ipXchange, Electronics components news for design engineers 1200 627

Baya Systems: Scalable On-Chip Interconnect IP for AI Compute

ipXchange, Electronics components news for design engineers 310 310

By Luke Forster


Products


Published


23 June 2025

Written by


At Embedded World 2025, ipXchange caught up with Baya Systems, a young startup solving a major pain point in AI system design: moving data efficiently between chiplets. Their focus is on building high-bandwidth, low-latency on-chip fabrics that are purpose-built for multi-chiplet implementations.

AI workloads today rely on tightly integrated compute blocks: CPUs, GPUs, and neural accelerators must communicate rapidly to perform real-time tasks. The challenge isn’t just in raw compute—it’s in how those blocks share data. Traditional interconnect methods often create bottlenecks. Baya tackles this head-on with silicon IP designed for dense, modular architectures.

Chiplet-first fabric design

Baya’s platform isn’t a generic NoC. Their team has developed a software-defined interconnect layer that supports different chiplet topologies and enables efficient partitioning. You can scale your design, reuse IP blocks, and maintain performance across chip boundaries.

Their approach gives SoC designers flexibility to switch between monolithic and disaggregated silicon. Baya’s IP allows close integration between compute units, without compromising on bandwidth or introducing latency. As AI systems grow more complex, this level of scalability becomes essential.

Strategic backing, global reach

Despite being less than two years old, Baya Systems has already attracted strategic investment from Intel Capital, Synopsys, Matrix Partners, and Maverick. The company is headquartered in Santa Clara, with a design centre in Bangalore and a growing UK office. This global footprint supports customers across the semiconductor supply chain.

They’re not just shipping IP—they’re helping customers rethink their chip architectures. For any team building next-generation intelligent compute platforms, Baya’s chiplet interconnect IP delivers both performance and flexibility.

Where it fits in your design

If you’re working on custom AI accelerators, multi-chiplet SoCs, or disaggregated compute platforms, this tech is worth exploring. The IP is already in early customer trials and comes bundled with software for rapid deployment.

You can watch our interview with Baya Systems to get the full picture—and see how this fabric-first approach could reshape your next design.

Comments

No comments yet

Comments are closed.

    We care about the protection of your data. Read our Privacy Policy.

    Get the latest disruptive technology news

    Sign up for our newsletter and get the latest electronics components news for design engineers direct to your inbox.