Cisco, like pretty much all network equipment vendors, is facing profit pressure. That might account for what sure seems like a hiring decision aimed at boosting its hardware creds, as Network World suggests. There are a lot of reasons Cisco might be doing this, and most of the reasons I’ll cite have likely contributed to the decision. We’ll look at them, and try to assess the goal, the impact, and the chance of success.
Cisco’s hire is Martin Lund, a veteran of Broadcom’s network switching line, Microsoft’s Azure initiative toward network operators, and Metaswitch, an open-source provider of telecom software. He’ll become Executive Vice President, Common Hardware Group, though it’s clear his skills extend beyond hardware. His background and experience play into some of the things I think are driving Cisco’s new initiatives, as we’ll see.
The first reason behind Cisco’s moves is improved differentiation. Cisco and rival Juniper have both been increasing their chip focus, proving that chips are a critical differentiator. The advent of white box switching based on merchant silicon (where Broadcom is a powerful force) has raised the stakes in terms of chip differentiation.
The second reason is getting a bigger piece of the pie. Operator spending on infrastructure has been in decline for several years, and enterprise network equipment spending is starting to come under pressure too. When the pie gets smaller overall, you need to get a larger part of to stay level with the past. To have to buy chips for switching or use in interfaces can dilute profits.
Reason three is to reflect new buyer priorities. Hosting network features, for example, is an emerging need and one that Cisco, with its long-standing UCS server position, should be strong in. Security is enough of an issue these days to generate interest in specializing hardware to address it, too. Finally, the increased talk about IoT and edge computing has induced most operators and vendors to look at latency management as a goal.
The network equipment market is changing, fastest in the telecom/MSO space but relentlessly even for enterprises, and M&A is both an indicator and a driver. Broadcom’s acquisition of VMware and HPE’s pending deal for Juniper are both, I think, a mixture of the value of consolidation in a commoditizing market and an attempt to work out of the commoditizing past. We need to look at the three Cisco drivers in light of each of these two deals.
Broadcom is already a threat to Cisco and to its switch/router rivals because its switching chips are used in the best and most competitively successful white-box devices. DriveNets, whose routers are used by AT&T in its core, is using Broadcom chips, so fending Broadcom off with better chips already makes sense. Add in the risk that Broadcom would “climb the stack” with VMware software deals and white boxes, or just leverage VMware’s position with enterprises to help white box sales and you can see the reason Cisco might want to take out some chip insurance.
Data center switching is the big network equipment priority for enterprises, and decisions there are driven more by IT issues than network issues. If Broadcom somehow makes white boxes and IT virtualization into a package, their sales force has a tight tie to the decision-makers. That same link could also play with network operators who are doing services based in part on routers and in part on data center feature hosting.
The Juniper/HPE deal could be even more relevant to Cisco’s interest in a hardware push. HPE is already a provider of network gear through its Aruba acquisition, but Juniper is a broader Cisco competitor whose AI tools have given it a strong market position even when Cisco had a weak quarter. HPE is also an influential data center vendor, so if data center trends are driving network decisions, it might have better connections to push Juniper gear. As a big server player HPE could also play into telco plans for feature hosting, and even create ties between routers and hosting. Recall that the “user plane” of 5G could benefit from a tight coupling of 5G control plane activity and routers, as I’ve noted in prior blogs.
The biggest problem that the HPE/Juniper combo could create for Cisco may be in security. When Juniper acquired 128 Technology it got an SD-WAN product that was session-aware, meaning that it had the ability to recognize the typical user-to-application or app-to-app relationships being established. Juniper built this into some router products so create “session-smart routers”, and this could create a significant new security tool that might well be used by HPE, just as Cisco closed its Spunk acquisition in the security space. Could we see HPE build a tight link between session-smartness and applications? I’m sure Cisco would hate to see that.
Cisco also surely hates the idea that white-box chip giant Broadcom and HPE’s link with arch-rival Juniper might cut off a Cisco move upward from pure connectivity into IT. If networking is commoditizing overall, it adds to the problem that the advent of VPNs cut off most enterprise router opportunity, and operator capex that currently fuels router spending is under pressure. Cisco may hope to move up the food chain toward IT, which is where applications that control the business case for all tech spending live, but two M&As mean competitors are already there.
Chip initiatives take time too. Unless Cisco already has something in the pipeline, it will be hard for a Cisco chip initiative to mature fast enough, particularly if HPE/Juniper realizes symbiosis quickly, which it very well might. The good news for Cisco is that neither HPE nor Juniper have been marketing giants, nor for that matter have VMware or Broadcom. Cisco might have time, but only might and not for sure. And either of their new M&A competitors might see that, and decide to be a bit more radical in their own plans. Even if they could not fully counter a massive Cisco chip move, they could take steps to define the market state into which Cisco had to introduce their new offerings. They should try to do this, and Cisco should try to prevent it.
