Broadcom's Jericho4 networking chip can transfer larger volumes of data at higher speeds, enabling customers to link up multiple smaller data centers to create a single big system for developing or running AI models. The chip can connect over 1 million processors across multiple data centers and handle about four times more information than the previous version. This addresses the need for data center capacity nearer to customers and helps speed up the answers users get from AI models.
Broadcom's silicon division has launched its Jericho4 networking chip, designed to connect data centers over 60 miles (96.5 km) apart and accelerate artificial intelligence (AI) computation. The Jericho4 chip, part of Broadcom's advanced AI networking portfolio, aims to enhance data center efficiency and security, addressing the growing demands of AI and cloud computing [1].
The Jericho4 chip introduces several key improvements over its predecessor. It can handle about four times more information than the previous version, allowing it to connect over 1 million processors across multiple data centers. This capability is crucial as AI models grow in size and complexity, requiring more distributed infrastructure [2]. The chip is built using TSMC's three nanometer process, ensuring high performance and efficiency.
One of the primary features of the Jericho4 chip is its ability to manage high volumes of data traffic. It employs high-bandwidth memory (HBM), similar to that used by Nvidia and AMD for their AI processors, to handle the data congestion that arises in large-scale networks. The chip can hold traffic in memory until congestion is alleviated, thereby improving network performance [1].
Security is also a significant focus for the Jericho4 chip. Broadcom has incorporated data encryption to protect against potential attacks that could intercept data as it travels between data centers. This is particularly important for cloud computing companies that require secure data transfers [1].
The launch of the Jericho4 chip comes at a time when the demand for data center infrastructure is surging. According to Business Insider, companies like Amazon, Microsoft, Google, and Meta are investing heavily in their infrastructure to keep up with the AI boom. By the end of 2024, the number of data centers in the US had nearly quadrupled, with Northern Virginia and Maricopa County, Arizona, emerging as significant hotspots [2].
This unprecedented build-out of data centers is driven by the need to process vast amounts of data and run complex AI models. However, it also comes with significant environmental costs, as large data centers can consume massive amounts of electricity and water [2].
In conclusion, Broadcom's Jericho4 chip represents a significant advancement in AI data center networking. Its ability to handle larger volumes of data at higher speeds, along with its enhanced security features, positions it as a key player in the growing AI infrastructure market. As data center construction continues to boom, the Jericho4 chip is poised to play a pivotal role in ensuring efficient and secure data transfers across distributed systems.
References:
[1] https://finance.yahoo.com/news/broadcom-launches-jericho-chip-advance-210207942.html
[2] https://www.businessinsider.com/big-tech-ai-data-center-spending-construction-map-2025-8
Comments
No comments yet