Ten years ago, telecom operators promised that edge computing would change everything. It didn't. So why is the industry having the same conversation again, and is anything actually different this time?
In this week's episode of The Five Nine, host Diana Goovaerts takes you inside some of the dozens of conversations she has had this year with experts who see the market from very different angles.
Catch the video at top, listen to the audio edition and read our transcript below, or watch this and future episodes on YouTube.
This podcast is written and hosted by Diana Goovaerts. It is edited by Diana Goovaerts, Matt Rickman and Nikki Grcic. Liz Coyne is our executive producer. Special thanks to guests Joe Constantine, Robert Blumofe, Nik Willetts and Joe Madden.
To learn more about the topics in this episode:
AT&T: Enterprises are starting to show interest in edge compute
Diana Goovaerts, Fierce Network: 10 years ago, telecom operators promised that edge computing would change everything, but it didn't. So why is the industry having the same conversation again, and is anything actually different this time?
Joe Constantine, Ericsson: We had MEC 10 years ago, it didn't go anywhere, I have a lot of respect for the MEC pioneers. They were right about the destination, but they were just too early about the journey.
Robert Blumofe, Akamai: I like the characterization of maybe early but not wrong. Again, you know, early might imply that some notion of failure. Again, I don't characterize edge computing that way. I think it's been a tremendous success. I've said that, "okay, edge probably isn't gonna be as big as core." I stand behind that statement, but it's not that one has to beat the other. They're really more complementary than they are competitive.
Nik Willetts, TM Forum: So what we're concerned about is operators betting the farm on that opportunity and missing the opportunity around evolution of the network.
Diana Goovaerts: I'm Diana Goovaerts, and this is The Five Nine.
This week's episode is actually a little bit personal for me. Throughout this year, I've had dozens of conversations searching for the answer to a question that keeps popping up: Is the edge opportunity really real for telcos, or is the industry just staring down the barrel of MEC 2.0?
This week, I am taking you inside some of those conversations I had with folks who see the market from very different angles.
And to get started, it helps to define what the edge even means, because, like many things in this industry, the edge means very different things to different people.
Joe Constantine: We see it as a continuum, a distributed framework where processing, analysis and storage are moved closer to where data is generated and used. Because what we wanna do, we wanna shorten the data travel, reducing backhaul and also delivering on low latency, high bandwidth and better control over data sovereignty.
We see edge in four areas: the private enterprise edge, the network edge, the extended public edge - co-location with hyperscalers - and obviously last but not least, the device and the gateway edge.
Diana Goovaerts: The next big question is what, if anything, has changed? Mobile edge computing, or MEC, flopped because the use cases requiring it just didn't materialize. So why should it work now? Ericsson's Joe Constantine gave me three reasons.
Joe Constantine: If we look at what are the fundamental changes we see three things.
First, the data scale is vastly different. So between the end of '23 and the end of 2029, in our Ericsson Mobility report, we actually expect global mobile data traffic to triple, and AI will be the key driver of that growth. We also anticipate uplink to grow by 10X by 2035. So routing all this traffic to centralized cloud isn't just only slow, it's like economically not even sustainable.
If we look at the second pillar, the networks are now actually built for time-critical workloads. If you look back at the first MEC plus 4G network, they were built for best effort. Today, 5G has time-critical communication that delivers actually 50 millisecond latency between 99.9 and 99.999% reliability. That's a vastly different network capability than we had before. We have 5G standalone with real-time slicing that delivers quality of service. We have cloud native cores that can be deployed and orchestrated everywhere. And then we have a proper tooling actually to manage these different devices, device, edge and cloud.
Third, why this is real, at least in our view, is we have AI inferencing is the growth opportunity that and an application that MEC did not have at the time.
Diana Goovaerts: Akamai's Robert Blumofe agreed that AI is the shift, and he drew a clear line between training and inference.
Robert Blumofe: A huge amount of the infrastructure has been built up around the training use case, and that's clearly a core centralized use case. You wouldn't use the edge to train a deep learning model of any kind. But as you transition to inference, then it's a whole different story because now you're interacting with people.
Edge is really about computing that is interacting in real time with humans, devices, agents and things like that. They serve a very different purpose. You know, and I've often said, by the way, that when it comes to where should your workload live, you wanna think about, "well, where is the data?" It's easier to move compute to the data than vice versa, as a rule of thumb.
