Chip suppliers gain leverage as AI infrastructure spending surges

Semiconductor suppliers are capturing a large share of AI infrastructure spending by controlling critical technologies across chip design, manufacturing and production equipment. Combined capital expenditure by Microsoft, Google, Amazon and Meta has reached 725 billion US dollars in 2026, reinforcing demand for accelerators, advanced fabrication and networking. The spending supports an interconnected supply chain whose strength depends on continued investment in larger models and data centers.

NVIDIA’s advantage extends beyond GPUs to CUDA, whose embedded software ecosystem makes switching platforms costly. TSMC supplies much of the advanced manufacturing capacity, while ASML holds a unique position as the sole supplier of EUV lithography machines. Broadcom benefits from customers developing custom accelerators and buying networking chips. AMD offers an alternative to NVIDIA through its accelerators and ROCm software, while Intel is pursuing a manufacturing comeback. Apple uses internally developed chips to support its integrated devices and cloud ambitions, and Arm collects architecture licensing fees and royalties.

These advantages also create vulnerabilities. Export restrictions threaten sales, dependence on TSMC exposes suppliers to geopolitical disruption, and customers’ custom chips could reduce purchases of general-purpose GPUs. Sustained expansion assumes AI investment will generate adequate returns. Lower computing utilization or procurement cuts could weaken orders just as suppliers commit to additional production capacity, exposing the fragility beneath the sector’s profitability.