Key Takeaways

  • Coherent recorded a 40.6% year-over-year jump in its Datacenter and Communications revenue, highlighting rapid hyperscale demand for optical connectivity.
  • POET Technologies, Lumentum, and Credo Technology are advancing supply chains tied to the 800G and 1.6T transitions.
  • Research from IEEE and Cisco indicates AI clusters are pushing operators toward optical links to overcome electrical bandwidth and power constraints.

AI data center operators are moving into a phase where every incremental GPU requires a proportionally larger slice of network capacity, and copper interconnects cannot scale fast enough to meet these demands. Coherent recently reported its Datacenter and Communications segment reached $1.362 billion in revenue, up 40.6% from the prior year. This segment now represents 75% of the company's total revenue, shifting from 41% pro forma a year earlier. Networking researchers have tracked this structural pivot toward optics, driven by the expanding power and bandwidth limits of high-density AI clusters.

According to IEEE research, hyperscale fabrics upgrading to 800G and 1.6T Ethernet frequently encounter electrical link distance limitations before reaching bandwidth ceilings, accelerating the transition to optical transceivers. Cisco has similarly mapped this shift in its high-speed Ethernet roadmaps, noting that as AI clusters expand in node count and radix, optical link budgets become a deciding architectural factor. These technical realities align with recent analyses of optical component shipments and AI data center expansion from Omdia and LightCounting.

Several specialized networking and photonics vendors are actively positioning for this infrastructure upgrade cycle. POET Technologies focuses on silicon photonics integration, offering optical engines that help operators manage thermal and power efficiency as data center footprints scale. Silicon photonics is increasingly central to the supply chain because it enables the integration of optical functions directly at the chip level, supporting the broader shift toward co-packaged optics.

Other key vendors addressing the high-speed connectivity gap include Credo Technology, Lumentum, and MACOM. Credo Technology supplies high-speed connectivity solutions required for dense scale-out networks, while Lumentum and MACOM provide the specialized lasers and optical components necessary to drive these connections over longer distances with less signal loss. By relying on optics rather than copper, operators can expand GPU clusters while managing the strict energy constraints imposed by massive AI training workloads.

Coherent remains a primary supplier of merchant optical transceivers for hyperscale infrastructure. The company's 40.6% segment growth illustrates the immediate financial impact of AI buildouts on networking hardware. Because 800G optical transceivers are currently the mainstream upgrade path for hyperscalers, suppliers capable of manufacturing these components at volume are capturing early capital expenditures associated with next-generation Ethernet deployments.

Research groups such as Dell’Oro Group have consistently linked AI infrastructure spending directly to parallel investments in networking capacity. As operators deploy larger GPU clusters, bandwidth requirements scale aggressively. Copper interconnects suffer from increasing electrical constraints over the longer physical distances required in large AI fabrics, forcing operators to adopt optical links to maintain throughput without exceeding power limits.

The optical networking market is currently transitioning to 1.6T interconnects, with industry roadmaps pointing to broader deployment between 2025 and 2027. Operators preparing for denser AI clusters are prioritizing optics to alleviate tightening power budgets and manage data traffic across expanding physical footprints. As hyperscale demand remains concentrated in these large-scale buildouts, the integration of advanced photonics will dictate how efficiently next-generation data centers can operate.