Xscape Photonics is a silicon photonics company spun out of Columbia University that is rethinking how data moves through AI data centers using light. As AI clusters grow, the electrical interconnects that traditionally shuttle data between chips and systems are hitting hard limits on bandwidth and energy efficiency. Optical interconnects — using photons instead of electrons — offer a path forward, and Xscape's bet is on multi-wavelength optics that pack far more data onto each fiber by transmitting many colors of light simultaneously.
At the heart of Xscape's technology is its proprietary CombX laser, which generates multiple distinct wavelengths of light on a single silicon photonics chip. By using many colors in parallel rather than a single wavelength, Xscape can multiply the bandwidth carried over each optical link without proportionally increasing the number of physical components. This wavelength-division approach is especially valuable for the dense, bandwidth-hungry networks that connect GPUs and accelerators in AI training and inference clusters.
The company has translated this technology into products including FalconX, an eight-wavelength laser device unveiled to transform how data moves within AI data center networks. By integrating high-performance, multi-color optical interconnects directly into AI networking, Xscape aims to relieve the interconnect bottleneck that increasingly constrains AI scaling, while improving energy efficiency — a growing concern as data center power consumption surges.
Xscape Photonics has attracted strategic and financial backing from across the technology and semiconductor industry. Its 2024 Series A round of $44 million was led by IAG Capital Partners with participation from Cisco Investments, NVIDIA, Altair, Fathom Fund, Kyra Ventures, LifeX Ventures and OUP. The company subsequently added $37 million to the round — announced in 2026, led by Addition with continued support from IAG Capital Partners and NVIDIA — bringing total Series A funding to roughly $81 million.
The involvement of Nvidia and Cisco, two of the most important players in AI compute and networking respectively, is a strong endorsement of Xscape's approach. As hyperscalers race to build ever-larger AI clusters, multi-wavelength silicon photonics like Xscape's could become an essential building block for keeping data flowing at the speed and efficiency AI demands.