Barclays' technology research team has outlined a group of networking and interconnect suppliers it sees as key beneficiaries of an anticipated boom in AI infrastructure spending. The analysts estimate that annual expenditure by Western hyperscalers and AI research labs could exceed $1 trillion, which they say is more than $300 billion higher than prevailing consensus forecasts. Barclays also noted there could be additional upside tied to sovereign AI initiatives and investment from China before the cycle reaches its peak in 2028.
The projected spending uptick is driven by what Barclays characterizes as an ongoing AI arms race, which in turn is increasing demand for digital and power infrastructure building blocks. In that context, the bank assembled a list of core "pick and shovel" suppliers concentrated on networking and interconnect solutions that support hyperscale AI clusters.
Companies highlighted by Barclays
- Broadcom - Identified for its Ethernet switching silicon, custom AI accelerators, and fiber optic components used in large AI clusters.
- NVIDIA - Cited for its Spectrum-X Ethernet platform, InfiniBand, and NVLink interconnect technologies designed for efficient AI cluster operation.
- Cisco Systems - Noted for its Silicon One switching platforms and an optics portfolio that supports 800G LPO and ZR/ZR+ modules.
- Corning - Spotlighted as a producer of high-density optical fiber, cable, and connectivity solutions tailored for AI-enabled data centers.
- Arista Networks - Recognized for high-speed Ethernet switches and support for pluggable optics used in scale-out and scale-up AI networks.
- Amphenol - Listed as a supplier of high-speed connectors, power solutions, and fiber optic interconnects used in AI and machine learning systems.
- TE Connectivity - Noted for providing 800G/1.6T electrical and optical connectivity solutions, including backplane and high-speed data cables.
- Lumentum - Identified for high-performance indium phosphide lasers and optical transceivers for hyperscale AI clusters.
- Coherent Corp - Included for its high-performance optical transceivers, optical circuit switches, and other optical products aimed at hyperscale AI data centers.
- Celestica - Cited for designing and manufacturing 800G and 1.6T open networking switches for hyperscale AI facilities.
Barclays also mentioned a number of additional U.S.-listed companies in this segment, specifically naming Astera Labs, Marvell Technology, MACOM, Credo Technology, Fabrinet, Keysight Technologies, and Clearfield as further players within the networking and interconnect ecosystem.
The bank's note frames these firms as critical infrastructure suppliers that enable the scale and performance required by large AI clusters, emphasizing components ranging from switching silicon and interconnect fabrics to optics and high-speed connectors. Barclays' estimate that annual AI infrastructure spending from Western hyperscalers and AI labs could top $1 trillion is presented as a projection that exceeds consensus by a significant margin, with the timeline pointing to a peak in 2028.
Investors and market participants tracking the data center, networking, and optical components sectors may view the Barclays list as a lens on where demand pressure could concentrate as AI-related capital expenditure grows.