HPE Acquisition of Juniper Would Mark End of An Era
At one point in the Godfather III, the last film in the highly regarded mafiosi saga, Michael Corleone, played by Al Pacino, exclaims: “Just when I thought I was out, they pull me back in.”
Although I am neither a mafioso nor an actor who plays one in a classic cinematic triptych, I can relate to the frustration Michael Corleone expressed in the Godfather III.
Since my retirement, you see, I have worked hard (how’s that for paradox?) to diversify and expand beyond the relatively esoteric market analysis of network infrastructure. Yes, I’m retired, but by no means do I consider myself a spent force. My objective here, and perhaps in future projects under consideration, is not to do what I’ve done before, to cover the same ground that I’ve already ploughed, but to pursue long-neglected interests that I had deferred during my many years of employment. Learning and discovery, not to mention creativity, should be lifelong pursuits.
I would hope that each of us is more than what we do for living. Still, in the world beyond our personal consciousness, we are identified with what we have done professionally. That’s how most people know us, a legacy defined by others’ perceptions, and one can never entirely cast it aside no matter how much one tries to move in other directions. Still, nobody wants to be defined by the past, like a dead insect trapped under glass in an entomologist’s laboratory.
Yet, as the news broke last night that Hewlett-Packard Enterprise (HPE) is reportedly in talks to acquire Juniper Networks for approximately $13 billion, a significant premium on Juniper’s market capitalization at yesterday’s market close, I felt that I should offer some insight to readers. I offer the following commentary not begrudgingly but willingly, even though I hadn’t planned to get pulled back into the arena of “inside-baseball” network-industry commentary. An admittedly inexpert analogy might involve an old rock band riven by ambivalence as a concert crowd requests nothing but a back catalog of golden oldies.
But forget all that. Let’s consider the context of the mooted acquisition of Juniper Networks by HPE.
End of an Era
If this deal goes through, I believe it in many ways will signal the end of an era in enterprise networking, by which I mean the development and sale of datacenter, campus, and branch networking to enterprises worldwide. I don’t mean that enterprises will no longer have need for such network infrastructure; for the foreseeable future, enterprises will continue to buy network infrastructure for obvious reasons. Networking is foundational plumbing.
What I do mean, though, is that the era of market dynamism and intense competition involving an abundant cast of vendors will have come to an end. If HPE acquires Juniper Networks, only Arista Networks will remain as a meaningful standalone networking vendor that offers an extensive product portfolio spanning the datacenter network, the campus network, and – admittedly to a lesser extent – the branch network. Yes, there’s Extreme Networks, too, but Extreme is mainly a campus player.
Not so long ago, the network-infrastructure market was teeming with vendors. The first generation of network vendors, of course, suffered a severe cull just after the dotcom bubble burst, leaving Cisco as the overwhelmingly dominant vendor. It’s true that Arista Networks, staffed heavily by former Cisco employees, launched in the mid aughts, but we should remember that Arista was funded by its founders, and not by venture capitalists. Venture capitalists were generally reluctant to back network companies that were in the business of competitively displacing Cisco, but Arista founders Andy Bechtolsheim and David Cheriton, whose Granite Systems had been previously acquired by Cisco, had no such compunction.
Gradually, as the years passed, the ranks of fallen networking companies grew: Brocade Networks (sold piecemeal to different acquirers), 3Com (sold to HPE), Force10 (sold to Dell), Avaya (sold to Extreme Networks), Enterasys (also acquired by Extreme), Mellanox (acquired by NVIDIA), were among those who sought, but didn’t always find, graceful exits.
A Brief Resurgence: Network Virtualization and SDN
Later, the network-infrastructure market experienced a brief resurgence of vitality, and an injection of venture capital to hopeful startup companies, with the advent of network virtualization and software-defined networking (SDN). But that wave has long since ebbed as the last decade advanced into its final years, leaving vendor detritus in its wake.
A notable network-virtualization success story was Nicira Networks, co-founded by Martin Casado, Nick McKeown, and Scott Shenker. In the summer of 2012, VMware announced its intent to acquire Nicira for $1.26 billion, and the latter’s technology was ultimately adapted and marketed as NSX.
For a variety of reasons, however, most other network-virtualization and SDN vendors struggled and eventually perished, usually sold for much lesser sums to established vendors. Juniper Networks, for example, acquired Contrail Systems in December 2012 for $176 million. Others who gradually sought exits, not nearly as lucrative in most cases as their financial backers had envisioned, were Big Switch Networks (acquired by Arista), Pluribus Networks (also acquired by Arista), Plexxi (acquired by HPE), Midokura (acquired by Sony Semiconductor Solutions), and Plumgrid (acquired by VMware). I’m sure I’m forgetting one or two.
