HPE finalizes Juniper acquisition, forms new AI-centric networking unit

Nearly 18 months after it was first announced, HPE today formally closed its $14B acquisition of Juniper Networks.

The deal was originally announced on January 9, 2024 and has undergone public scrutiny ever since, with regulatory reviews in the UK, EU and the US. It was the US that proved to be the final hurdle, with the Justice Department suing to block the deal in January of this year. That obstacle was overcome on June 27, paving the way for the closing of the deal today.

With the deal now closed, work begins in earnest to integrate Juniper into HPE. As part of the deal, there is a new HPE Networking business unit, led by former Juniper CEO Rami Rahim, which will contain the operating units of both HPE Aruba and HPE Juniper. The plan is to retain both brands, as the company continues to see value in them. What isn’t yet entirely clear is precisely how the new combined portfolio will shake out.

“This is day one, and we’re just getting started,” Rahim said during a press conference about the close of the deal. “But I can tell you, my objective here as the leader of this combined networking business is to build the best networking business, really on the planet, one that’s founded in innovation, and we plan on doing this first by starting to focus on our customers and partners.”

Department of Justice mandates Mist AI licensing, forces Aruba Instant On divestiture

While the precise reworking of the HPE Networking portfolio is being determined, what is clear is how HPE is responding to concerns from the US Department of Justice (DOJ). The DOJ’s antitrust review resulted in two specific requirements that will directly impact the technical roadmap and competitive positioning of the merged entity.

First, HPE must license Juniper’s Mist AI operations algorithms to competitors through an auction process. This affects the core machine learning models that power Mist’s predictive analytics. However, the licensing requirement is narrowly scoped to preserve HPE’s competitive advantages.

“We have agreed with the DOJ to offer a license, through an auction, to specific aspects of Juniper Mist, which is just the AI operations part,” HPE CEO Antonio Neri explained during the press conference.

The distinction is technically significant. Competitors will gain access to Mist’s anomaly detection and predictive failure algorithms, but not the underlying network operating system, hardware abstraction layers, or the customer-specific data models that make these algorithms effective in production environments. HPE retains control over data collection mechanisms, telemetry processing pipelines, and the integration points with physical network infrastructure.

“The intellectual property obviously stays with us, but we have to be able to support that as we go forward,” Neri clarified. “But again, it’s only the AI operations portion of the Juniper Mist stack.”

Rahim emphasized that the larger value in the Mist AI stack is the actual data. He noted that Juniper has over 10 years of learning from real world deployments, and that is extremely difficult for anyone else to replicate.

The second DOJ requirement mandates divestiture of HPE’s Aruba Instant On portfolio, which targets small business deployments with cloud-managed access points and switches. When asked about the divestiture, Neri attempted to minimize the impact.

“The Aruba Instant On is a very new business that we built over the last three years or so, and it is completely separate from the rest of the traditional HP Aruba platform, or Aruba Central,” Neri noted. “It is a unique offer targeting the SMB segment of the market, and more specifically, the ‘s’ of SMB; it is a very small business for us.”

AI for networks, networking for AI

A core element of the integrated portfolio will be AI. Rahim explained that there are two components to the strategy: AI for networks and networking for AI.

AI for networks focuses on applying AI to network operations and management. This involves using Mist’s predictive analytics to identify performance issues before they impact applications, automatically optimize environments, and provide prescriptive remediation for network problems. The integration will extend these capabilities beyond wireless networks to include wired infrastructure, security policies, and cross-domain correlation with compute and storage systems.

“We’ll have greater range and flexibility, because we’ll be able to now, with the strength that both HPE and Juniper bring to the table, satisfy more use cases with more deployment options, and then we’re going to have greater integrated security,” Rahim said.

Networking for AI addresses the infrastructure requirements for AI workloads themselves. This includes the high-bandwidth, low-latency fabrics required for GPU-to-GPU communication, as well as the integrated cooling and power distribution systems that modern AI clusters demand.

“The second opportunity of networking for AI is equally exciting. Here we’re building the large scale data centers, the factories for AI that are incredibly, incredibly important in today’s environment,” Rahim explained. “Together, we’ll be able to innovate faster across all layers of the technology stack, from silicon to systems to software, and we’re going to have a complete solution.”

Open Source commitment will continue

Both Juniper and HPE have a long history with open source technologies and that is going to continue.

“Open source has always been an important part of Juniper’s, and I’m sure also HPE’s, innovation agenda,” Rahim said in response to a question from Network World. “For example, in networking, open source in areas like Sonic has picked up some momentum and traction in the industry. For that reason, Juniper has decided to innovate in this area, and I’m sure that will sort of extend into the new HPE networking business.”

The expanded portfolio creates new opportunities for open source contributions. Rahim expects that expansion will extend Juniper’s contributions beyond just networking to also include compute and storage.

HPE is no stranger to open source networking either. The company is a founding member and key contributor to the Ultra Ethernet Consortium (UEC) which recently published its 1.0 specification.

Rahim emphasized that the combined company will remain focused on open standards as well.

“I think this combination gives us this awesome ability to combine technology components to provide a soup to nuts solution for our customers that can make their life tremendously easier, to deploy AI learning, AI inference,” Rahim said. “But I think a core tenet of providing those solutions is going to be to leverage open interfaces, APIs, because that’s what our customers want.”

Combined portfolio will tackle more use cases

While HPE is not providing a specific integration timeline, both Rahim and Neri emphasized that the company is taking a measured approach. It will preserve existing investments while providing migration paths.

“Nobody, no customer, no one is going to be left behind, contracts will be honored for the life of the products,” Neri stated.

The strategic imperative moving forward is to integrate AI, cloud-native, and networking technologies together. While there is some overlap across the existing portfolio, Rahim argued that often the technologies take different architectural approaches to solving networking problems.

“Together, the opportunity here is to build something that will capture more use cases, whether they need cloud, or on premises with integrated security, the sky is the limit in terms of how these things come together to capture greater levels of innovation,” Rahim said. “I feel like a little bit of a kid in a candy store right now, looking at technology components all around me, and my team and I are just super excited to roll up our sleeves and get that roadmap that everybody is of course very interested in seeing.”

Source:: Network World