On Monday, August 4, Broadcom introduced its latest Jericho network chip designed to connect data centers over 60 miles (96.5 kilometers) apart while significantly boosting AI computing speeds.

The new Jericho4 chip builds on previous models with enhanced features that improve network traffic performance within and between large-scale data centers. As AI workloads grow exponentially, linking thousands of GPUs becomes essential, demanding faster and more sophisticated network solutions. Leading cloud providers like Microsoft and Amazon rely on these advanced chips to ensure smooth and efficient data flow.

Security remains a critical factor when data travels across physical data center boundaries, where interception risks increase. Broadcom's engineering team developed the Jericho4 chip to support massive deployments—according to Ram Velaga, Senior Vice President and General Manager of Broadcom's Core Switching Group, a single system can incorporate approximately 4,500 chips.

To tackle network congestion, Jericho4 uses high-bandwidth memory (HBM), the same technology employed by AI processor leaders like NVIDIA and AMD. Beyond performance improvements, the chip also enhances security through robust data encryption.

Manufactured using TSMC's cutting-edge 3nm process, the Jericho4 chip marks a significant step forward in data center networking and AI acceleration.