Broadcom Advances Edge AI Infrastructure and Application-Aware Networking at MWC25

Broadcom's Edge AI Infrastructure Advancements

At MWC25, Broadcom Inc. unveiled significant progress in edge AI infrastructure and connectivity solutions. Vijay Nagarajan, VP of strategy and marketing for the Semiconductor Solutions Group, highlighted the company's achievements over the past year, emphasizing their commitment to open, scalable, and power-efficient AI solutions 1

1

.

Broadcom's advancements span across various technologies:

1. SerDes (Serializer/Deserializer): Crucial for maintaining strong signal-to-noise ratios in high-speed data transmission.
2. Ethernet and XPUs: Enhanced capabilities to support growing AI and machine learning demands.
3. DOCSIS 4.0 chipset: Integrates AI and machine learning for improved real-time processing in home networks.

Nagarajan explained, "SerDes needs to make sure that your signal gets across from one end of the link to the other despite all the noise that you would see in the electrical circuits" 1

1

.

Expanding AI Capabilities to the Edge

Broadcom is extending its AI expertise beyond large-scale infrastructure to edge devices:

1. Neural engines in broadband gateways for enhanced video processing.
2. AI-driven optimization in modems, smart applications, and passive optical network devices.
3. 50-gig PON devices with neural capabilities for improved network and power management.

These advancements aim to improve network optimization and real-time decision-making at the edge. Nagarajan drew an analogy between edge AI and the spinal cord, emphasizing its role in rapid, localized processing before data reaches centralized systems 1

1

.

Application-Aware Networking and Technological Convergence

Sanjay Uppal, SVP and GM of the VeloCloud Division at Broadcom, introduced VeloSky, a new product enabling application-aware networking for telecom operators 2

2

. This innovation allows for:

1. Dynamic resource allocation based on application needs.
2. Convergence of multiple networks (5G, fiber, satellite) at the application level.
3. Improved performance for critical AI workloads and healthcare systems.

Uppal stated, "You can combine two 5Gs together, you can combine satellite with 5G or with fiber -- in fact, any type of network" 2

2

.

Adapting to AI-Driven Traffic Patterns

The rise of AI applications is reshaping network traffic patterns:

1. Increased upstream traffic: Some AI applications require 100 times more upstream than downstream data.
2. Challenges for traditional network designs optimized for downstream traffic.
3. Need for flexible solutions to support real-time processing and machine-to-machine communication.

Uppal emphasized the critical role of networks in enabling agentic AI: "No network; no agentic AI. And that's really critical" 2

2

.

Future Implications for Telecom and AI

The convergence of AI, application-aware networking, and dynamic connectivity is poised to redefine enterprise network infrastructure. This shift presents new monetization opportunities for telecom operators and enables capabilities previously deemed impossible. As AI-powered devices and agent-based computing become more prevalent, the ability to intelligently allocate network resources in real-time will be crucial for unlocking the full potential of next-generation connectivity.