Kandou AI

Kandou AI, founded in 2011 by the accomplished mathematician Dr. Amin Shokrollahi, operates as a fabless semiconductor company from its headquarters in Lausanne, Switzerland. The company's inception traces back to Dr. Shokrollahi's extensive two-decade career at the forefront of information communication research, which includes inventing Raptor codes and serving as Chief Scientist at Digital Fountain (acquired by Qualcomm). His academic and entrepreneurial journey, moving from mathematics to real-world applications, is central to Kandou's foundation, which was also established with co-founder Harm Cronie.

The firm specializes in high-speed, energy-efficient chip-to-chip link solutions, aiming to redefine wired connectivity. It operates on a dual business model, generating revenue through the licensing of its intellectual property (IP) and the sale of standard semiconductor products. Kandou serves a broad market that includes clients in data centers, hyperscale computing, consumer electronics, and 5G infrastructure. The company has successfully raised a total of $326 million over nine funding rounds, with its latest Series E funding in August 2023 securing $72.3 million to fuel market expansion and product development.

Kandou’s core technological offering is its proprietary Chord™ signaling architecture, a multi-wire signaling method that generalizes traditional differential signaling. This technology allows for the transmission of more bits over a correlated set of wires (e.g., five bits on six wires), significantly boosting bandwidth and pin efficiency while reducing power consumption by up to 50% compared to conventional technologies like PAM4 and NRZ. This has led to the development of several product families, including the 'Matterhorn' series of USB4® retimers, the 'Regli' family of retimers for PCIe 5.0 and CXL 2.0, and the 'Glasswing' PHY IP for ultra-short reach applications. Recently, the company has sharpened its focus on the artificial intelligence sector, rebranding to Kandou AI and leveraging its copper MIMO interconnect technology to create chiplet-based AI fabric solutions. These solutions are designed to make AI systems more scalable and affordable by enabling low-cost GPU usage, creating network-attached memory pools for AI training, and developing massive switchless memory systems with CXL controllers.

Keywords: chip-to-chip connectivity, fabless semiconductor, Chord signaling, high-speed interconnects, energy-efficient semiconductors, retimers, SerDes, intellectual property licensing, signal conditioning, USB4, PCIe 5.0, CXL 2.0, AI hardware systems, memory fabric, chiplets, data center solutions, Amin Shokrollahi, multiprotocol retimer, Copper MIMO, interconnect technology