Networking in 2022

For most who read my blog, network issues and opportunities are paramount. The good news for that group is that the Street almost universally sees networking at the top of the hardware opportunity list. The bad news is that when you look at the companies the Street really likes, you don’t see any of the major network vendors, and in fact just a couple vendors that most would consider “network vendors” at all.

In assessing the network space, my own strategy is to say that there are three factors that we have to consider. The first two, service provider and enterprise networking trends, are covered in this blog. The final factor, the Internet, is covered in the next blog, the one for December 22nd, which will be my last blog of 2021. There’s an interplay between this blog and the next, just was there was an interplay between Monday’s blog on software and the cloud, and this one. There’s no easy way to cover what’s going to happen in 2022, so I’ll ask you to bear with the approach I’ve taken and refer to items in whichever of the three blogs you need to address your own interests.

OK, let’s get started. The top trend in networking, from the Street’s perspective, is 5G. They’re citing “strong 5G handset demand”, which is why the top company pick in “networking” is Apple. They’re also citing strong 5G-related telco capex, of course, and some “pent-up” spending on enterprise networks. Let’s look at these points first.

There’s no question that 5G modernization of telco networks is the biggest source of incremental (non-modernization) capex for the telco space. I’ve noted in many blogs that 5G is the only significant technology that has budget approval, and that approval helped in 2021. In 2022 it may be so critical as to swamp other issues for vendors who want to sell to telcos. However, it’s important to separate “strategic” 5G that would (purportedly) create new mobile network demand, from “tactical” 5G which is just more of the orderly evolution of wireless generations we’ve seen over the last 20 years.

It is true that 5G can deliver better wireless speeds to consumers, but far less true that this makes any real difference to most mobile users. Even video delivery doesn’t require per-user bandwidth more than LTE provides. My straw poll of a couple hundred tech people who’ve made the LTE-to-5G transition shows that less than 10% are “confident” they can tell when they’re on 5G versus an earlier technology, without looking at an indicator on the phone.

Why then the big Street bandwagon for 5G handset vendors? Most of the Street analysts I talk with say that they believe that the “leading-edge” consumers will “demand” 5G because it’s the latest and greatest. My straw poll says otherwise; over half of those who have gone to 5G smartphones did so because 5G came on the latest models, and it was time for them to upgrade. The next-largest segment said that they wanted 5G because they believed that coverage available on earlier generations would likely get gradually worse over time. Those who actively wanted the “differences” of 5G were the smallest group. I most note here, though, that my poll included people who were more likely knowledgeable about mobile network trends; the average consumer might not have the same view.

5G is important to the Street in telco infrastructure, as I’ve already noted, because it’s budgeted. There is another factor, which is that 5G could favor a broader vendor list in the mobile network space than we’ve had in the past. Three vendors have dominated mobile networking; Ericsson, Huawei, and Nokia. Huawei is suffering from the pushback against Chinese companies for security reasons, and interestingly neither Ericsson nor Nokia has beaten Cisco in share price over the last year. It seems pretty clear that the Street hasn’t favored mobile incumbents, despite the 5G push.

The reason may be the open-networking trend, which in 5G means Open RAN and O-RAN. Nokia is widely seen by the telcos as the vendor most open-committed, and they’ve beaten rival Ericsson in share price over the last year too. However, part of the stock-price trends relate to growth opportunity, which generally means that the Street likes smaller players with more upside than giant incumbents who can’t gain much market share without a major revolution. Arista, one of the Street favorites, handily beats all the mobile incumbents and Cisco in share price appreciation.

5G would logically mean an increase in traffic, which means IP equipment in general would be expected to show strength. Cisco, the market leader, is at the bottom in terms of share price growth, with rival Juniper doing a bit better and Arista doing a lot better than either of these. Putting things in Street terms, they’re expecting to see Juniper gain some market share and Arista to gain a lot.

Arista’s big plus, says the Street, is the “hyperscalers”, meaning public cloud and other high-density compute players. Juniper gets a nod in this same space, but there’s also Street enthusiasm for Juniper’s AI push, which promises to address operations challenges that are clearly developing for any large-scale network operator, including the cloud providers. Between Cisco and Juniper, the Street tends to like Juniper best.

Then there’s Ciena. They’ve run consistently behind all the network vendors I’ve mentioned in terms of stock price appreciation, until their last quarterly report, at which point they jumped into a favored position (along with Arista, who still outpaces them in stock price). There are some good reasons for this, but there’s also a systemic shift that we need to look at first.

There was a big change in network equipment vendor share pricing around the first of November 2021, one that didn’t mirror a major change in the indexes. Arista, Ciena, and Juniper all saw increases in share price, and I think this is reflecting a move by the Street to favor companies that they saw as having a better chance of gaining market share. The trend didn’t impact Cisco or the mobile vendors, which tells me that the Street was looking more at conventional growth, both in data center switching and perhaps to an extent, in routing. The fact that Arista saw the most appreciation means to me that the Street thinks switching is the better space.

