Chip startup Retym has secured $75 million in funding as part of a larger $180 million investment to develop networking chips optimized for artificial intelligence (AI) computing in data centers. The company, which specializes in digital signal processing (DSP) chips, aims to address the growing demand for faster data transfer across large data centers—a critical component in AI advancements. With the increasing reliance on AI models like ChatGPT, data centers are facing bottlenecks in connectivity, which Retym plans to resolve.
Retym’s DSP chips are designed to enable quick and reliable information transfer over long distances, a necessity as thousands of chips work together to power AI models. The company’s upcoming chip will focus on optimizing data flow over distances between 30 and 40 kilometers, though it will be capable of handling ranges from 10 to 120 kilometers. Retym’s innovation lies in its use of a modulation technique that ensures data integrity during transportation, an essential feature for AI computing tasks.
Currently, the DSP chip market is dominated by companies like Marvell Technology, but Retym aims to carve out its own niche by solving a critical challenge in AI infrastructure. The need for high-speed, reliable connectivity between chips has grown as AI models become more complex. Retym’s focus on long-distance data transfer sets it apart from other chipmakers in the space, with its CEO, Sachin Gandhi, emphasizing the importance of coherent DSP chips in next-generation AI deployments.
The $75 million funding round, part of a broader Series D investment led by Spark Capital, will help Retym bring its first product to market. The startup, founded four years ago, is currently testing its chip, which will be manufactured using a five-nanometer process. As AI technology continues to evolve, Retym’s innovative approach could play a key role in the future of cloud and AI infrastructure.
Source: Reuters