Over the last year, we have reported on the need to bolster data center network infrastructure to meet the performance demands of AI and other compute- and data-intensive workloads. Many think meeting that need is the reason behind HPE’s acquisition of Juniper Networks.
The deal, worth $14 billion, was formally announced yesterday. HPE’s hopes for the joint company are high. “There has been only one large vendor,” said HPE Chief Executive Officer Antonio Neri in interviews and press briefings. He was referring to Cisco, which offers a wide range of compute and networking solutions. He continued, “Today we are actually creating a second option — a much more modern option that ultimately customers were looking for.”
In news interviews about the deal, Juniper CEO Rami Rahim noted that the two companies will focus on building data centers and running networks for AI-enabled operations.
Complementing an existing networking lineup
HPE has made several major acquisitions in the networking space over the last decade or so. It acquired 3Com for $2.7 billion in 2010. And followed that up with the acquisition of Aruba Networks for $3 billion in 2015.
Juniper brings a large portfolio of data center switches and routers to the table. Over the last year, it has focused on automation to make the devices easier to manage. And more recently, it has bolstered that automation with AI. Additionally, the company melded security into its networking offerings, touting what it calls a zero-trust data center architecture.
Beyond acquisitions to ensure enterprises have a suitable network infrastructure for AI, HPE was one of the founding members of the Ultra Ethernet Consortium (UEC), an industry group that seeks to build a complete Ethernet-based communication stack architecture for AI and high-performance computing (HPC) workloads. (Other founding members of the group include AMD, Arista, Broadcom, Cisco, Eviden (an Atos Business), Intel, Meta, and Microsoft.)
AI brings a focus to the data center network
AI has forced many enterprises to look at ways to re-evaluate their network infrastructure. Efforts like the Ultra Ethernet Consortium are just one of many industry and vendor efforts.
Other activities include the development of new networking chips that can handle artificial intelligence (AI) and machine learning (ML) workloads, such as those offered by Cisco, Broadcom, and others. Additionally, the industry has seen the introduction of new network and AI workload accelerates, including Infrastructure Processing Units (IPUs) and Data Processing Units (DPUs).