Kandou AI Selects Baya Systems to Power Its Attack on the AI Memory Wall

Joint solution reduces risk and accelerates time-to-silicon for advanced semiconductor platforms

Baya’s WeaveIP™ fabric and WeaverPro™ software accelerating architecture development for Kandou’s 448G copper AI connectivity platform

Santa Clara, CA and Saint-Sulpice, Switzerland — July 14, 2026 — Baya Systems, the leader in software-defined fabric IP for AI and high-performance computing, and Kandou AI, the pioneer in high-speed connectivity for AI infrastructure, today announced a strategic IP licensing agreement. Kandou AI has licensed Baya’s WeaveIP™ fabric and WeaverPro™ software platform and is actively deploying both across its architecture and development workflow for the next generation of its connectivity platforms to overcome the memory wall constraining further AI development. The collaboration also demonstrates how software-defined fabric architecture and advanced connectivity work together to address the data movement bottleneck limiting next-generation AI systems.

Attacking the Memory Wall

AI model growth has outpaced memory bandwidth by orders of magnitude. While compute has scaled 60,000x over two decades, interconnect has managed just 30x. This clear mismatch is stymieing further AI development, since data simply can’t move at the speeds compute requires.

Kandou AI’s patented Copper MIMO technology (aka Chord™ Signaling) is closing that gap by sending correlated signals at real-time data processing speeds of 448 Gbps and beyond, providing the industry with strong alternatives to optical interconnects for a broad class of AI infrastructure applications. By using Copper Multiple Input Multiple Output (MIMO) technology, Kandou can deliver 10X the efficiency of conventional alternatives at reduced power, costs and greater scalability. The economic impact is direct: faster, denser, cheaper AI systems.

While Kandou advances the physical layer of AI connectivity, Baya’s software defined fabric platform optimizes data movement for complex heterogeneous compute, memory, I/O on die and across multiple chiplets. Together, the technologies enable customers to architect AI systems that are not only faster, but also easier to design, validate, and scale.

"Solving the memory wall is a system-level challenge. Our breakthrough signaling technology is only one part of a broader AI system. We need a fabric architecture that can move data with the precision and efficiency our products demand," said Srujan Linga, Co-Founder and CEO of Kandou AI. “Baya’s WeaveIP complements our interconnect technology. We provide physical scalability, with WeaveIP providing logical scalability to system design. Users are seeing in practice that WeaverPro significantly accelerates development timeline”.

Software-Driven Architecture, Faster to Silicon

Kandou AI selected WeaverPro to assist the exploration and validation of on-chip interconnect configurations, such as traffic class partitioning, QoS policy and protocol adaptation, before reaching the critical Register-Transfer Level phase of design and development. Because Baya's tiled implementation methodology assembles verified fabric primitives into a configurable architecture, Kandou AI anticipates significant acceleration of development cycles and time to market.

"The tiled implementation approach is a genuine competitive advantage for teams working at our pace," Linga continued. "Baya’s technical expertise and collaborative engagement has made them a true extension of our own engineering organization."

A Partnership Built on Technical Alignment

The engagement reflects a broader convergence: both companies are building for an AI infrastructure landscape where data movement, which gates the scaling of compute, is the chief challenge to be solved. It accelerates the trend in the semiconductor industry towards modular AI architectures, where innovation depends on combining best in class connectivity, compute, memory and system fabrics.

"Kandou AI represents exactly the kind of customer Baya was designed to serve," said Dr. Sailesh Kumar, Founder and CEO of Baya Systems. "They are attacking a real and fundamental problem, and they need a fabric platform that pushes beyond the data movement limitations of traditional NoC and enables them to realize their ambition. WeaveIP and WeaverPro offer the architectural freedom and implementation velocity to achieve that without compromise.

"The engagement further validates Baya’s growing role as the architectural foundation for next-generation AI infrastructure. As customers increasingly adopt chiplets, heterogeneous compute, and custom AI accelerators, software-defined fabrics are becoming a critical layer for optimizing data movement across the entire system.

About Kandou AI

Kandou AI (kandou.ai) is pioneering the next generation of connectivity for AI infrastructure. Its innovative Copper MIMO (aka Chord™ Signaling) technology — developed over more than a decade and rooted in information theory — enables high-speed, energy-efficient copper-based interconnects that extend performance beyond what conventional SerDes can achieve. Kandou AI is backed by leading technology investors including Maverick Silicon, BVP, Capital Group, SoftBank, Synopsys, Cadence and Alchip.

About Baya Systems

Baya Systems is leading the next wave of foundational, high-performance, and modular semiconductor systems technologies that are chiplet-ready and accelerate intelligent compute everywhere. Inspired by the baya bird’s nest-weaving ability, Baya integrates best-in-class compute, communication, and I/O components into seamless, energy-efficient solutions. Its software-based design and exploration platform enhances performance, yield and reusability, enabling cutting-edge, cost-effective innovation across multiple industries. Baya is backed by leading investors including Matrix Partners, Maverick Silicon, Synopsys Inc., and Intel Capital. For more information, visit Bayasystems.com or follow us on LinkedIn.