Although not strictly an SDN or network virtualization company, Cumulus Networks, which sold a Linux network operating system and associated network management software (as part of network disaggregation), was eventually sold to Mellanox before the latter was acquired by NVIDIA.
A few of the startup companies from that era are still extant, but they have not had enjoyed the success their investors and founders anticipated. There was, for instance, a flurry of activity in disaggregated routing, predicated on anticipated demand from carriers, but most of those vendors have found that telecom customers are mixed blessings.
With the rise of cloud and cloud-native application environments, traditional network infrastructure, though still required, no longer retains its traditionally constituted perceived value.
In some cases, the need for enterprise refreshes of datacenter-network infrastructure has been reduced or eliminated completely by the migration of workloads to public clouds. Further, the aftereffects of COVID have influenced how and how often many campus networks are refreshed. Even at the branch, where an SD-WAN goldrush ensued for a few heady years, the market has matured, growth has slowed, and consolidation has ensued, with dozens of vendors winnowed down to a fortunate few.
The golden age of enterprise network infrastructure, defined as a hardware and software used to design and build networks in on-premises enterprise environments, is dead.
As our historical review attests, the enterprise networking market has been tottering toward decreasing relevance for some time. The network will still be there, undergirding and supporting the delivery of applications and data, but it’s not nearly compelling enough to draw the interest of power-law venture capitalists and the most ambitious entrepreneurs. More lucrative propositions are at hand, exemplified by the current wave of AI mania.
Moreover, as Kubernetes grew in popularity, driving adoption of containers and microservices, the perceived and real value of networking moved up the stack. This is not a new observation, from me or from others. At IDC, we first recognized this shift a few years ago. With technologies such as Cilium and eBPF, networking and network security moves into a sandbox in the network kernel. This development changes not only the role and nature of networking, but who designs network architecture and operates the higher-layer network to support applications and developers. The lower layers of the network will still be there, of course – where could they possibly go? – but that’s not where the action will be.
HPE-Juniper: Analysis and Implications
With our historical digression complete, let’s turn to HPE’s reported interest in Juniper and how an acquisition, integration, and assimilation might unfold. In Juniper, HPE would acquire a stronger portfolio in routing and datacenter switching, and it would also gain ownership of Juniper’s highly regarded, AI-driven Mist software, which intelligently automates Juniper’s campus portfolio of wireless and wired network equipment. It is here, in the campus, that potential overlaps will occur involving HPE’s Aruba networking portfolio. Presuming the acquisition proceeds, we’ll have to see how those overlaps are resolved. Customers should demand clear answers.
On the SD-WAN side, there’s also an overlap, but HPE already has solid representation on that front through its prior acquisition of Silver Peak Systems, which made an auspiciously timed pivot many years back from WAN optimization to SD-WAN. (Riverbed Networks, the WAN optimization market leader, was slower to adapt, providing gratuitous confirmation of the innovator’s dilemma, and never truly recovered from its intransigence.)
While the business and technology media will focus almost exclusively on the Juniper’s AI-based Mist technology, because the fashionable IT industry is obsessively chasing the shiny AI ball, there are other aspects of this acquisition to consider carefully.
Juniper’s routing portfolio, for example, should give HPE a larger presence at telco accounts, as well as more opportunity to cross-selling of its other compute and network infrastructure. Juniper routing and Silver Peak SD-WAN is one such combination, but by no means the only one.
In cloud accounts, too, especially at the hyperscalers, Juniper’s routing portfolio gives HPE entrée to new opportunities. For business reasons, relating to its enterprise focus and product portfolio, HPE hasn’t had a substantive presence at hyperscalers, but Juniper changes the picture. To retain Juniper’s hyperscale patronage in routing, HPE will need a light touch, leaving the Juniper hyperscale account teams, which focus heavily on engineer-to-engineer mind melds, to their existing practices and engagement models.
In the enterprise datacenter, Juniper’s Apstra software, which automates and validates design, deployment, and operation of multivendor datacenter network, could become a useful asset, providing wedge that might allow cross-selling of Juniper datacenter switching as well as HPE datacenter infrastructure.
The deal between HPE is only rumored at this stage. It has not happened, and it might not happen. That said, the signals received thus far are relatively strong, and confirmation of an agreed deal could be imminent.
The deal won’t only signal a significant merger between two longstanding companies, but also the end of a fascinating era in the networking industry.