CIOTech Outlook Team | Tuesday, 05 August 2025, 03:11 IST
Broadcom's silicon division launched the Jericho4 networking chip, engineered to connect data centers up to 60 miles apart while accelerating AI computations. The chip enhances network performance for cloud computing giants like Microsoft and Amazon, addressing the growing demands of AI-driven workloads.
As AI applications become increasingly complex, they require linking thousands of graphics processing units (GPUs) across vast networks. According to Ram Velaga, Senior Vice President and General Manager of Broadcom’s Core Switching Group the Jericho4 chip is designed for massive scalability, with a single system capable of integrating approximately 4,500 chips. This scalability ensures efficient data movement within and between data centers.
To tackle network congestion, Jericho4 incorporates high-bandwidth memory (HBM), similar to that used by Nvidia and AMD in AI processors. "The switch is actually holding that traffic (in memory) till the congestion frees up," Velaga explained. "It means you need to have a lot of memory on the chip." The chip’s design accounts for longer data transmission distances, necessitating substantial on-chip memory to maintain performance.
Also Read: Airtel Launches Sovereign Cloud with Singtel, Globe, Africa Tie
Security on externally transferred data is essential outside the boundaries of data centers, considering that there are high risks of interception. Jericho4 improves security because the data is encrypted and therefore not exposed over the network. The chip, woven through an advanced three-nanometer manufacturing technology of TSMC, includes enhanced performance and efficiency.
The Jericho4 introduces features that increase network traffic capacity, ensuring seamless operation across large-scale data center networks. As AI workloads intensify, Broadcom’s latest chip positions the company to meet the evolving needs of cloud computing infrastructure with robust, high-speed, and secure connectivity solutions.