints.

Significant bottlenecks exist in catching up with advanced process technology. Although SMIC's N+3 process demonstrates competitiveness in some metrics—SemiAnalysis measurements show its minimum metal pitch is 32.5nm, about 10% more compact than Intel's 18A process's 36nm pitch—this is achieved purely with DUV without EUV lithography machines. In key indicators such as process maturity, yield, and production cost, N+3 still lags behind TSMC's N6. If US export controls continue to tighten, SMIC will face substantial bottlenecks in equipment acquisition for process development beyond N+3.

High capital expenditure will continue to suppress profit margins. First-quarter depreciation and amortization costs reached $10k, a year-over-year increase of 25.7%; capital expenditure was $2.51B. The depreciation pressure from capacity expansion is a major factor limiting further gross margin improvement. CMB International also noted that "depreciation will remain the main factor restricting margin recovery."

Uncertainty exists regarding the sustainability of AI infrastructure investment. The current explosion in AI chip demand is built on the scale of hyperscaler capital expenditures. If cloud vendors slow down their AI investment pace, or if the commercialization progress of AI applications falls short of expectations, the current supply-demand tightness could reverse. DIGITIMES also cautioned in its outlook about "bubble concerns in AI infrastructure investment."

Conclusion

The global semiconductor industry is in a structural upward cycle driven by AI. Gartner's forecast of industry revenue exceeding $1.3 trillion in 2026, with AI chips accounting for a third of the share, outlines the scale and intensity of this cycle. Against this backdrop, SMIC's growth logic is shifting from "consumer electronics-driven capacity expansion" to "structural upgrades driven by AI demand spillover and localization substitution."

Based on first-quarter data, the company has validated this trend in key indicators such as revenue scale ($17.62B), capacity utilization (93.1%), and gross margin (20.1%). The supply-demand reversal in mature process nodes is improving its profit base, the explosion in AI inference demand is creating new incremental markets, and the supply chain restructuring driven by geopolitics provides it with an irreplaceable strategic position. However, bottlenecks in advanced process technology, profit margin suppression from high capital expenditure, and uncertainty about the sustainability of AI investment constitute constraints.

SMIC's ultimate degree of benefit in this cycle depends on the evolution of three variables: the sustainability and magnitude of the mature process price hike cycle; the technological breakthroughs and yield ramp-up progress of N+3 and subsequent process nodes; and the speed and scale of order transfers from domestic AI chip design companies to local foundries. These three variables interact to determine whether China's largest wafer foundry can convert cyclical industry prosperity into a structural competitive advantage.

FAQ

Q1: What is the level of SMIC's most advanced process? Can it undertake AI chip foundry demand?

SMIC's most advanced process is N+3, which, according to SemiAnalysis teardown, is equivalent to TSMC's N6 level. This process has been used for the foundry of Huawei's Ascend series AI chips. Compared to TSMC's top-tier processes like 3nm and 5nm, there is still a generational gap. However, against the backdrop where domestic AI chip design companies cannot rely on overseas foundries, N+3 represents the highest level of foundry capacity available domestically, holding scarcity value.

Q2: How does AI demand affect the market landscape for mature process wafer foundry?

The expansion of AI chip capacity is crowding out global wafer foundry resources. At the same time, major players like TSMC and Samsung are gradually reducing mature process capacity, leading to a rapid tightening of the supply-demand structure for 8-inch and 12-inch mature process nodes. TrendForce expects the price hike effect to extend into 2027, with prices for some process nodes already showing a 5% to 10% increase in the second to third quarters of 2026. This directly drives profit improvement for SMIC, which is primarily focused on mature process nodes.

Q3: How did SMIC perform in the first quarter of 2026?

In the first quarter, sales revenue reached $1.36B, surpassing the single-quarter $2.5 billion mark for the first time, a year-over-year increase of 11.5%; gross margin was 20.1%, up 0.9 percentage points sequentially; capacity utilization was 93.1%. The company expects second-quarter revenue to increase 14% to 16% sequentially, with gross margin between 20% and 22%.

Q4: What is the long-term impact of US export controls on SMIC?

The US has designated SMIC as a "controlled facility" and proposes a comprehensive ban on the export of semiconductor manufacturing equipment to it. On one hand, this limits SMIC's ability to acquire key equipment such as advanced EUV lithography machines, creating bottlenecks for process development beyond N+3. On the other hand, it accelerates the transfer of orders from domestic chip design companies to local foundries, with SMIC's China region revenue share rising to 88.9%. In the short term, the order return effect dominates; in the long term, the equipment embargo poses a technical ceiling.