Silicon Valley startup Xscape Photonics has launched the EagleX Laser Evaluation Kit, offering early access to its ChromX multi-color laser platform designed to overcome bandwidth bottlenecks in next-generation AI data centers. The kit features the industry’s first wafer-scale silicon photonics (SiPho)-based multi-color laser source capable of emitting up to 16 wavelengths.
Xscape’s technology targets the growing scalability challenge as AI workloads drive skyrocketing demand for data center capacity. Traditional data center optics typically support only four wavelengths per fiber—insufficient for AI-driven architectures. The EagleX Kit supports 8 and 16 wavelengths in the O-band, compliant with CW-WDM MSA, and is powered by a single off-the-shelf DFB pump laser, enabling hyperscalers to dramatically increase bandwidth without excessive power consumption or cost.
The EagleX kit marks the first commercial release of the ChromX platform since Xscape closed a $44 million Series A in late 2024. The platform is aimed at next-generation fabric connectivity for AI clusters, with applications in AI interconnects, optical computing, co-packaged optics (CPO), and near-packaged optics (NPO). Founded in 2022 by SiPho experts from Columbia University and veterans of Broadcom, Intel, and others, Xscape aims to reshape optical scaling for AI hardware over the next decade.
- Xscape launches EagleX Kit for testing 8- and 16-color multi-wavelength laser source
- SiPho-based ChromX platform targets AI data center interconnect bottlenecks
- Enables hyperscalers to expand bandwidth while reducing power and cost
- Applications include AI interconnects, CPO, NPO, and optical computing
- Live demos at DAC 2025, San Francisco, June 22-25 (Booth #2308D)
“Today’s AI data centers are simply not efficient enough. The bottlenecks created by existing networking infrastructure only allow users to see a fraction of GPU performance,” said Vivek Raghunathan, Co-Founder and CEO of Xscape Photonics. “This platform provides a new vector of bandwidth scaling to power AI hardware roadmaps for the next decade.”
