Semiconductors: The Tax on Every AI Dollar Spent

Every dollar spent on AI infrastructure flows through silicon first. That makes the semiconductors theme less a sector bet and more a toll road on the entire technology economy. The 83 companies in this graph include chip designers, fabrication-equipment giants, materials specialists, and the connectors that stitch it all together — a full vertical stack from raw wafer to finished system.

The cleanest expression of that toll-road logic sits in the equipment layer. Before Advanced Micro Devices or Micron Technology can ship a single chip, fabs must buy the machines that etch, deposit, and inspect at atomic scale. That upstream position is structurally sticky: chipmakers cannot switch equipment vendors mid-node without requalifying an entire process, which takes years.

Equipment Makers Control the Chokepoint

KLA Corp owns process-control and inspection. Lam Research dominates etch and deposition. Applied Materials touches nearly every step of wafer fabrication. All three benefit from the same demand signal — rising chip complexity — regardless of which designer wins the AI accelerator race. More layers in a chip means more passes through their machines. Tighter geometries mean more inspection runs. The secular trend toward denser, more powerful chips is a direct revenue driver for this group.

Teradyne adds the testing dimension. Every chip that leaves a fab gets tested, and the shift toward high-bandwidth memory and AI accelerators raises test time per unit. That is a volume and duration tailwind simultaneously.

Materials and Connectivity Complete the Stack

Advanced nodes demand ultra-pure chemicals, specialized gases, and precision materials that standard suppliers cannot match. Entegris sits in that position — supplying the materials that enable sub-5nm manufacturing. Switching costs are as high here as they are in equipment; contamination risk makes fabs deeply conservative about their supply chains.

Connectivity is the other overlooked layer. Amphenol makes the interconnects that move data inside servers, between chips, and across rack infrastructure. As AI clusters scale from hundreds to hundreds of thousands of GPUs, interconnect density and signal integrity become critical. Coherent Corp addresses the optical side of that same problem, supplying the transceivers that move data at the speeds AI training demands.

Further out in the design stack, ARM Holdings licenses the instruction-set architecture that underlies the majority of chips shipped globally — smartphones, servers, and increasingly AI edge devices. Astera Labs targets the connectivity bottleneck specifically inside AI infrastructure, designing chips that keep GPUs and memory from throttling each other.

How to Hold the Theme

The ETF library offers several entry points. XSD — the SPDR S&P Semiconductor ETF — tracks the sector directly with equal weighting, giving more exposure to mid-caps than a market-cap index would. Broader funds like QQQ and XLK carry significant semiconductor weight but dilute the exposure across the wider technology landscape. For investors who want the theme concentrated, XSD is the cleaner instrument.

The risk is cyclicality. Semiconductor equipment spending follows fab-investment cycles, and downturns compress order books hard. The current cycle is unusually extended because AI capital expenditure has kept leading-edge capacity demand elevated well past the typical correction window. That does not eliminate the cycle — it defers it. Investors in this theme are making a judgment that AI infrastructure spending remains durable enough to sustain above-trend equipment and materials demand through the next several years.