Will Edge Computing Pay Off Big for Mobile Operators?

Will mobile operator interest in multi access edge computing drive less revenue and strategic upside than hoped for? That would not be an unusual outcome for telco initiatives generally, but especially when considering moves outside their core connectivity services domain.

In principle, edge computing as a service could allow mobile operators to assume new roles in the computing ecosystem, in addition to the connectivity role. Some mobile operators have tried  to operate their own neutral host data centers, for example. Some still do. 

AT&T briefly tried to become a computing supplier in the 1980s with its NCR acquisition. 

At least for the moment, tier-one telcos have eschewed the effort and risk of becoming "edge computing as a service" providers and instead are banking on advantages related to the core connectivity role and some incremental revenue or value as providers of real estate (racks, power, cooling, security). 

source: STL Partners

While clearly less risky and capital consumptive, such partnerships might not drive large amounts of incremental revenue, though, as always, edge computing roles could support their connectivity revenues, and create some amount of real estate incremental revenue.

One might argue that is about the best they can hope for, at the moment. 

Though other roles might eventually emerge, the immediate tack seems to include an acknowledgement that the opportunity to become “edge computing as a service” suppliers is highly limited, if possible at all. 

5G "Needs" Edge Computing

The value, strategic or revenue potential of 5G is contingent on developments in related areas including edge computing, internet of things and machine learning. The reason is that the value of edge computing is the enablement of new use cases and applications built on the assumption of ultra-low-latency access to computing resources, often with simultaneous ultra-high bandwidth.

In other cases, 5G networks themselves will require edge computing support to operate the core 5G transport and access networks. So mobile operators will be big and early users of edge computing, for internal purposes.

Also, consider edge computing and 5G as used to support apps on consumer mobile devices. Most of the new use cases and revenue drivers depend on connectivity, but are not directly “owned” by connectivity providers. That being the case, the "value" of 5G is not as great as some think. Most consumer 5G users of smartphones will simply be exchanging a 4G subscription for a 5G account. The amount of incremental revenue will be nil, in many cases.

source: State of the Edge 202

By 2028, 4G and 5G mobile consumer and residential consumer applications will dominate the edge computing footprint, for example, says the State of the Edge 2020 report. A growing number of those apps would not be consumable without edge computing plus the latency performance and bandwidth made possible by 5G. 

source: State of the Edge 202

By about 2028, only eight percent of edge computing resources will be used to support retail public network voice or messaging. Gaming along might represent 18 percent of edge computing resource consumption. 

The edge computing power footprint for mobile and residential consumers will reach 16938 megawatts and 10843 MW, respectively by 2028, the report says. 

Edge computing power footprints for service providers and enterprise IT is expected to increase 7117 MW and 5800 MW, respectively. 

source: State of the Edge 2020

Mobile operators especially will be significant users of edge computing to support their own internal operations, including their virtualized 5G core and access network facilities. 

source: State of the Edge 2020

How Long Before Artificial Intelligence Hits an Inflection Point and is Adopted on a Mass Scale?

It would be hard--perhaps literally impossible--to find any forecasts of 5G or artificial intelligence adoption that do not slope upwards and to the right. What often matters quite a deal to ecosystem stakeholders, though, is how many years elapse before the curve hits some clear inflection point. 

The reason is obvious: firms want to create products for the market before it hits the inflection point, and not “too early or too late.” Perhaps 5G is not so much the issue, as deployments already have begun, so charting progress is fairly easy.

source: ABI Research

The various forms of AI are harder, in part because AI is a capability that will be incorporated into many products. Customers will not be “buying AI” so much as some other solution to a business problem. And for all the hype, few business managers yet report they are using AI. 

A study by the U.S. Census Bureau, for example, finds that firms reporting they already use AI in any form (often machine learning) are less than three percent of respondents. 

A rule of thumb about technology adoption therefore suggests we are about seven percentage points of adoption away from an inflection point virtually everyone expects will happen. So why seven points?

source: ABI Research

Most useful advanced technologies tend not to go mainstream until adoption reaches about 10 percent. That is where the inflection point tends to occur. That essentially represents adoption by innovators and early adopters. 

source: LikeFolio

One often sees charts that suggest popular and important technology innovations are adopted quite quickly. That is almost always an exaggeration. The issue is where to start the clock running: at the point of invention or at the point of commercial introduction? Starting from invention, adoption takes quite some time to reach 10 percent adoption, even if it later seems as though it happened faster. 

source: Researchgate

Consider mobile phone use. On a global basis, it took more than 20 years for usage to reach close to 10 percent of people. 

source: Quora

That is worth keeping in mind when thinking about, or trying to predict, advanced technology adoption. It usually takes longer than one believes for any important and useful innovation to reach 10-percent adoption. 

source: MIT Technology Review

The bottom line is that if the classical curves hold, the inflection point for AI adoption will happen when surveys report that 10 percent of respondents are aware they have bought and are using AI solutions.

AWS Wavelength Live on Verizon 5G Network in Boston and Bay Area

 AWS Wavelength now is live for developers in Boston and the Bay Area, embedded into Verizon’s 5G network. 

AWS Wavelength embeds AWS compute and storage services at Verizon 5G edge locations, allowing Wavelength apps to run without leaving the Verizon 5G network, reducing latency.

What is not clear is how the business model works. Users of Wavelength must be AWS customers first. 

“With just an AWS account, you can deploy your 5G applications in Wavelength Zones and seamlessly connect to applications and services in AWS Regions,” Verizon says. “Simply log-in to the AWS Management Console, request access at https://aws.amazon.com/wavelength/, and enable the Wavelength Zones you want to use for your account.”

