Foundry sourcing in 2026: TSMC, Samsung, Intel Foundry, GlobalFoundries, and the geopolitics of node access

Foundry sourcing is the most concentrated procurement decision in semiconductor hardware. TSMC holds roughly sixty percent of global foundry revenue and an even higher share of leading-edge node capacity. Samsung Foundry is the second player at advanced nodes. Intel Foundry Services is rebuilding capacity with the 18A and 14A nodes. GlobalFoundries serves trailing-edge specialty nodes. UMC and SMIC fill out the rest. Lean SupplAI was built to track foundry capacity, node availability, and customer-allocation patterns precisely because in semiconductor procurement, foundry choice is upstream of every other decision.

The geopolitical environment shifted materially in 2024 and 2025. The US CHIPS Act, EU Chips Act, and Japan's foundry incentives have funded major capacity expansion outside Taiwan. The shift is real but slower than the headlines suggest, and the procurement implication is nuanced: leading-edge capacity is still concentrated in Taiwan, but trailing-edge and specialty nodes are diversifying meaningfully.

TSMC: the dominant choice

TSMC operates at every relevant node from 28nm down through 2nm (production ramp scheduled 2025-2026). Customer concentration includes Apple (largest customer, exclusive at the leading edge), NVIDIA (largest GPU customer), AMD, Qualcomm, MediaTek, and Broadcom. The Arizona Fab 21 site is ramping 4nm capacity for US customers, with 3nm and 2nm planned for later phases. For programs requiring leading-edge nodes, TSMC is typically the only credible choice through 2027.

Samsung Foundry: the second source

Samsung Foundry runs leading-edge nodes (3nm, 4nm, 5nm) plus mature lines. Customer wins include Tesla, Qualcomm (some Snapdragon lines), and parts of Google's Tensor program. Yield and reliability gaps versus TSMC have narrowed in 2024 and 2025. For procurement teams looking for leading-edge alternatives to TSMC, Samsung is the most production-credible option, with the Texas Taylor fab adding US capacity from 2025-2026.

Intel Foundry: the rebuilding play

Intel Foundry Services (IFS) is in the rebuild phase under the 18A node, with first volume production scheduled 2025-2026 and 14A planned for 2027. Early customer wins include Microsoft and Qualcomm. The PowerVia backside-power technology and gate-all-around (GAA) architecture are differentiators if execution holds. For procurement teams considering Intel Foundry, the calculus is risk-tolerance: capacity and US-domestic sourcing benefits versus newer-process risk.

GlobalFoundries: trailing-edge specialty

GlobalFoundries left the leading-edge race in 2018 and now focuses on differentiated specialty nodes: 12LP/12LPP (FD-SOI), 22FDX (FD-SOI for low-power and RF), and BCD processes for power management. The customer base includes automotive (Bosch, Continental, AMD's Pensando), industrial, and IoT programs. For mature-node sourcing in 2026, GlobalFoundries with TSMC's mature lines and UMC are the dominant non-Chinese options.

The CHIPS Act and capacity geography

The US CHIPS and Science Act has allocated roughly fifty-three billion dollars in incentives, with the largest awards going to TSMC Arizona, Intel Ohio and Arizona, Samsung Texas, Micron New York, and GlobalFoundries New York. The EU Chips Act similarly funds capacity expansion at TSMC's Dresden fab, Intel's Magdeburg site, and existing European fabs. By 2027 the US-domestic share of leading-edge capacity will be meaningfully higher than 2024, but Taiwan will remain dominant for the most-advanced nodes.

How Lean SupplAI tracks foundry capacity

Lean SupplAI indexes foundry capacity by node, fab site, customer-allocation patterns, expansion roadmap, and geopolitical exposure. For procurement teams running chip programs, Lean SupplAI surfaces node availability against the program's volume and timeline requirements, with the supporting capacity data visible inline.

What sets Lean SupplAI apart