Capacity expansion at existing fabs. Co-CEO Zhao Haijun stated that wafer starts are "far exceeding" earlier forecasts, and SMIC is actively evaluating adding equipment at its existing factories to add more mature-node capacity . Analysts forecast 22%/14%/12% revenue growth in 2026/27/28 from higher utilization and ASP gains .

Benefiting from the 'AI siphon effect.' As global tier-1 foundries like TSMC and Samsung shift capacity to premium AI accelerators and HBM, a mature-node capacity crunch has emerged, driving foreign clients to shift orders back to SMIC . Q2 2026 shipments hit 2.9 million 8-inch-equivalent wafers, up 14% quarter-on-quarter . Chinese foundries' share of global legacy-node capacity in the 22nm to 40nm range is expected to reach 37% in 2026 and 41% in 2027, up from 32% in 2025 .

Financial results surged. Q2 2026 net profit more than tripled year-over-year to $479.2 million, beating analyst estimates, with revenue crossing $3 billion for the first time .

Specific Products Driving Demand Through 2027

Not all chips are created equal in the AI era. The following mature-node categories are the primary growth engines for SMIC:

BCD (Bipolar-CMOS-DMOS) power-management chips. This is the single most-cited product line. BCD chips are essential for power management in AI servers and data centers, and SMIC's co-CEO said orders are already locked in through the end of 2027 . This platform supports efficient power management and is essential to data centers, which require enormous amounts of electricity .

Power-management ICs (PMICs). A broad category of mature-node chips used to regulate power in AI servers, general-purpose servers, and edge AI devices. TrendForce notes that AI server demand has lifted 8-inch foundry utilization for power-related ICs toward 85–90%, up from 75–80% in 2025 . Products such as PMICs and power discrete devices remain largely dependent on 8-inch processes .

MCUs (microcontrollers) and analog chips. Zhao outlined five pathways through which AI demand flows into SMIC's fabs, with MCU and analog platforms among the strongest categories. These are considered second-tier causation pathways, where AI demand indirectly pulls orders into SMIC's fabs .

Logic and other auxiliary chips. SMIC also sees strong demand for the lower-profile "around-the-GPU" chips — logic ICs, display drivers, and other supporting components used in AI server motherboards and edge AI devices . As Zhao put it, these are the chips that make the "expensive parts work."

The Key Takeaway

The AI boom is not solely about cutting-edge GPU nodes. It is creating a durable multi-year tailwind for SMIC's mature-node (28nm and larger) power management and specialty chips, with pricing power and capacity expansion set to continue through at least 2027. The "siphon effect" — where global foundry capacity shifts away from mature nodes to premium AI accelerators — has created a structural tightness that SMIC is uniquely positioned to exploit, driving both volume and pricing gains for years to come.

Capacity expansion at existing fabs. Co-CEO Zhao Haijun stated that wafer starts are "far exceeding" earlier forecasts, and SMIC is actively evaluating adding equipment at its existing factories to add more mature-node capacity . Analysts forecast 22%/14%/12% revenue growth in 2026/27/28 from higher utilization and ASP gains .

Benefiting from the 'AI siphon effect.' As global tier-1 foundries like TSMC and Samsung shift capacity to premium AI accelerators and HBM, a mature-node capacity crunch has emerged, driving foreign clients to shift orders back to SMIC . Q2 2026 shipments hit 2.9 million 8-inch-equivalent wafers, up 14% quarter-on-quarter . Chinese foundries' share of global legacy-node capacity in the 22nm to 40nm range is expected to reach 37% in 2026 and 41% in 2027, up from 32% in 2025 .

Financial results surged. Q2 2026 net profit more than tripled year-over-year to $479.2 million, beating analyst estimates, with revenue crossing $3 billion for the first time .

Specific Products Driving Demand Through 2027

Not all chips are created equal in the AI era. The following mature-node categories are the primary growth engines for SMIC:

BCD (Bipolar-CMOS-DMOS) power-management chips. This is the single most-cited product line. BCD chips are essential for power management in AI servers and data centers, and SMIC's co-CEO said orders are already locked in through the end of 2027 . This platform supports efficient power management and is essential to data centers, which require enormous amounts of electricity .

Power-management ICs (PMICs). A broad category of mature-node chips used to regulate power in AI servers, general-purpose servers, and edge AI devices. TrendForce notes that AI server demand has lifted 8-inch foundry utilization for power-related ICs toward 85–90%, up from 75–80% in 2025 . Products such as PMICs and power discrete devices remain largely dependent on 8-inch processes .

MCUs (microcontrollers) and analog chips. Zhao outlined five pathways through which AI demand flows into SMIC's fabs, with MCU and analog platforms among the strongest categories. These are considered second-tier causation pathways, where AI demand indirectly pulls orders into SMIC's fabs .

Logic and other auxiliary chips. SMIC also sees strong demand for the lower-profile "around-the-GPU" chips — logic ICs, display drivers, and other supporting components used in AI server motherboards and edge AI devices . As Zhao put it, these are the chips that make the "expensive parts work."

The Key Takeaway

The AI boom is not solely about cutting-edge GPU nodes. It is creating a durable multi-year tailwind for SMIC's mature-node (28nm and larger) power management and specialty chips, with pricing power and capacity expansion set to continue through at least 2027. The "siphon effect" — where global foundry capacity shifts away from mature nodes to premium AI accelerators — has created a structural tightness that SMIC is uniquely positioned to exploit, driving both volume and pricing gains for years to come.