This is the latest in a brief series of articles about white-box switching. Earlier installments profiled Cumulus and Pica8 and covered a new open-source switch from Centec.

A common assumption is that white-box switching will lead to cheaper switching for everybody. That is, the white box itself — off-the-shelf hardware that's not built for particular networking software — will provide the competition to drive down prices and, for today's vendors, margins.

At this early stage, though, it's hard to tell if that's really how it's going to play out, or how mild or severe the effects will be.

Even proponents agree that capex — the cheapness of the boxes themselves — won't drive this market. Capex gets customers' attention (who doesn't like the idea of spending less money?) but it's really the operational savings that hook them.

Those savings don't come automatically, though. "The economics or the ROI difference might not even be that large any more," says analyst Seamus Crehan of Crehan Research.

The price difference between generic and name-brand switches is "narrower than it used to be" in the first place, he says. Then there's the cost of making a change to white-box switching. "It's no trivial task to change your supply chain to a white-box model," Crehan says. "You also need to have a software stack, and you need to invest in and maintain that software stack over time."

That might help explain the resilience of Cisco, which has kept its market-share lead through multiple phases of low-price threats. "We can look back and say it must be about more than price," Crehan says.

Even so, price is likely to have some effect. In July, analyst Rod Hall of JPMorgan Chase wrote an extensive report concluding that the data-center switching market could shrink to $6.8 billion in 2015 compared with $7.6 billion in 2012.

"We do not believe many observers expect the [data-center] switching market to decline over the next three to five years, but our analysis suggests this scenario is a real possibility if things play out along the lines of our central case," Hall writes.

Hall did have to make a lot of assumptions, which his report carefully enumerates. He's measuring only the data-center market there, not telecom or enterprise. And he had to take a guess at what the operating system would cost on a white box. The $6.8 billion result is based on a middle estimate where the software costs 30 percent of the ODM hardware. If that figure is 60 percent, the 2015 estimate grows to $8.9 billion.

Quantifying the specific effect on Cisco is tougher. Hall took a swing at it, coming up with 2017 earnings-per-share range of $1.66 to $2.29. (Cisco's earnings per share for fiscal 2013, which ended in July, were $1.99. All the earnings-per-share figures in this story are non-GAAP.)

"While this range is relatively wide we would point out that it implies negative EPS growth over the next few years for Cisco," Hall writes — and because of that so-called negative growth, he's considering Cisco stock to be materially overvalued. "More optimistic scenarios are certainly possible but even then earnings growth is challenged unless one assumes that 'bare metal' hardware fails to take off. We believe this is unlikely."

In fact, Hall's analysis assumes that 85 percent of the data-center switching market will convert to bare metal and that the non-switching parts of Cisco will exhibit a 4.9 percent compound annual growth rate. The range of earnings possibilities comes from adding other assumptions; the worst-case scenario, for instance, assumes 30 percent of campus switching goes to white boxes as early as next year.

The conversion to white boxes has started in some quarters, particularly the huge data centers run by the likes of Facebook and Google. Pica8 CEO James Liao was able to find customers even in his company's earliest days, taking advantage of contacts from Stanford University. Analysts don't realize how many deployments are out there, he says.

"It's a tough market to track, because if you're somebody like a very large cloud provider who goes to a white-box manufacturer, it's hard for people externally to have good visibility on that," Crehan says.

Concentrating solely on the data center, Crehan guesses that white boxes account for about 2 million ports per year, about 5 percent of the market. White-box sales will almost certainly grow faster than the overall market does, but it's going to be a gradual increase, he emphasizes. "Maybe it gets to the 15 percent range or so within, say, five years," he says.

It's hard to say what effect that will have on the incumbent vendors' sales. But the white-box movement has already had an effect on their way of thinking, Crehan says.

"Just look at the past year, look at the amount of new announcements and innovation from the existing data-center OEM switch vendors. We've seen much more openness, much more incorporation of SDN," he says.

You can see it in the incumbents' eagerness to claim turf in software-defined networking (SDN). Here are some recent examples:

• Cisco's Insieme Revealed at Last: All Your Policy Are Belong To Us
• Juniper's Metafabric Is a Chip Off the Old QFabric
• Extreme Networks Joins Broadcom's Trident II Parade
• Alcatel-Lucent's Nuage Adds a Gateway to Its Virtualization Offering
• HP Opens the Doors on an SDN App Store
• Brocade Doubles Down on the Data Center