Kandou AI, a company focused on high-speed, energy-efficient connectivity and system solutions for artificial intelligence infrastructure, announced the closing of a $225 million strategic funding round. The oversubscribed Series A was led by Maverick Silicon, with strategic participation from SoftBank Group Corp., Synopsys, Inc., Cadence Design Systems, Inc., and Alchip Technologies Limited. Several existing investors also joined the round.
The funding comes as AI models continue to expand rapidly, creating challenges in connecting GPUs and CPUs to growing memory requirements. According to the company, performance in AI clusters is increasingly limited not by compute power but by the bandwidth and efficiency of data interconnects between compute and memory. Traditional networking and interconnect technologies, originally developed for pre-AI applications, are struggling to scale with these demands.
Kandou AI states that its silicon-proven interconnect technologies are designed specifically to address these bottlenecks in data movement.
The capital will support three main priorities: ramping up manufacturing of its high-performance AI connectivity chips, deepening engagement with hyperscale and AI infrastructure customers worldwide, and advancing its chip and IP roadmap toward multi-terabit connectivity.
Srujan Linga, Co-Founder and CEO of Kandou AI, said the round validates the company's vision and will accelerate efforts to transform AI hardware using copper-based interconnect and system IP. Initial products target AI systems and high-speed rack connectivity, aiming to exceed the 448G barrier, with future expansion into broader AI infrastructure. The company plans to collaborate with hyperscale and enterprise customers on next-generation systems.
Kandou AI's approach combines information theory and semiconductor design through patented signaling and SerDes technologies. These reimagine data transmission over copper interconnects, targeting improvements in bandwidth, reach, and energy efficiency. The company indicates this could allow scalable AI systems at lower cost and power levels.
Manish Muthal, Senior Managing Director at Maverick Silicon, noted that next-generation AI cluster performance depends on efficient data movement between chips, an area where Kandou AI has focused its efforts. He expressed support for the company's work on scalable, energy-efficient AI systems.
Neeraj Paliwal, senior vice president of Silicon IP product management at Synopsys, highlighted that AI performance is increasingly constrained by data movement between compute and memory. He said Kandou AI's advanced coding techniques for high-speed copper interconnects align with industry goals for multi-terabit connectivity, and integration with Synopsys' pre-verified IP is intended to accelerate time to market.
Nimish Modi, senior vice president and general manager of Strategy and New Ventures at Cadence, pointed to inter-chip data movement as a key bottleneck in scalable AI infrastructure amid data-intensive workloads. He stated that Cadence is supporting Kandou AI through its EDA solutions to develop next-generation connectivity technologies.
The announcement follows recent company milestones, including the successful tapeout of its next-generation SerDes technology, momentum in its retimer business, the opening of a new design center in Hyderabad, India, and the addition of industry veterans to its executive team.
The round values Kandou AI at $400 million.
