
Chip Startup Retym Raises $75 Million to Transform AI Data Center Interconnectivity

Retym, a rapidly expanding semiconductor startup, has raised $75 million in its most recent funding round, taking its total funding to date to $180 million. The firm is using the investment to create sophisticated networking chips tailored for artificial intelligence (AI) computing within data centers, Retym said on Monday.
The proliferation of AI use cases has established a growing need for rapid data movement across and within large data centers. Retym is focused on building digital signal processing (DSP) chips that maximize data transmission efficiency, an important need for AI workloads. The company will seek to overcome connectivity bottlenecks by optimizing data movement between thousands of chips that are connected to one another to train and deploy AI models such as ChatGPT.
Presently, the market for DSP chips is controlled by market leader Marvell Technology. Yet, Retym is looking to become a strong contender by centering on an innovative solution for addressing data center networking issues. The CEO of the company, Sachin Gandhi, underlined the significance of smooth connectivity in massive AI computing, where multiple chips have to function in collaboration effectively.
"We are committed to developing high-performance DSP chips that will empower next gen AI infrastructure and cloud computing implementations," Gandhi said. "Our technology will enable quicker, more accurate data movement over large distances, removing existing bottlenecks in AI data centers."
Retym's initial DSP chip is designed to carry data over distances of 10 kilometers to 120 kilometers, with the optimization target being 30 to 40 kilometers. The chip uses a sophisticated modulation method to ensure data integrity and avoid signal corruption over long distances. This technology makes Retym's solution especially well-suited for mass-scale AI deployments.
"They set out to solve one of the most difficult problems providing high speed, long haul data connectivity," said Navin Chaddha, managing partner at Mayfield, a venture capital company that has invested in Retym. "This approach differentiates them in the competitive semiconductor industry."
Retym is using Taiwan Semiconductor Manufacturing Company's (TSMC) state-of-the-art five-nanometer process to produce its first chip. The company's engineers are now testing and validating the chip, optimizing its performance before its release to the market.
The Series D financing round was spearheaded by Spark Capital, a testament to investors' faith in Retym's vision and technological competencies. The four year old startup has come a long way since its inception and continues to be on course to bring its first product to market this year.
As the adoption of AI keeps increasing, Retym's revolutionary DSP chips can potentially become a key player in transforming the data center networking infrastructure, improving efficiency, and opening up new avenues in AI-based computing.
The proliferation of AI use cases has established a growing need for rapid data movement across and within large data centers. Retym is focused on building digital signal processing (DSP) chips that maximize data transmission efficiency, an important need for AI workloads. The company will seek to overcome connectivity bottlenecks by optimizing data movement between thousands of chips that are connected to one another to train and deploy AI models such as ChatGPT.
Presently, the market for DSP chips is controlled by market leader Marvell Technology. Yet, Retym is looking to become a strong contender by centering on an innovative solution for addressing data center networking issues. The CEO of the company, Sachin Gandhi, underlined the significance of smooth connectivity in massive AI computing, where multiple chips have to function in collaboration effectively.
"We are committed to developing high-performance DSP chips that will empower next gen AI infrastructure and cloud computing implementations," Gandhi said. "Our technology will enable quicker, more accurate data movement over large distances, removing existing bottlenecks in AI data centers."
Retym's initial DSP chip is designed to carry data over distances of 10 kilometers to 120 kilometers, with the optimization target being 30 to 40 kilometers. The chip uses a sophisticated modulation method to ensure data integrity and avoid signal corruption over long distances. This technology makes Retym's solution especially well-suited for mass-scale AI deployments.
"They set out to solve one of the most difficult problems providing high speed, long haul data connectivity," said Navin Chaddha, managing partner at Mayfield, a venture capital company that has invested in Retym. "This approach differentiates them in the competitive semiconductor industry."
Retym is using Taiwan Semiconductor Manufacturing Company's (TSMC) state-of-the-art five-nanometer process to produce its first chip. The company's engineers are now testing and validating the chip, optimizing its performance before its release to the market.
The Series D financing round was spearheaded by Spark Capital, a testament to investors' faith in Retym's vision and technological competencies. The four year old startup has come a long way since its inception and continues to be on course to bring its first product to market this year.
As the adoption of AI keeps increasing, Retym's revolutionary DSP chips can potentially become a key player in transforming the data center networking infrastructure, improving efficiency, and opening up new avenues in AI-based computing.