Broadcom’s silicon division on Monday launched its next-generation Jericho networking chip, designed to connect data centers up to 60 miles (96.5 km) apart and accelerate artificial intelligence workloads. The Jericho4 chip introduces enhancements that increase the volume of network traffic moving across large-scale data center networks, both within and between facilities.
As AI development grows increasingly demanding, cloud giants like Microsoft and Amazon rely on thousands of GPUs working in tandem, requiring faster, more advanced networking chips to handle data efficiently. Security is paramount when transferring data beyond data center walls, guarding against potential interception during transit.
Ram Velaga, senior vice president and general manager of Broadcom’s Core Switching Group, explained that the Jericho chips are engineered for massive scale, with a single system potentially incorporating around 4,500 chips. To address network congestion, Jericho4 uses the same high-bandwidth memory (HBM) technology favored by AI processor designers like Nvidia and AMD. “The switch is actually holding that traffic (in memory) till the congestion frees up,” Velaga said. “It means you need to have a lot of memory on the chip.”
The chip’s memory capacity must also grow with the distance data travels to its destination. Beyond boosting performance, Jericho4 enhances security through data encryption. Broadcom built the chip using TSMC’s advanced three-nanometer manufacturing process, underscoring the focus on cutting-edge technology.
Broadcom’s Jericho chip builds on the company’s expertise in networking semiconductors by supporting ethernet fabric routing, enabling AI workloads to be distributed across multiple sites without the traditional bottlenecks. Analysts say this technology could democratize AI development by allowing smaller cloud providers to pool their resources into virtual supercomputers.
Following Broadcom’s earlier efforts to improve Nvidia’s AI processors within data centers, Jericho extends capabilities beyond single facilities, potentially reducing the need for costly new infrastructure builds. This wider reach is a key advantage as companies look to scale AI efficiently.
For partners like Meta and Apple, the Jericho chip offers strategic benefits. Apple’s collaboration with Broadcom on server chips expected in 2026 highlights the growing AI hardware partnership. Connecting data centers up to 60 miles apart, Jericho lowers latency and energy use, aligning with industry goals for sustainable expansion.
Investors have responded positively to Broadcom’s AI-driven growth outlook, with Jericho seen as central to its portfolio. While integration challenges and competition from Cisco remain, the chip’s scalability could usher in a new era of federated data centers, breaking geographic limits on AI development.
Broadcom’s broader AI strategy emphasizes connectivity and modular infrastructure, positioning the company as a crucial player beyond just chipmaking. With deployments like Meta’s upcoming servers, the industry may move toward highly interconnected AI systems that span regions and continents.
Pump Radar
Stock Alerts