Coherent launched PhotonLink on September 21, 2026, bringing lasers, silicon photonics, precision optics, fiber assemblies and detectors into an integrated platform aimed at next-generation AI data centers. The company says the platform supports co-packaged optics, near-packaged optics and emerging chip-to-chip optical links as operators push for more bandwidth with lower power use. Coherent’s launch announcement describes the platform as an end-to-end optical signal-chain offering backed by expanded indium-phosphide and specialty-fiber manufacturing.

Why AI Networks Need More Optics

Large AI clusters move enormous amounts of data between accelerators and switches. Traditional pluggable optical modules remain important, but long high-speed electrical paths consume power and become harder to manage as link speeds rise. Moving optical conversion closer to switching and compute silicon shortens those electrical paths. Coherent says PhotonLink is designed around that shift, covering light generation, beam shaping, transmission, detection and conversion back to an electrical signal near an xPU or switch chip. PhotonLink’s technical overview lists InP continuous-wave lasers, silicon-photonics modulators, VCSEL arrays, fiber-attach units and InP and silicon-photonics detectors among the supported technologies.

CPO and NPO Move the Optical Engine Closer

Co-packaged optics places optical engines alongside switching silicon inside the package environment, while near-packaged optics keeps the optical hardware close to the package. Both approaches reduce the distance that extremely fast electrical signals must travel before becoming light. The engineering goal is not simply faster fiber. It is a more efficient connection between compute, switching and the optical network. Coherent’s ECOC 2026 demonstrations include a 6.4-terabit-per-second NPO optical engine and 3.2-terabit-per-second transceiver technology, showing how quickly optical bandwidth density is increasing. Coherent’s ECOC technical announcement also highlights optical circuit switching, advanced materials and thermal-management work.

The Power Budget Matters

Networking power increasingly competes with accelerators, memory and cooling for a data center’s electrical budget. Every watt removed from the interconnect can create more room for useful compute or reduce heat that cooling systems must carry away. NVIDIA has also been developing co-packaged optical networking for AI systems, illustrating how the industry is treating optical integration as part of the system architecture rather than as a separate cable problem. The embedded NVIDIA presentation explains the basic CPO approach and why shorter electrical paths can reduce networking power.

Manufacturing Becomes Part of the Story

PhotonLink also shows why AI infrastructure is becoming a manufacturing problem as much as a chip-design problem. Lasers, silicon photonics, lenses, fibers and detectors have to be assembled and tested with tight tolerances. Coherent says it is expanding six-inch indium-phosphide and specialty-fiber manufacturing capacity to support the platform’s ramp. That makes precision assembly, optical test and packaging capacity important pieces of the AI hardware supply chain, not secondary components. HPCwire’s September 22 coverage confirms the platform’s focus on complete optical solutions for AI infrastructure.

What to Watch Next

The next engineering test is adoption at scale. CPO and NPO must deliver high bandwidth while meeting reliability, thermal, serviceability and manufacturing-yield requirements. PhotonLink gives Coherent a broad platform spanning several parts of that stack. AI clusters are making optical networking a core compute-infrastructure technology, and the companies that can manufacture the complete optical path reliably may become increasingly important as accelerator deployments grow.

BitcoinVersus.Tech Editor’s Note:
We volunteer daily to ensure the credibility of the information on this platform is Verifiably True. If you would like to support to help further secure the integrity of our research initiatives, please donate here: 3C9o19EH5HSiwEPyCTmEKzxhNCbo2X6TTb