Enterprise and Telecom Spending Will Drive Edge Deployment Despite Hyperscaler Head Start

When considering edge computing, it is impossible to ignore the headway being made by hyperscalers in bringing innovative solutions and making them available in selected areas. Looking at the progress being made in North America, the three major hyperscalers all have made significant investments: Amazon Web Services (AWS) has brought Local Zones to twenty-one US cities and has plans to roll out to eleven European cities by the end of this year, Microsoft has plans to deploy two Azure Edge Zones in the USA this year, and Google has deployed their network edge solutions at over 140 locations globally. The global edge roll-out for all hyperscalers is under way. Most of the investments in hardware that have been made by the hyperscalers have been deployed into metro city areas with a population over three million, which ensures there is ample opportunity to recoup a healthy return on those investments—in other words, they are harvesting the low hanging fruit that is presented by these wealthy metro populations and their associated industries. Perhaps with the exception of Google’s cloud platform, the size of a typical hyperscaler edge deployment is relatively large, with estimates ranging upwards from 2,500 servers per deployment center. This is an expensive model that is unlikely to be used outside of wealthy major population centers whose usage of the facility will justify the spend. A recent study performed by ABI Research indicates that hyperscaler spending on edge computing servers will grow modestly over the coming years but the large exponential growth in server shipments to be used to provide edge computing services will be seen in alternative sectors.

The use cases demanding low latency compute capabilities that process data at or near the point of creation in near real-time will not be satisfied cost-effectively by deploying multiple, large, de-centralized hyperscaler backbone extensions. Instead, these demands will be met by deploying a network of smaller, more agile, yet powerful compute environments that are well connected, or they will be met by the enterprise themselves, deploying solutions on site that will meet their needs. It is almost certain that many of these deployments will have a hyperscaler element to them, either a software element such as Google’s Anthos or hardware such as AWS Wavelength Zones, but crucially the hyperscalers will need to collaborate with carriers, enterprises, and neutral hosts to deploy those services. Carrier and enterprise spending will dominate infrastructure investments at the edge.

The hyperscalers may yet retain the upper hand since they have demonstrated their ability to roll out edge services. They have already captured developer interest in multiple verticals and applications that benefit from operating at the edge that are being written and deployed using hyperscaler ecosystems. In keeping with hyperscaler deployment strategies, they can leverage economies of scale and oversubscription models to deliver edge servers at minimal cost, either through partnerships or by commoditized or self-integrated infrastructure. The result is lower overall costs for the hyperscalers, meaning that their deployments will be described by more favorable economics and faster rollouts.

The important factors that will shape the dynamics of upcoming edge deployment will be demand and the ability to effectively meet that demand. So, whilst the bulk of edge infrastructure spend will be made by the two groups that have the real estate to supply effective solutions, the enterprise itself and carriers, the spend split is less well understood and the crucial factor that will decide this split is data security.

Where the enterprise has obligations to ensure the end-to-end security of the data it processes at the edge, it is far simpler, if not more expensive, to deploy dedicated solutions on site. If the enterprise can use a public edge facility whilst fully meeting its data handling obligations, then public edge services will be the cheaper and easier option for the enterprise to take.

If the carriers and their technology partners can offer public edge services that meet the strict data handling requirements imposed on the enterprise by their governing bodies and national regulators, then they will steal a large slice of the edge market away from the enterprises who do not necessarily want to shoulder the cost or responsibility of providing their own edge services.

Whilst the use cases that the enterprise will bring to the edge are not fully known, the regulations that control data they will be processing are known, and where they are not known implicitly, they are in principle. ABI Research firmly believes that if the carriers and their technology partners can solve these data privacy problems before the enterprises start to elaborate on their edge use cases, then they will gain a much larger share of the market and gain a much more significant return on their investments.