Edge computing is on the rise. Telecom operators should seize the opportunity to compete in the next stage of the cloud infrastructure market
With double-digit growth rates, the cloud infrastructure and data center (DC) market continues to deliver outstanding performance. Leading players in this market enjoy valuations with multiples of 20 times EBITDA and beyond, far higher than telecom operators. Initially, operators had big ambitions in this market. But given the scale and capital expenditure necessary to compete successfully, operators such as AT&T, Verizon, Lumen, and Telecom Italia, have curbed their ambitions or even exited the cloud DC market completely over the past years. Today, they find themselves serving commoditized pipes carrying data that is monetized in ecosystems beyond their reach. But how does edge computing now change the odds back in favor of these telecom operators?
The Linux Foundation defines edge computing as ”the delivery of computing capabilities to the logical extremes of a network in order to improve the performance, operating cost and reliability of applications and services.” With such a broad definition, it is clear that edge computing manifests itself across a spectrum of implementation options. Figure 1 shows an overarching view of the continuum between edge computing and central cloud computing, including the characteristics of both the devices and DCs.
Figure 1: Edge computing characteristics. Source: Strategy& analysis
Edge computing offers a viable opportunity for telecom operators to re-enter the cloud DC market and benefit from its growth potential. Market researchers expect growth rates for the cloud business in excess of 15% per annum. For edge cloud services, CAGRs are even higher, admittedly starting from a low base. (Allied Market Research, for example, has projected a 33% CAGR for edge cloud services for 2020 to 2025).
To succeed, telecom operators need to exploit the advantages they hold over today’s cloud DC leaders, such as Equinix, Digital Realty and the cloud hyperscalers. They already possess the necessary capabilities and assets for a highly distributed compute infrastructure at the edge of their networks. They can deploy the edge IT equipment at their mobile sites or local exchanges, and use their technical and management capabilities to operate and maintain it efficiently.
While telecom executives are committed to investing in 5G radio access network technology, they don’t appear so sure about edge computing. We believe, however, that edge computing would greatly increase the return on 5G investment. This is because it enables attractive monetization schemes that are closely connected to 5G features, such as network slicing for guaranteed latency and throughput. These features will be in high demand as they facilitate high-performance and mission-critical use cases of edge computing.
Initial use cases for edge computing include real-time video analysis in the business-to-business segment, and high-performance cloud gaming for the consumer segment. Such use cases generate the need for powerful computing resources near to mobile end devices which are designed to be simple and lightweight, and which cannot process heavy compute tasks. These use cases rely on compute-intensive analytics of locally generated or contextualized data to deliver real-time insights and compelling visual scenes. They are by definition incompatible with communication round-trip latencies in the 50-100 millisecond range that would be expected for transporting large data masses across the network to reach the next hyperscale DC.
Because of the edge computing architecture, such use cases can still take advantage of the known benefits of cloud services, particularly elasticity and flexibility. The variety of edge computing use cases can be best understood in terms of their common must-have requirements that drive the need for an edge computing architecture. While such archetypes of edge computing use cases do not depend on the individual maturity and commercial opportunity of the use cases, they share certain characteristics, such as mobility across defined or unknown areas and integration with disperse sensor-actuator fleets. Figure 2 summarizes these archetypes.
Figure 2: Edge Computing Archetypes and Characteristic Requirements. Source: Strategy& Analysis
Distributed resources for compute and storage are housed in edge computing data centers. We have observed a continual migration of compute and storage resources from the central tier 1 data center locations, close to the large internet exchanges, towards more regional tier 2 and tier 3 locations. To get even closer to the end devices, we expect this trend to continue traveling through the levels of the network architecture until the point is reached where increasing costs outweigh the benefits. Figure 3 shows the range of potential locations for edge computing data centers in relation to the telecommunication network.
Clearly, when the computing and storage resources are more distributed, the set-up investment and cost of operations increase dramatically. Telecom operators, given their customary heavy investment in the installation and maintenance of network equipment, are well acquainted with such outcomes. In contrast, hyperscale cloud providers and large-scale colocation providers have yet to become confident at managing distributed asset deployments at such scale.
Figure 3: Edge Locations: Options for Telcos. Source: Strategy& analysis
Hyperscale cloud providers have already started to place some of their resources in regional DCs. At the same time, they have started to offer solutions for placement in their telecom partners’ network infrastructure.
Nevertheless, telecom operators are well placed to tap into the edge computing opportunity independently. While the hyperscalers bring state-of-the-art technology architecture and a powerful IaaS/PaaS/SaaS portfolio to the edge computing ecosystem, it remains to be seen which types of businesses will emerge in response to the hyperscalers’ activities. The strong reliance of many use cases on 5G connectivity and monetizable features, such as network slicing for guaranteed quality of service, put the telecom operator in an excellent position to exploit their customer access and offer combined connectivity and edge computing deals.
A telecom operator’s own edge computing offering must be closely linked to the network services, and ideally rests on an independent architecture controlled by the operator and its technology partners. Moreover, partnerships with technology players make it possible to hit the ground running and participate early in this growing market.
Among the possible available locations for edge computing data centers, telecom operators should make use of their existing real estate assets at the locations in the core network, network aggregation points and access network. The same applies when offering colocation sites to edge computing providers. While it is still necessary to re-architect many of such real estate assets in order to host additional equipment, they provide an unrivaled head start in the race to the edge.
The edge computing ecosystem offers telecom operators different ways to position themselves. Depending on their business focus, they may define their play as:
Any of these plays will entail a capability to service either owned or partner equipment across a large number of distributed sites. This would include security and smart hands services. In order to provide edge IaaS and PaaS services, the telecom operator will need to strengthen their cloud development and operations capabilities. These capabilities were not needed until recently, but are now in high demand within the sector.
Moreover, a successful telecom edge computing play needs a robust partnership strategy. This would cover technology partners, go-to-market partners, and even hyperscale partners, to complement any of their own offerings and gain early traction in the market.
We expect most telecom operators to opt for a two-tiered edge play. First, forge close ties with hyperscalers and provide the scarce, highly connected edge space to accommodate the hyperscaler ecosystem. Second, partner with technology players to build and market telecom operator’s own edge IaaS/PaaS to platform players, e.g., in the SaaS or content space aiming to disintermediate hyperscalers.
Several telecom operators are making strides in this direction, and we are looking forward to seeing even more edge initiatives emerge.