The DoJ finally, after making a totally nonsensical objection to the HPE/Juniper deal, followed up with a somewhat compensating irrelevant settlement deal, leaving things just about where they’d been before the whole mess began. But, of course, where that might be in an overall market sense is still up in the air.
The CEOs of the two companies have made their statements, but as usual in this sort of thing, there’s not much insight to be gained by looking at them. Juniper now becomes an element of HPE Networking, with Rami Rahim as president and GM of the group. Enterprises have told me consistently that they thought the deal would be good for competition, meaning of course competition with Cisco, and they’re likely right, but even there, the questions are “Why?” and “Who benefits?”
Both companies have stressed the AI angle all along, but given that AI is the current media darling topic, almost everything being done gets tied to it whether there’s a good reason or not. In this case, AI might be one of two decisive reasons the deal could benefit buyers and the two merging companies alike, and give Cisco some angst along the way. It could even hit Nvidia and Broadcom, but it could also be a total miss and mess. It depends on what the entity does now, and we’ll look at that in both a technical and pure sales/marketing mechanics sense, starting with the latter.
Generally, the more you can sell a prospect, the more value you can gain from having control of the sales process. Juniper is very strong in the data center and WAN (big switches and routers) area, where HPE has been less competitive in the past. Juniper is also the network vendor that has gotten the most AI cluster hosting attention from enterprises I’ve chatted with, which suggests that if AI is a driver of data center growth on a large scale, HPE will now be able to exploit it more easily.
This is important to HPE because, like IBM, HPE has a good level of strategic influence in accounts (they do fall noticeably behind IBM, though). They could now push AI hosting projects with a better shot of getting the networking piece of the project, which means that they’re more likely to expend the sales/marketing collateral to boost the deals. That’s important because of some past research into enterprise purchasing habits I’ve covered for decades.
There are three kinds of data enterprises gather from prospective vendors when they’re buying. One is differentiators, the thing that separates choices and ranks them. Another is objection managers, things that have to be overcome to make those in the project approval chain comfortable, but the most important thing is the enablers, the specific things that contribute to the business case. If you don’t have enablers, you don’t have a project, but vendors tend to focus on differentiators because enablers necessarily require a building of the business case itself, a time-intensive process that educates the buyer and can easily end up making a deal for a competitor. Having a broad portfolio raises the benefits of a win, thus making the risks associated with enabler-building tolerable.
The technical side of the deal lies in cementing that unified benefit. The biggest problem enterprise CIOs face is the exploding complexity of the infrastructure that runs and connects applications to users. Enterprises have, for many decades, been challenged in hiring and retaining technical specialists in hardware, software, and networking, given the increased competition from vendors, startups, and consulting firms who can typically pay better and offer better career growth. This challenge renders them incapable of dealing with the complexity of the present, much less that of the future. When you ask CIOs what they need to assure, the dominant answer is “quality of experience”, but how do you do that when the totality of an experience is divided among a host of elements, each of which may be independently managed?
Given all of this, it’s clear that HPE Networking can offer a broader AIops solution, a larger footprint on the elements of QoE. It’s also clear that HPE has more account influence than Juniper had, and in fact my data suggests that’s true even for Juniper’s own accounts. Data center networking is driven by what the data center is built to house, data and data processing. HPE has a wide set of offerings for that, and Juniper has offerings to connect it. Good symbiosis.
Mist AI, as I noted in the past, is the most-recognized network AIops tool among the enterprises who offered comment. Since enterprises don’t tend to comment spontaneously on whether they’d like Mist AIops in HPE data center tools, all I can offer is that almost 90% of current Juniper Mist users thought that the HPE acquisition would improve their operations overall, double the rate of favorable views in the market at large. To me, that suggests that Juniper Mist users are eager for a broader ops footprint.
There are always questions, of course. One obvious one is that the exact way that fully integrated AIops will be done (Mist and HPE Compute Ops Management) isn’t yet known (there was a question on it on the call HPE/Juniper did after the deal was done). Another is how HPE will overcome the fact that most of the HPE prospects are not familiar with Mist, and thus not automatically happy about having it as an element in (or basis for) unified AIops. A third is whether the concession to DoJ on licensing the Mist AIops code to a select set of companies will erode the competitive benefit, and the final one is whether AIops is in itself a transformational opportunity that will create “green fields”. If not, then the competitive challenge to Cisco is still to be realized.
It likely comes down more to HPE’s work on AI than on anything limited to AIops. If there is an in-house AI cluster build, and if HPE can get a big chunk of it, then this deal could pay off in a big way. If not, then it’s hard to say whether the acquisition will really pay off for either company. I’ll watch developments here and keep you posted.
