TSMC's AI Boom Lifts Chinese Foundries via Mature-Node Overflow

What happened

Global foundry revenue reached $86 billion in the first quarter, up 23% year-over-year, driven by AI chip demand. TSMC posted 41% revenue growth in the quarter and 30% growth through May, as customers ordered more leading-edge wafers and advanced packaging for AI accelerators. The surge in AI orders filled TSMC's advanced-node capacity, causing mature-node orders-chips for power management, display drivers, and consumer electronics-to shift to second-tier foundries. Chinese foundries SMIC and Nexchip grew 12% and 19% respectively, while Taiwan's UMC and Vanguard rose 10% and 14%. The overflow was amplified for Chinese fabs by domestic semiconductor localization and rising wafer prices across both 8-inch and 12-inch processes. TSMC also raised its AI accelerator revenue growth forecast to the mid-to-high 50% range through 2029 and announced plans to expand capacity with up to 10 new or upgraded fabs in Taiwan and continued investment in Arizona. UBS raised its TSMC price target to NT$3,400, citing sustained AI demand and expected capital expenditure increases through 2028.

Why it matters

The dynamic illustrates how TSMC's focus on leading-edge AI work is reshaping the broader foundry market. While TSMC captures the high-value AI chip business-driving its revenue and reinforcing its dominance-the mature-node orders it can no longer accommodate are flowing to competitors a tier below. For TSMC, this means sustained pricing power and elevated margins as it allocates scarce advanced capacity to the highest-value customers, but it also means second-tier foundries gain revenue and market share in mature nodes. Chinese foundries benefit disproportionately from the overflow because domestic customers are sourcing locally and wafer prices are rising structurally, creating a tailwind even without leading-edge competition. The shift also highlights supply constraints across the AI chip supply chain-EUV lithography, high-bandwidth memory, and advanced packaging-that are affecting cloud capacity, AI project timelines, and infrastructure planning for enterprises. TSMC's capacity expansion and price increases expected by 2027 signal that these constraints are likely to persist, affecting both AI infrastructure buyers and foundry competitors.

Bigger picture

The foundry sector is bifurcating as AI demand concentrates at the leading edge. TSMC's leadership in advanced nodes and packaging leaves it uniquely positioned to capture AI infrastructure spending, while second-tier foundries absorb the overflow and serve domestic markets. The $86 billion quarterly foundry revenue underscores the scale of the AI-driven boom, but it also exposes the supply chain vulnerabilities-tight EUV capacity, sold-out HBM production through 2026, and packaging bottlenecks-that shape how quickly AI workloads can scale. Geopolitical factors add complexity: U.S. CHIPS Act funding is supporting TSMC's Arizona expansion to $165 billion, creating a domestic hedge, but Taiwan remains the center of leading-edge production. Meanwhile, Chinese foundries are gaining ground in mature nodes, supported by localization and rising wafer prices, even as they remain cut off from EUV technology. Samsung's push to secure 2-nanometer orders from Tesla and Intel's foundry ambitions add competitive pressure, but neither has matched TSMC's yields or scale at 3 nanometers and below. For enterprise buyers, the supply constraints mean fewer available GPU instances, longer lead times, less predictable pricing, and a need for flexibility across accelerator options and cloud commitments.

What to watch

TSMC's July 16 second-quarter earnings call will provide updated guidance on AI demand, capital expenditure plans, and capacity expansion. Investors will look for signals on whether AI orders remain robust, how quickly the 2-nanometer process ramps in the second half of 2026, and whether price increases expected by 2027 materialize. The progress of TSMC's Arizona fabs and advanced packaging partnership with Amkor will indicate how effectively the company is diversifying geographically while managing higher overseas costs. For Chinese foundries, watch whether domestic localization and wafer price strength persist through year-end, as Counterpoint expects, and whether SMIC and Nexchip can sustain double-digit growth without access to leading-edge technology. Broader supply chain indicators-ASML's EUV shipment pace, Micron's HBM capacity additions beyond 2026, and TSMC's CoWoS packaging throughput-will signal whether bottlenecks ease or tighten further. Finally, track how enterprise AI buyers adjust infrastructure plans in response to supply constraints, including shifts toward custom accelerators, multi-cloud strategies, or on-premises GPU clusters.