Broadcom Unveils Jericho4: A Game-Changer for AI Data Centers

Broadcom just dropped a powerhouse chip that could redefine how AI workloads run across massive, geographically spread data centers. The Jericho4 isn’t just fast—it’s built for the future of cloud AI.

In a bold leap for AI infrastructure, Broadcom has launched its latest Jericho4 chip, built specifically to meet the skyrocketing networking demands of artificial intelligence in cloud data centers. Announced Monday, the chip connects facilities up to 60 miles apart, delivering the kind of bandwidth and memory muscle modern AI needs to thrive.

AI models today are more powerful than ever—but they also require an intense web of GPUs and servers, often spread across large-scale data centers. Companies like Amazon and Microsoft rely on these setups to train their next-gen AI systems. But when you’re moving massive amounts of data across long distances, even milliseconds matter.

That’s where Jericho4 steps in.

Key Takeaways:

  • Connects data centers over 60 miles apart for faster AI workloads.
  • Packed with high-bandwidth memory (HBM) to reduce network congestion.
  • Supports 4,500 chips in a single system, enabling massive scalability.
  • Encrypts data for improved cross-site security.
  • Built on TSMC’s 3nm process for better power and performance efficiency.
Broadcom,Jericho4,AI Data Centers,Jericho4 chip,cpu

According to Ram Velaga, Broadcom’s SVP of Core Switching, Jericho4 chips can “hold data in memory until congestion clears”, reducing delays and increasing throughput. This is critical in AI, where terabytes of data must move seamlessly from node to node.

Security, too, is a major win here. With cyberattacks on the rise, Jericho4 includes built-in encryption, helping protect data as it travels beyond traditional data center walls.

And thanks to being manufactured using TSMC’s cutting-edge 3nm node, the chip balances performance with energy efficiency—something AI operations desperately need. In short, Broadcom isn’t just playing catch-up—it’s aiming to set the standard for how AI data centers operate in a multi-site, high-demand world

Also Read

Leave a Comment