Verizon might be content simply to drive AWS traffic onto its 5G network, with no direct revenue contribution. AWS might pay Verizon interconnection or some other similar fee for the right to domicile on the 5G network (leasing real estate, as would be the case in a data center). It seems unlikely AWS would do revenue sharing. 

The parties seem to have said nothing about the business arrangement in public.

The Only Certain Winners in Edge Computing are Server Suppliers

Firms Chase New Edge Computing Roles and Revenues

Many possible roles and business models exist for edge computing: on the device, on the enterprise premises or at some other location in a metro area, supplied by a hyperscale computing as a service supplier, a connectivity provider, directly by enterprises as a private network or by other third parties providing neutral host computing. 

To the extent that content delivery networks were an early form of edge computing, Akamai can claim it has the largest edge computing business, at the moment. “The Akamai Intelligent Edge

Platform has grown to include over 300,000 servers in over 4,000 locations and nearly 1,500 network partners,” the company says. 

“We’re the largest provider of edge computing services by far,” says Tom Leighton, company CEO. 

Separately, Cloudflare, which initially provided content delivery network services, now is branching out into edge computing services, working with Vapor IO and EdgeMicro for the local data center real estate. 

Also, Vodafone Business is building a private private 5G network for Centrica Storage. That is not, strictly speaking,  an “edge computing” deployment, but supports Cenrica’s own enterprise edge computing capabilities. 

The three developments illustrate the different potential business models for edge computing (content delivery) and infrastructure edge computing, or multi-access edge computing services offered by connectivity providers. Akamai’s CDN services are a form of edge computing. 

Cloudflare is branching out from CDN services to cloud computing as a service. Vodafone likely is acting as a system integrator. 

Then there is Amazon Web Services Outposts, a fully managed service that extends AWS infrastructure, AWS services, application programmer interfaces (APIs), and tools to virtually any data center, colocation space, or on-premises facility.

Outposts is one example of how some parts of the edge computing business might not require use of edge computing facilities owned or managed by telcos or other edge computing providers offering neutral host computing and storage.

source: Rackspace

Instead, AWS Outposts essentially is one way to support hybrid computing, with AWS edge computing inside enterprise data centers. AWS also will support other forms of edge computing that put the actual servers at some location outside the enterprise but possibly within a metro area, in partnership with some telcos, for example. 

In those cases the computing or storage as a service is supplied by AWS, the colocation facilities by a connectivity service provider. Outposts is not designed for small retail customers but for enterprise data centers, as is clear from the data center requirements. 

In other cases, new entities creating edge computing facilities within a single metro area to support their own operations might--as AWS has done--also allow third party use of services at those facilities. Retailer Walmart provides an example.

AWS Outposts is for Hybrid Computing

Amazon Web Services Outposts is a fully managed service that extends AWS infrastructure, AWS services, application programmer interfaces (APIs), and tools to virtually any data center, colocation space, or on-premises facility.

It is one example of how some parts of the edge computing business might not require use of edge computing facilities owned or managed by telcos or other edge computing providers offering neutral host computing and storage.

source: Rackspace

Instead, AWS Outposts essentially is one way to support hybrid computing, with AWS edge computing inside enterprise data centers. AWS also will support other forms of edge computing that put the actual servers at some location outside the enterprise but possibly within a metro area, in partnership with some telcos, for example. 

In those cases the computing or storage as a service is supplied by AWS, the colocation facilities by a connectivity service provider. Outposts is not designed for small retail customers but for enterprise data centers, as is clear from the data center requirements. 

Who Wins Edge Computing?

The vision advanced by proponents of multi-access edge computing includes new roles for a variety of potential ecosystem providers, ranging from hyperscale computing as a service providers to data center operators; connectivity service providers to neutral host access providers. 

All sorts of hardware and software infrastructure providers also are fundamental in any scenario, to supply central processing units, memory, artificial intelligence or analytical capabilities, racks and other associated gear. 

The range of potential data center providers also includes some non-traditional providers such as very large retailers who might allow third parties to use edge computing capabilities originally developed to support internal operations. 

source: STL Partners

The issue for all potential providers is “how big” the edge computing opportunity might be as a driver of new revenue, which roles make most sense and are most feasible. 

At a high level, the domains include computing hardware and software (CPUs, memory, racks, power supplies); edge facilities (real estate); turned up computing as a service, storage as a service or platform as a service. 

“Platform as a service” often is hardest to comprehend, as most observers understand the value and role of CPU/memory (core compute); real estate (data centers) or fully hosted software as a service. Of course, that always is the case whenever choices are made somewhere between “do it all yourself” and “buy a fully-managed service.”

source: Red Hat

At this point, it is too early to predict which current participants in the ecosystem will emerge in new roles in edge computing, and to what extent. Nor is it clear how much additional value or revenue edge computing will drive for various suppliers in various roles. 

It already seems clear that, ultimately, edge computing is still about “computing,” and that means significant upside for hyperscalers and some data center providers. How much “platform” operations will be important for connectivity providers is not yet clear, but some of us who have watched telco innovation efforts over decades might say tempered expectations are realistic.

Telcos have not generally been highly successful at entering new ecosystem roles in computing or applications. That might be even more true in an era of open source, open networks (the internet) and virtualized platforms and networks where scale economics prevail. 

In a closed environment, telcos might have exercised more control. That is largely impossible when “connectivity” is functionally separated from all the other platform functions (storage, servers, operating systems, middleware and runtime environments as well as from applications, data ownership and end-user-facing business models. 

MEC might be important for connectivity service providers, at least to support their own virtualized core network and access network operations. How much incremental value MEC can provide others in the ecosystem has to be determined. That, in turn, will determine how big a revenue generator MEC might be, for connectivity service providers.