Site icon Converge Digest

Marvell and Mojo Vision Partner on Micro-LED Optical Interconnects

Marvell Technology and Mojo Vision announced a long-term collaboration to develop micro-LED-based optical interconnects aimed at next-generation AI data center infrastructure. The effort targets short-reach connectivity between accelerators, memory pools, and other components in large-scale AI clusters, where system architects are seeking higher bandwidth density, lower power consumption, and reduced latency.

The joint development program combines Mojo Vision’s micro-LED platform with Marvell’s connectivity technologies, including high-speed SerDes and integrated electrical circuitry. According to the companies, the architecture integrates micro-LED emitters, photodetectors, multicore fiber bundles, and software-defined alignment to move terabits of data per millimeter at very low power and latency over short distances. The companies said development of the technology has been underway for more than a year and will support multiple form factors for hyperscale and cloud data center deployments.

Marvell said it also served as the largest investor in Mojo Vision’s 2025 Series B Prime financing as part of a multi-generational agreement to co-develop the technology. Micro-LED optical interconnects are emerging as one approach to address the growing I/O bottleneck inside AI systems, where traditional pluggable optics and copper links face constraints as AI clusters scale to thousands of XPUs.

“Micro-LED technology represents an important technology capability for high-performance data center connectivity applications,” said Xi Wang, senior vice president and general manager of the Connectivity Business Unit at Marvell.

🌐 Analysis: Micro-LED optical I/O is gaining attention as AI infrastructure pushes beyond the limits of traditional copper interconnects and conventional pluggable optics. Companies including Ayar Labs, Lightmatter, and others are exploring alternative optical architectures—such as co-packaged optics, photonic chiplets, and massively parallel optical links—to address bandwidth density and energy efficiency inside large-scale AI clusters.

Exit mobile version