Could the cloud really become the network? That’s a question that many (including me) have asked over the last decade. When the view was “everything is moving to the cloud”, that would be an almost-inevitable outcome except for broadband access, but since I’ve rejected that universal-cloud theory from the first, and since I’ve reported that enterprises have also rejected it, the question is whether cloud attitude shifts would alter the idea that networking, at least in a WAN sense, would be subducted into the cloud. Since I’ve chatted recently with several dozen enterprises on detailed cloud plans, I have some new data to share.
Before I start, I need to make it clear that “the network” I’m talking about is the business WAN, what today is the realm of MPLS VPNs. I’ll touch on two other kinds of “network” here at the end, the Internet and the business and data center LAN, because the forces acting on the WAN also act on them, and because these three network domains each act at least a bit on each other.
We also need some definitions. What is a business WAN, and what is “the cloud”? A business WAN is what connects a company’s sites together. The cloud is any strategy that offers hosting as a service from a pool of resources. That’s intended to be a simplification, because it’s the basic concepts that are shifting here so we have to disconnect them from current offerings and implementations to see the real trend.
If everything moved to the cloud, hosting-wise, then every business site would simply connect there; it’s the data center’s magnetism that justifies the notion of an independent business WAN. That’s why the universal cloud model creates the universal cloud network. However, there are also other forces at play, working against the independent business LAN.
Cost is the most critical one. Enterprises say they pay from ten to twenty times as much to connect a site to their MPLS VPN as the fastest and most expensive broadband Internet service that’s available. Of 77 enterprises who tell me they’ve switched at least some of their sites to a broadband Internet model (via SD-WAN), only three said the result was unsatisfactory in terms of QoS and availability. There is a growing enterprise view that, in particular where fiber broadband is available, it’s inevitable that more and more sites shift to use it. Among the 283 who commented on this, 250 said that they believed that within ten years they’d have “most” of their sites on SD-WAN, and 74 said they thought it would only take five.
Another factor is the impact of Internet selling and support. One CIO points out that his company has a set of applications whose access is controlled by what you could call “roles”, a casual browser, a customer, a partner. Why not an employee? As more power is offered directly to buyers and partners, the way those roles are supported converge with what’s available to company employees. That makes it harder to justify separating the workflows for all these roles, and more logical to converge them all on the Internet.
The cloud, the real cloud, is a factor at this point. The majority of enterprise cloud plans focus on the front-end or GUI-level relationships, so the cloud collects a variety of role workflows and concentrates them onto a small number of cloud-to-data-center trunks. The need for an independent business WAN to do that is thus at least minimized, and likely eliminated.
I don’t think there’s much question that the independent business WAN is going to disappear as these forces act on networking. I don’t think there’s much question that most of it will end up being absorbed by the Internet/cloud combination. This, of course, means that not only will enterprise WAN gear be diminished or eliminated, but that operator revenues for MPLS VPNs will slip considerably unless they find other service strategies to adopt and push to enterprises. The obvious move, one that most operators are already taking, is to offer SD-WAN services carefully, to ensure they get at least the leavings of the MPLS VPN space but don’t accelerate its demise.
In the end, it’s likely that the Internet is the real beneficiary here. Cloud providers aren’t likely to want to get into the access business; too little profit and too much investment. Since the user/cloud side of the traffic originates on the Internet, it’s not too difficult to see the possibility that corporate data centers would simply connect to the Internet, and through that to their cloud components, via SASE/SD-WAN. The alternative would be to pay cloud providers for a direct trunk, which is business the cloud provider isn’t likely to spend much time chasing, IMHO.
So the enterprise WAN network isn’t really going into the cloud. That option is likely totally off the table because everything isn’t going to the cloud, and everything here isn’t just all of enterprise IT, it’s also all of the Internet. Because, truth be told, the cloud doesn’t want it. Yes, they’d love all of enterprise IT, but it’s now clear that’s not an option unless providers dump prices so much that the cloud business tolerates margins like the telcos have. Which is why the cloud providers don’t want all of the Internet either. They don’t want to go lower on the value chain, they want to go higher. They need telcos to handle the low side.
Carrying some traffic is essential to the cloud providers, but they get data transit payments to cover the cost, and in any event you can’t process data you can’t get. If the cloud providers could get all the hosting, not just enterprise application hosting, they’d end up eating everything, including the Internet, but they don’t want that. They want high-margin hosting, high-margin business. They’d rather invest in AI, where there might be a hope of something with high margins, than in something whose low potential has been proven by the whole telecom industry.
What this says is that the cloud providers really face the same challenge that enterprise IT faces, that network vendors face; the challenge of new business cases. You can’t command good margins doing commodity stuff, and there’s nothing more commoditized than maintaining something whose benefits have already been reaped. The cloud’s growth, like growth in network spending, in telecom, whatever, depends on growing willingness to pay, which means the development of new missions. Not just new technologies; something new like AI is just another thing waiting for benefits to justify it. New justifications. I think that cloud providers and AI providers alike, including those like Microsoft who are both, are starting to see that. Maybe that will lead them to try to do something about it.
