AI clusters aren’t just servers. They’re interconnected systems where GPUs communicate at enormous bandwidth over high-speed optical links, and where the memory bandwidth between chips determines model performance.
Two technology categories sit at the heart of this architecture: optical networking and advanced semiconductor packaging (particularly HBM — High Bandwidth Memory). Both are less visible to mainstream investors, and both offer compelling exposure to AI infrastructure buildout.
Optical Networking
Lumentum Holdings (LITE)
Lumentum makes lasers, optical components, and photonic products used in high-speed networking and telecommunications. Their coherent optical products are used in the high-bandwidth links that connect AI data centers and clusters.
The AI angle: As AI deployments scale, the networking bandwidth requirements between clusters and data centers grow dramatically. Optical interconnects are the only technology capable of delivering that bandwidth economically. Lumentum is a direct beneficiary of increased optical network deployment driven by AI traffic.
What to watch: Datacom vs. telecom revenue mix, coherent transceiver order growth.
Coherent Corp (COHR)
Coherent (formerly II-VI Incorporated) manufactures optical components, transceivers, and laser systems. They compete with Lumentum in several markets and have significant exposure to both AI networking and industrial applications.
The AI angle: Coherent has been calling out AI-driven networking demand as a growth catalyst. Their 800G and 1.6T optical transceiver products are designed for the high-speed connections that AI infrastructure requires.
What to watch: AI/datacom transceiver revenue growth and design win announcements.
Advanced Packaging and HBM Ecosystem
BE Semiconductor Industries (BESI)
BE Semiconductor (also known as Besi) is a Netherlands-based company that manufactures advanced semiconductor packaging equipment — specifically bonding systems used in chip assembly. Their hybrid bonding technology is expected to become critical for next-generation HBM (High Bandwidth Memory) production.
The AI angle: HBM is the high-bandwidth memory stacked directly on AI accelerators (like Nvidia’s H100 and H200). Producing HBM requires advanced packaging technology that Besi specializes in. As AI drives demand for more and faster HBM, Besi’s equipment becomes more essential.
The nuance: This is a semi-equipment play, not a direct chip play. Semi-equipment companies benefit from capital investment cycles, which can be lumpy. Timing matters more here than in some other infrastructure categories.
Camtek (CAMT)
Camtek makes inspection and metrology equipment for semiconductor manufacturing — including advanced packaging processes used in HBM production. It’s a smaller, less-followed name in the AI infrastructure ecosystem.
The AI angle: Quality inspection is a critical step in advanced packaging. As HBM production volumes increase to meet AI demand, Camtek’s inspection equipment sees higher utilization and potential for new orders.
What to watch: Advanced packaging revenue as a percentage of total revenue, HBM production announcements from SK Hynix, Samsung, and Micron.
Key Risk: Efficiency Improvements
Both optical networking and memory packaging face a structural risk: if AI models become dramatically more efficient (requiring less compute and memory per inference), demand growth could slow. This has been an ongoing trend — models are getting more efficient rapidly. The question is whether enterprise adoption grows fast enough to offset efficiency gains. Most evidence suggests it does, but it’s a risk to monitor.
This is not financial advice. International stocks (BESI is Dutch-listed; COHR and LITE trade on US exchanges) carry additional currency and regulatory risks.
Ready to Get Started?
Invest in AI infrastructure stocks commission-free at Webull.