Juniper Networks released their second quarter earnings last week, and the numbers were pretty strong. The company reported a 13% increase in revenues y/y and 4% sequentially, and beat forecasts on both revenues and earnings. They also reported increases in product orders, and they’re working through their order backlog as supply chain issues resolve. Significantly, enterprise sales were the strongest piece of the story, accounting for 45% of total revenue, while service provider and cloud provider sales were weak, gaining only 1% for the former and being off 6% in the later. What can we say about all of this, for Juniper and for networking overall?
The key technology element in Juniper’s success was Mist AI and the product areas (Juniper calls them “Mistified”) associated with it. Juniper has broadened their original Mist acquisition from WiFi to wireline in general but SD-WAN in particular, and this has been a major success for the company. Their resulting “AI-driven Enterprise” portfolio has the best name recognition among those enterprises I’ve chatted with who are actively looking for a new or highly modified enterprise network strategy, but the number of enterprises involved is relatively small, so there’s some statistical risk here. Just under three-quarters knew about it, which is a quarter more than any other vendor’s overall architecture managed to achieve. Not surprisingly, that solution showed 63% growth in Juniper’s quarter, and that leads me to believe my sampling of enterprises on the topic is valid.
Juniper is also seeing traction with its Marvis Virtual Network Assistant, which they enhanced with ChatGPT. Marvis doesn’t yet get the recognition that Mist does, and none of the enterprises I talked with had adopted the ChatGPT piece, though 3 said they were “trying it out”. Still, it seems that Juniper’s success is increasingly tied to AI, and that AI is pulling through Juniper devices and thus at least partially responsible for the uptick in device sales. Furthering the enterprise story, Apstra showed strong gains in customers, orders, and revenues. Apstra is a data center operations automation tool and figures in Juniper’s Cloud-Ready Data Center story.
For the industry, I think all of this means that network equipment is not the best source of differentiation in its own space. It’s the software-based add-on elements that matter the most. That obviously raises the question of whether white-box technology, which everyone agrees was over-hyped (what isn’t these days?) when it first came along, might now have a chance to penetrate enterprise networks overall. I think we’re still a way from having a white-box enterprise revolution, but I also think that vendors need to be watching the risk, and looking at how they could defend against it.
Juniper adjusted their full-year revenue guidance downward, reflecting what Juniper says is a fairly broad market softening that we’ve seen this year as central banks raise interest rates to fight inflation. Guidance weakness resulted in the stock taking a hit, and while it stabilized this week, it hasn’t regained its previous level. Revenue from the cloud is expected to remain soft, and Juniper offers the same macro-economic reasons for that, but I think that they’re missing a more fundamental issue in computing and they’re still perhaps not fully leveraging a service provider trend that could favor them significantly.
The biggest problem with the cloud, according to enterprises I’ve interacted with, is that it’s been overdone in multiple ways. First, cloud hype has driven cloud projects beyond the extent that normal business justifications could take it. Senior management, who are typically exposed to the less-technical mass-market publications and positions, were taken with the notion that the cloud was going to take over everything. That was never going to happen. Second, cloud projects have overrun on costs significantly, largely due to the fact that development teams have focused more on innovation than on optimization. The cloud overshot its own rational business case, and so it’s retrenching to fit reality. That isn’t transitory, it’s a permanent shift that will favor the data center.
We have, with generative AI, another hype wave to contend with. There are plenty of good AI stories out there, and Juniper is actually the top player in “network AI”, with a massive three-quarters of enterprises putting them there, double the number who cite all other vendors (my comment on statistical significance still applies here). However, this isn’t a generative AI story, and the real value of generative AI to enterprises so far lies in pedestrian applications like drafting press releases and technical documents. It’s probably smart for Juniper to stick some ChatGPT story into their AI, but it wouldn’t be smart to think that this was likely to move mountains for them. Similarly, generative AI isn’t going to create a long-term cloud opportunity because willingness to pay for it is low. None of the enterprises I’ve talked with say they have any significant financial commitment to the use of generative AI at this point, and only 18% think they might develop such a commitment in 2023. Even more significantly, there’s no significant commitment to hosting generative AI among enterprises, so it can’t boost sales of switches.
The real question for Juniper in the long run, IMHO, is how they’ll promote their Cloud Metro vision. In abstract, Cloud Metro is a fusion of two trends. First, generally, there is a trend toward a tighter coupling of networks to hosting platforms. Enhancements to enterprise productivity, support for IoT, gaming and metaverse applications, and even AI all demand lower latency for tighter coupling between applications and the real world. Second, there is a trend toward integrating hosted features with connection services. You can see this most clearly in 5G and NFV applications, but there is little doubt that if what AT&T sometimes calls “facilitating services” are to be deployed, they’ll be based on hosted elements. This hosting is most likely to be done at the metro level, where latency and economy of scale reach their optimum balance.
Juniper mentions metro twice on their call, and I still believe that metro is what could magnify Juniper’s overall opportunity. That probably should start by framing Cloud Metro as a solution that applies a combination of tight network-to-hosting coupling and superior management, both of which are valuable to enterprises in a different context. This is really nothing more than a positioning exercise because Juniper already has the individual pieces. Combine Apstra with switching solutions and Mist/Marvis management and you have a great enterprise story as well as a framework for realizing the carrier cloud and even cloud provider opportunities.
This might be a good time for Juniper to take a look at that positioning. As I noted in another blog, the face-off between NVIDIA and Broadcom is introducing the possibility that networking is going to evolve closer to the GPU chip in order to manage latency in generative AI applications. That’s not the only latency-sensitive application, of course. IoT in general is latency sensitive, and so is the social metaverse and perhaps even the next logical steps in worker productivity enhancement.
Since the dawn of commercial computing in the 1950s, we’ve had three periods where business IT spending grew significantly faster than GDP. Each corresponded to a shift in computing that brought IT “closer”, meaning more directly involved with, worker activity. The next logical step in that progression is “point-of-activity empowerment” where IT actually participates in how the worker does the job. That level of IT participation demands that the applications improve rather than constrain worker action, which means latency has to be very limited. IBM used to push “sub-second response time” in their marketing, but it was too early in the progression of empowerment to be effective. Now could be the time.
Positioning isn’t everything, of course. Most companies, including Juniper, have had a long-standing battle between sales and marketing, and in the last couple of decades, when delivery of information was more a constraint on empowerment than production of information, sales usually won out. Past history, IMHO, shows that when it’s time for a paradigm shift it’s important to get out in front and lead, and that can only be done through marketing and positioning.
I think Juniper is in a great position here, but not in a position that vests the company with automatic success and expanded revenue. What they do in the next couple years will likely determine whether they gain share significantly on rivals like Cisco, or are consigned to a subordinate position. I think Cisco’s quarter is likely to be strong, so the number of opportunities Juniper has to gain on their rival may be declining.