Going back now to the Ciena story, the Street believes that it’s got a strong data center story, which is true. Some analysts also appreciate the story that operators are likely to invest more in capacity than in capacity management, favoring a lot of fiber-level stuff to simplify operations up at the packet layer. The second story is more significant in my view than the first, given that as the Street admits, Ciena is the leading vendor in the data-center fiber space.

OK, that’s a summary of the Street’s view of networking. Let me now go over what I’m hearing, and what I think. It’s sometimes significantly different.

The first difference is that the Street focuses on technology segments but you really need to look at network topology segments to get the right picture. The most important trend in networking is metro. Metro is where access networks concentrate traffic. Metro is the deepest point where you can apply service intelligence efficiently. Metro/edge is the Last Frontier of hyperscaler deployment, whether you believe the cloud will move out or telcos will take up carrier cloud.

The reason this is important is that the architecture of the metro infrastructure determines the device makeup within it. All of the factors I’ve cited influence the infrastructure, meaning that the extent to which the develop, the pace of the development, and whether metro infrastructure is actually architected or is just evolving, establish what device types, and what vendors, might win.

Right now, the Street is assuming that “hyperscaler” opportunity is really just expansion of the public cloud deployments, and that 5G isn’t driving any conspicuous edge deployment at all. That’s possible, of course, and if it happens then metro equals 5G “devices” (which might well be tightly integrated metro server boxes) plus backhaul fiber. Nokia (leader in the Street view for 5G) and Ciena (leader in fiber capacity) win in network equipment, with Arista and Ciena still winning in data center.

In order for this not to happen, you need one or both of two new forces. The first is metro architecture and the second is carrier cloud. If a vendor were to position a cohesive vision of how metro, 5G, edge computing, and new services all combine in a single architectural vision for metro, that vendor could empower its own strategies by showing how they create the metro/edge of the future. On the other hand, if a vendor could define how operators could/must/should deploy their own edge computing, and tie that to 5G, they could make carrier cloud data centers a priority, link it to 5G, and win with it.

The impact of metro architecture and carrier cloud would be felt most directly in the way that operators see white-box technology. The metro area is the hottest focus of investment, where most 5G technology other than towers and radios would be placed. Right now, there is no real “metro architecture” in play, and certainly the big network vendors aren’t competing for a win in the metro. That means that the open-model strategy promoted by 5G in general and O-RAN in particular could define the metro network, and 5G and the hosting of 5G-related services (if they develop) could define edge computing. That would likely encourage white-box deployments in the metro area, which would then validate white-box technology in the places where operator budgets are focused. Obviously, that could create issues for the incumbent network vendors.

Moving on to the enterprise, we come to a major shift that gets essentially no notice from the Street—virtual networking. Virtual networks are a part of cloud data centers, and they’d likely be part of the carrier cloud too, but in those missions they’re generally invisible, and most of the cloud providers have used their own technologies in those places. In the enterprise, we’ve had “SDN” in data centers, which really means virtual networking and often VMware’s NSX, but the big news has been SD-WAN.

Enterprises aren’t likely to make convulsive changes in their network technology. The majority of them fear vendor lock-in and have embraced multiple vendors in multiple places, but they’re disinclined to mix vendors within a single area of their network. Since we have no real drivers of change in the enterprise (5G is such a driver for network operators), technology evolution depends on modernization budgets, and those rarely fund total switches of technology, or vendors. But virtual networks are different.

The SD-WAN mission of virtual networks typically adds a device to sites that had been previously left off MPLS VPNs because of service costs or availability. That device, though, is essentially a branch router, and so it doesn’t represent a major change in architecture. There are also typically software or appliance additions in the data center and perhaps other major sites, to terminate the SD-WAN virtual network and reintroduce traffic in a normal form. This basic mission of SD-WAN has actually gained a lot of traction, and it’s important because enterprises are roughly five times as likely to select a vendor for SD-WAN that’s not incumbent in their networks, as to add a new “network vendor” overall.

Under the covers of SD-WAN’s initial VPN-extending mission, there’s another realization, which is that SD-WAN is an enterprise virtual network in general, one that’s adept not only of networking thin sites, but networking clouds and cloud applications. In this role, it can provide a much higher level of security than ordinary firewall-based solutions could offer. A very few SD-WANs, including Juniper’s 128 Technology acquisition, offer a strong zero-trust model.

Security is the most-supported enterprise network priority, with about a third of enterprises saying that their budget for security is equal or greater than their budget for network expansion. If SD-WAN taps into security, it could bring virtual networking to the enterprise on a large scale, and that would transform how enterprises use networks and IP. It would then likely transform how enterprises buy IP services, which transforms how they’re sold and built by operators.

Virtual networks would tend to tap features from the transport level, and of course they would also remake how we view network security. Both these truths tend to worry incumbent router vendors in particular, but the fact is that the biggest impact of virtual networks might be that they’d tap a bit of the pressure for white-box networks in the enterprise.

Networking, then, is highly dependent on two points raised here—metro evolution for network operators and virtual networking evolution for the enterprises. A final force in networking evolution is the Internet, and that will be the topic of the last of my blogs on what to expect for next year.