Lightmatter is taking a major step toward solving one of the biggest bottlenecks in modern AI infrastructure: bandwidth. The company has announced parallel collaborations with Global Unichip Corp. (GUC) and Cadence, aimed at speeding up the real-world rollout of Co-Packaged Optics (CPO) inside data centers.
As AI models grow larger and more demanding, data centers are running into limits that traditional electrical interconnects struggle to overcome. Moving data between chips and across servers is increasingly expensive in power, heat, and complexity. That’s where co-packaged optics comes in. CPO brings optical connectivity much closer to the compute, helping deliver higher bandwidth and improved efficiency compared to conventional approaches that rely heavily on copper connections and pluggable optical modules.
Lightmatter’s role centers on its photonic technology, designed to enable faster, more efficient data movement using light instead of electricity. By pairing that photonics expertise with GUC’s abilities in ASIC design and implementation, the collaboration targets more practical and scalable pathways to integrate optical interconnects alongside high-performance custom silicon. The goal is to make CPO architectures easier to deploy in the types of chips and platforms hyperscalers and enterprise data centers already depend on.
In parallel, Lightmatter’s collaboration with Cadence focuses on the design and engineering side needed to bring CPO to market faster. Cadence is widely used for electronic design automation workflows, and aligning photonic integration with mature chip design flows can reduce friction for teams trying to adopt co-packaged optics. This matters because CPO isn’t just a single component upgrade; it’s a system-level shift that requires careful coordination across packaging, signal integrity, power, thermal performance, and manufacturing readiness.
Together, these collaborations point to a clear message: the industry is moving aggressively toward co-packaged optics as AI networking demands accelerate. By combining photonics innovation, ASIC development expertise, and established design tooling, Lightmatter, GUC, and Cadence are aiming to break through bandwidth constraints and make next-generation data center connectivity more achievable.
For data center operators and AI infrastructure builders, this kind of progress could translate into higher throughput, better energy efficiency, and a stronger foundation for scaling future AI workloads—without hitting the same interconnect limits that are increasingly slowing today’s systems.
