Nubis Communications has introduced its first product, the XT1600 optical engine, designed to be thoroughly optimized for machine learning and artificial intelligence (ML/AI) networks. Founded in 2020, the company has since raised over $40 million to fully fund the development and volume production of its first product.
Samples of the XT1600 optical engine are now available to customers. It has a total capacity of 1.6Tb/s optical transport and 1.6Tb/s optical receive carried on 16 fiber pairs, allowing full fan-out and flexible allocation as 16 independent 100Gb/s full-duplex channels, four independent 400Gb/s full-duplex channels - duplex channels, or two 800Gb/s full-duplex channels. The XT1600 engine is optimized for unretimed direct-drive operation, connecting directly to the host ASIC SerDes designed for copper links. This direct drive approach significantly reduces the power, size and cost of the optical solution, which may be critical for the continued growth of ML/AI clusters in power-constrained data centers.
Our customers have been telling us that the power and size of traditional optics is unsustainable," said Dan Harding, CEO of Nubis Communications. They are building the world's largest and highest performing data interconnect clusters and rely on us to provide higher density, lower power, and more efficient optics to support the expansion of their networks."
Founder Peter Winzer added, "Our solution is optimized to work with modern SerDes, matching not only their spacing to their edge I/O density, but also their energy-efficient direct drive capabilities. The result is that AI gas pedals or similar large ASICs can achieve full bandwidth connectivity within the data center at a fraction of the power compared to traditional optical solutions. In addition, our approach is well suited for emerging box architectures as well as native small-chip implementations to enable tighter integration in the future."
