Chip startup Retym has shaken the semiconductor industry with a bold move, raising $75 million in 2025 alone as part of its total $180 million funding pool. The company plans to disrupt the networking chip landscape for artificial intelligence (AI) computing in large-scale data centers. With a focus on digital signal processing (DSP), Retym aims to solve one of the biggest bottlenecks in modern AI infrastructure — the fast and reliable transmission of data between chips spread across long distances.
The Growing Challenge of AI Networking
As AI models like ChatGPT, DALL·E, and others continue to grow in complexity, they demand massive computational resources. These models run across thousands of interconnected chips within sprawling data centers. Moving enormous amounts of data quickly and accurately between these chips becomes a critical challenge. AI tasks often involve splitting computations among multiple processing units, and the efficiency of those tasks depends heavily on how fast and reliably the system can share data.
Currently, Marvell Technology dominates the DSP chip market, especially in AI and cloud data centers. However, Retym’s entrance introduces fresh competition, innovation, and a clear ambition to lead in a sector essential to the future of AI.
Retym’s Mission: Redefining Long-Distance Data Transmission
Retym, pronounced “re-time,” focuses on developing coherent DSP chips. These chips serve a key role in maintaining data integrity while transmitting signals across considerable distances. CEO Sachin Gandhi emphasized the urgency of solving this communication bottleneck:
“We are focusing on building coherent DSP chips for the next generation deployment of AI infrastructure and cloud.”
Instead of going after low-hanging fruit, Retym engineers chose the more complex challenge — optimizing chips for medium to long-distance data transfer. Their first chip will move data efficiently across distances ranging from 10 kilometers to 120 kilometers, with a sweet spot between 30 km and 40 km.
These specifications directly target real-world needs in metro-scale data centers where servers span multiple buildings or campuses. By refining DSP chips to work flawlessly across these ranges, Retym addresses a crucial gap in AI infrastructure — the inability of current chips to maintain both speed and signal integrity over longer distances.
Technical Innovation at the Core
Retym’s DSP chip doesn’t rely on basic signal transfer. It utilizes an advanced modulation technique that ensures data integrity during long-distance transmission. When bits of information travel across optical fibers, they often face distortion, noise, and loss. Retym’s solution minimizes these issues by embedding intelligence into the chip’s signal modulation and demodulation process.
Navin Chaddha, managing partner at Mayfield (one of Retym’s investors), praised the company’s approach:
“They took the approach of solving the harder problem with longer distances.”
Instead of retrofitting existing designs, Retym started from scratch. Their engineering team opted for a ground-up design, optimizing every detail to meet the specific demands of AI data center communication. That decision may cost more in time and resources, but it promises better performance, scalability, and future-readiness.
Backing from Top Investors and Industry Giants
Retym’s ability to raise $180 million since its founding four years ago proves investor confidence. The latest Series D round, led by Spark Capital, brings in a fresh injection of $75 million in 2025. With these funds, Retym accelerates development, testing, and commercialization.
Other key backers, including Mayfield and several unnamed strategic investors, provide not only financial muscle but also deep industry expertise. These partners bring guidance in areas such as semiconductor supply chains, cloud infrastructure, and market expansion — all essential for a young chip startup aiming for dominance.
Leveraging Advanced Semiconductor Fabrication
To ensure top-tier performance, Retym selected Taiwan Semiconductor Manufacturing Co. (TSMC) as its fabrication partner. TSMC’s 5-nanometer manufacturing process enables Retym to pack more performance and efficiency into a smaller chip footprint. That decision enhances power efficiency — a major concern in AI computing — and ensures better heat management, faster speeds, and compact design.
Retym’s engineers are currently testing and validating prototype samples, and they remain on track to bring their first product to market in 2025. This timeline puts Retym in a strong position to capture early demand from AI-focused cloud providers and hyperscale data centers.
Disrupting an Entrenched Market
The networking chip segment has seen limited innovation over the past few years, especially in the DSP category. With Marvell currently controlling much of the space, data centers face limited options for high-performance, low-latency communication hardware tailored for AI needs. Retym challenges this status quo by introducing fresh technology specifically designed for AI and cloud-native architectures.
The startup’s choice to specialize in coherent optics — an area often associated with telecom networks — and adapt it for AI data centers marks a unique and forward-looking strategy. Instead of relying solely on electrical transmission, Retym embraces optical networking powered by advanced DSP to bridge the increasing physical gaps between chips in large facilities.
Looking Ahead: Growth, Expansion, and AI Acceleration
Retym’s roadmap extends beyond a single chip. The team already has plans for future generations that will support higher data rates, longer distances, and smarter signal processing capabilities. These improvements will enable Retym to support the growing wave of AI workloads expected over the next decade.
As large language models, generative AI, autonomous systems, and enterprise AI continue to expand, data centers will need infrastructure that can grow with them. Retym wants to become the backbone of that infrastructure, handling the invisible but crucial job of making sure all those chips communicate flawlessly.
Final Thoughts
Retym isn’t just building another chip — the company is laying the foundation for the next era of AI computing. By tackling one of the most complex and essential problems — high-speed data movement over long distances — the startup positions itself as a key enabler of AI at scale.
With robust financial backing, cutting-edge technology, and a clear vision, Retym stands poised to disrupt a market ripe for innovation. As the world races to build smarter machines and more powerful AI models, the need for high-performance networking chips will only grow — and Retym plans to meet that demand head-on.
Read More: European Startups Focused on Ethical AI
Also Read: Startup Mahakumbh 2025: Why It’s a Must-Attend