Chip startup Retym announced on Monday that it has secured $75 million this year, contributing to $180 million in funding to develop networking chips for artificial intelligence computing in data centers.
Retym specializes in chips designed for digital signal processing (DSP), facilitating rapid data transfer between large data centers. This capability has become increasingly crucial with the growing demand for AI technologies. Training advanced AI models, such as those powering ChatGPT, requires thousands of interconnected chips supported by networking infrastructure.
Currently, Marvell Technology dominates the DSP chip market. However, according to CEO Sachin Gandhi, Retym—pronounced “re-time”—aims to address a significant bottleneck in data centers. As AI computing relies on multiple chips working in unison, efficient connectivity between them is becoming a critical factor in optimizing performance.
“We are focusing on building coherent DSP chips for the next generation deployment of AI infrastructure and cloud,” Gandhi said.
The chip startup, Retym’s first chip will be engineered to transfer data across distances ranging from 10 to 120 kilometers, with an optimal performance range of 30 to 40 kilometers. To maintain data integrity during transmission, the DSP chip utilizes a specialized modulation technique that prevents corruption.
“They took the approach of solving the harder problem with longer distances,” said Navin Chaddha, managing partner at Mayfield, a venture fund that has invested in Retym.
Retym is utilizing Taiwan Semiconductor Manufacturing Co.’s five-nanometer process for its first chip, which its engineers are currently testing and validating.
Spark Capital led the company’s Series D funding round. Founded four years ago, Retym aims to launch its first product this year.
