Hewlett Packard Enterprise (HPE) is trying to prove that its $14bn (£10.45bn) takeover of Juniper Networks is already translating into momentum in one of the most competitive corners of the AI infrastructure market.
Speaking at Mobile World Congress (MWC) in Barcelona, Rami Rahim, the former Juniper chief executive who now runs HPE’s networking business, told City AM the company had moved from outlining the rationale for the merger to delivering products and integrating the two firms.
“Last year… at Mobile World Congress, we were really talking about ambition and of what we wanted to do. This year, a year later, we’re talking about execution”, he said.
The deal closed last year after regulatory scrutiny in the US, creating a networking division that combines Juniper’s routing and switching tech with HPE’s Aruba networking business.
The combined unit now sits at the centre of HPE’s attempt to position itself as a major infrastructure supplier for the next wave of AI data centres.
“We have brought together two iconic businesses, amazing teams, great technology”, Rahim added.
Among the recent launches are new routing platforms designed for high-capacity data centre networks, as well as a liquid-cooled switch aimed at AI training and inference environments.
The company has also begun parts of Juniper’s Mist platform with HPE Aruba’s networking software.
Rahim said maintaining product development during the merger was an important sign for customers.
“It’s really important that we demonstrate to our customers and our partners that even in the midst of this large amount of work, we’re not losing sight of the importance of staying ahead of the game on innovation,” he said.
AI data centres push network capacity higher
The merger comes as demand for networking capacity soars alongside the rapid expansion of AI infrastructure.
“The demand for capacity right now is huge”, Rahim said. “The last time I’ve seen this kind of demand for network infrastructure was in the early days of the internet when it was just exploding to what it is today”.
While networking typically represents a smaller share of the cost of an AI data centre compared with compute hardware, its role is critical to how efficiently those systems run.
“If you’re spending millions, hundreds of millions, potentially billions of dollars on expensive GPUs, you want to be able to use them at as close to 100% utilization and efficiency as possible,” he said. “A big component of your ability to do that is going to depend on the network”.
HPE hopes its Juniper acquisition will give it a broader networking portfolio covering campus, branch, data centre and routing systems.
Rahim also argued that scale would help Juniper’s tech reach a wider customer base. “At Juniper, we had great technology… but we were doing this in spite of the fact that we didn’t have the kind of reach that a company like HPE enjoys today,, he said.
Beyond data centres, telecom operators are also looking at how AI could create new sources of revenue and place additional pressure on their networks. Operators are trying to balance the rapid growth in data traffic with limited new income from connectivity services.
They’re seeing growth in the capacity that they need to carry on their network without necessarily enough new revenue to compensate for the investment.
For some, that has meant exploring sovereign cloud or AI infrastructure projects.
Rahim said many telecom groups attending MWC were considering such initiatives as they look for new business models.
“Many of the telcom operators… have sovereign cloud initiatives that they’re executing on or they’re seriously considering,” he said.
At the same time, businesses are reassessing how they use public cloud services for sensitive data used in AI systems, and many companies are moving toward hybrid systems where data remains closer to where it is generated.
“They find it too risky to put it all in public cloud or to use them with public LLMs”, he said. “They are seriously reconsidering their cloud architecture… to one that’s going to necessarily be a hybrid cloud architecture”.
#HPE #leans #Juniper #deal #drives #demand #faster #networks