Diana Goovaerts: But not everyone is sold on the edge opportunity. AvidThink founder and principal analyst Roy Chua has argued throughout the year that despite the change AI has brought, the use case just isn't there yet. Anything mission critical will be processed on device or on campus, and anything not can afford the latency to be sent to a more centralized data center facility for processing.
Red Hat's telco chief Fran Heeran told me in May that it doesn't make sense to put intelligence in a mission critical situation anywhere other than on the device itself. And even TM Forum's CEO admitted in June that at least as of right now, the edge opportunity is a niche one, echoing something that AWS's Amir Rao told me in May.
And there are questions about the edge opportunity's long-term viability.
Nik Willetts: From TM Forum's point of view, there is an edge opportunity. It is interesting, but it's quite niche and specialized. Sometimes it's fixed to things like fixed manufacturing. I think the whole debate around physical AI is interesting, but the real question we need to ask ourselves is twofold. One is, can we move fast enough to provide the global localized edge, sovereign edge capability that manufacturers will need?
Because we're not talking about products built for one country at a time. We're talking about wanting total coverage of the globe with these kind of products. Can we do that fast enough, faster than AI chips will evolve to be on the device? And then how much real demand is there? How many real use cases?
Diana Goovaerts: For the sake of argument though, let's say the edge opportunity is real. What implications does that have for compute requirements and the network?
Nik Willetts: We see a number of operators building out small GPU clusters. Most of those are running at sub 20% occupancy right now. Demand may be building, but you're also competing with capabilities from hyperscale and so on, which may not be as edge as we're talking about, but can still be pretty local.
Diana Goovaerts: Ericsson and Akamai believe the edge will be part of a broader network continuum that will serve future AI use cases. While it isn't super prevalent today, our edge optimists pointed to physical AI as the use case that will make edge capabilities critical.
Joe Constantine: We think of AI compute as a continuum, where AI runs seamlessly on a device to radio access network, edge and to core and cloud. So the idea from us is centralize where you can, it's economically viable, but distribute what you must do.
And the discussion we're having is that fabric that we believe will be the foundation for the next generation technology fuses the mobile networks plus AI, AI plus cloud.
If we look at robotics, to your example, where does it sit? Does it sit in a robot or does it sit in a network? I think gonna be both. You would have to have both. I don't think you can do it in one side, but without the other. I don't think the device itself, without any connection to other devices and without any connection to the real world models and context awareness can actually operate in a society.
Robert Blumofe: In that case, you want to extend the continuum all the way onto the device, because in most cases, there's a lot of functionality that you simply cannot leave the device for.
You know, things that are literally mission critical, like the robot moving a joint to stay balanced, that can't leave the device. If it's a car, you know, the actual immediate decision of turning the steering wheel to the left or turning the steering wheel to the right, that's a decision that can't even leave the car.
But other aspects of the decision, you know, maintenance and entertainment, many, many other aspects can leave the device, but there's no point in backhauling them all the way to some centralized location.
Diana Goovaerts: But then you might ask, why would a telco rather than a cloud giant win this business?
Joe Madden, Mobile Experts: I'm very focused on the physical AI and robotics because I see that growing into a trillion-dollar business over time, and a substantial part of that can, can belong to the telcos. A cloud provider such as Amazon or Google, they have the cloud computing resources. They've developed gen AI models. They have huge infrastructure for those things. But they're - as companies, they're unable to provide a latency or a jitter guarantee.
So the deterministic timing that I was talking about in high dexterity robots, that's something that's beyond them. For the wide world and robots that move around, it has to be a wireless connection with a latency guarantee. That's where the cloud computing guys have no options.
Robert Blumofe: You know, the hyperscalers obviously very strong on core, a lot of capabilities there. They don't really have any meaningful edge computing capabilities.
Diana Goovaerts: So is the edge opportunity for telcos really real? After all of these conversations, my answer is simple: it depends on the use case. The one thing everyone seems to agree on is that operators can't afford to bet the farm on a technology without a clear customer and a clear use case, but they also can't really afford to wait.
That's all the time we have for now. Make sure you like and subscribe wherever you are listening or watching, and make sure you stay tuned to fierce-network.com during the week for news and analysis. We will catch you next time.