Kandou AI Targets Data Center Bottlenecks With New Chips

The funding highlights growing demand for faster and more cost-efficient interconnect technologies to address data bottlenecks as AI infrastructure scales.

Kandou AI has raised $225 million in a funding round backed by investors including Maverick Silicon, SoftBank Group, Synopsys, Cadence Design Systems and Alchip. The round values the company at about $400 million, Chief Executive Srujan Linga said. It is part of a growing group of startups working to solve data transfer bottlenecks in AI systems, an area gaining attention as computing demands surge.

The company develops interconnect technologies that improve how data moves between chips in artificial intelligence systems. As AI workloads grow more complex, the speed of data transfer between graphics processing units has become a key bottleneck in large-scale data centers. These systems rely on multiple GPUs working together to process large volumes of data, and slow connections between chips can limit overall performance and increase infrastructure costs.

Kandou AI is focusing on copper-based interconnects as a lower-cost alternative to fiber-optic solutions, which are widely used but more expensive in large-scale AI deployments. The company said its technology could improve communication between GPUs, helping reduce the cost of building and operating such systems.

Copper-based interconnects are gaining attention as companies seek more efficient data transfer in AI systems. While fiber-optic connections are widely used for long-distance transmission, they are more expensive and power-intensive in data centers. Kandou’s approach aims to improve speed while reducing energy use.

“We are building a fundamentally new and efficient way of connecting GPUs within data centers,” Linga said, adding that the approach could help accelerate the expansion of scalable AI infrastructure.

The company also offers chip-level interconnect solutions designed to improve communication between components within processors. Its Glasswing technology uses a proprietary method to transmit data more efficiently while reducing power consumption and design complexity.

Kandou AI has shifted its focus toward AI infrastructure in recent years to capitalize on rising demand for high-performance computing systems. The company said it is working with selective large-scale cloud operators and aims to expand manufacturing of its connectivity chips as demand for AI infrastructure accelerates.

Disclaimer: This article is based on information from Kandou AI’s official announcement and other publicly available sources. NervNow has not independently verified the claims.

MORE IN AI & INFRASTRUCTURE
IBM and NVIDIA Expand Partnership to Scale Enterprise AI Deployment
Nebius-Meta $27B Deal Highlights Big Push in AI Infrastructure Expansion
Kluizs in Talks to Raise $25 Million as AI Data Center Demand Grows
AI in Retail: What Reliance, Tata are Building that Other Retailers are Not