Raymond Liao

7 October 2020

7 min read

The amount of data in the world is increasing at a rate well beyond the capability of centralized computing resources to handle, including even enormous hyper-scale clouds. Performant systems must therefore have their computing resources gravitate toward where the data is being generated out in the field in the real world.

To solve these problems, edge computing is pushing infrastructure -- i.e., compute, storage, security, and networking -- into physical locations that are closer to end users and their devices. At Samsung Next, we are looking for leading open source, microservices, and cloud-native development startups serving customers in both the enterprise and network edge segments.

State of the market

Edge computing is an emerging field that is being pursued across three giant industries: enterprise IT (a $4T market), telecom providers (a $1Tn+ market), and public cloud infrastructure (a $300B market). As a result, we believe edge computing at minimum is a $1B market opportunity but could grow beyond that while leading the next generation of enterprise IT.

To date the cloud computing market has grown to more than $200 billion, but according to Gartner it still only represents 5.6 percent of total IT spending. For reasons that include data strategy, cost, and privacy, most of the remaining enterprise IT workloads could end up at the edge instead of in the cloud.

Even though the edge computing market is huge, we believe it is becoming increasingly accessible to early-stage startups. There are several reasons for this:

  • Developer-first go-to-market strategies have become a valid approach given the success of AWS and our portfolio company Packet, who was acquired by Equinix early 2020. This is actually an unfair advantage that startups have against legacy incumbents who have been focusing on top-down enterprise sales.
  • Large telcos and cloud providers are increasingly emphasizing open APIs and contributing to open source projects in an effort to avoid OEM lock-in. This trend breaks down the barriers of entry to domain knowledge and system stack.
  • Finally, the adoption of microservices and “loosely coupled” system architectures give smaller players a better chance of being integrated into production systems of large incumbents.

We believe there will be a significant need for rearchitecting infrastructure to enable applications that require faster time to deployment, lower cost of ownership, and lower latencies. There will need to be faster networks and servers in closer proximity, as well as new approaches, new equipment, and new architectures deployed at the middle and last mile. The market is still early, fully of dynamics of challenges and opportunities.

Component-Edge1b
Market Segmentation

We segment the edge computing market from device to cloud into four key categories: Device Edge, Enterprise Edge, Network Edge, and Public Cloud.

Enterprise Edge

Enterprise IT is migrating to the cloud, but it represents only about 6 percent of spend so far. Enterprises want to benefit from cloud computing as part of a DevOps drive for efficiency and cost reduction, as well as cloud-native application development for business agility.

However, many enterprises are still worried about losing control of their data and have security or privacy concerns. Edge computing can be complementary to hyperscale data centers, as workloads are migrating to the cloud and those workloads require infrastructure edge functionality. Edge infrastructure CAPEX to support enterprise IT is forecast to grow dramatically as enterprise IT cloud workloads begin to be deployed at the infrastructure edge.

According to Gartner, more than 90 percent of enterprises have begun exploring single-use cases for edge computing, and more than 50 percent of enterprise generated data will be created and processed outside the data center or cloud by 2022. By the end of 2021, more than 50 percent of large enterprises are expected to deploy at least one edge computing use case versus less than 5 percent in 2019.

Enterprise Edge is a giant market ranging from large data center opportunities like the converged storage and networking infrastructure, to industrial IoT applications in various verticals. COVID-19 and the remote work trend will further accelerate the IT transformation in enterprises.

Network Edge

In the context of communications service providers (CSP), the Network Edge consists of a hierarchical structure of a vast access network of remote communications and computing nodes, as well as a more concentrated core network with higher capacity networking and computation.

An access network is a primary market for edge computing because it consists of compute platforms that are collocated with access sites to house network access equipment, such as cellular radio base stations, xDSL, and xPON access sites. A core network consists of regional data centers and central offices where access controller, switching equipment, and other service gateway functionality is commonly deployed.

CSPs have been aggressively pushing for the decoupling of traditionally vertically-integrated solutions (i.e., control and user plane separation), and are moving toward virtualization and software-defined networks. The goal is to reduce CAPEX with white box hardware powered by merchant silicon (e.g., SmartNICs and switching ASICs) and OPEX with cloud native operations..

Given heavy investment and an urgency to develop alternative OEM suppliers, the introduction of 5G makes the problem more acute for the telco industry. In other words, 5G is not just about delivering more wireless bandwidth, but also about embracing open source, open APIs, and cloud-native operations.

Device Edge

The Device Edge consists of devices and servers on the “downstream” side of the last-mile network, including self-contained end-point devices such as smartphones, wearables, and automobiles; gateway devices such as IoT aggregators, switching and routing devices; and on-premise server platforms.

During the last decade, we have seen IT vendors championing IoT fog computing while the home appliance industry pushes for smart home solutions. So far, any progress made has been modest and focused mostly on device interoperability within one vendor, while local cloud has been used for mostly photo or video storage and sharing.

The market is still hindered by cross-vendor system integration challenges at upper protocol layers and applications. Some progress from the Enterprise Edge could address these challenges and deliver a more vibrant developer ecosystem.

In the near- to mid-term we expect 5G “killer apps” like AR/VR, automotive, network gaming, and consumer IoT to become more mature and deliver success for the Device Edge.

Public Cloud

Hyperscale cloud providers have recognized the opportunity above as well. They are aggressively rolling out edge products and services to reach all the above segments.

Areas of opportunity

Cloud computing has enabled scale, innovation, connection, and agility at the backbone internet back end but edge computing could complement cloud computing by providing more real-time value, data production, and intelligence at the access networks.

For entrepreneurs, we believe their unfair advantages against large incumbents will come from: open source projects, cloud native developer-first ecosystem, and SaaS apps. These are mapped into three corresponding layers: the data plane (aka the user plane in 5G term), the control plane, and the application plane.

In the data plane, because of the availability of high-performance merchant silicons and their SDKs, open source projects are not handicaped anymore against traditional networking OEMs. There are more rooms to innovate in routing, network security, and application performance for both the Enterprise and Networks Edge markets.

In addition, there is also a data privacy opportunity because the data does not always have to be moved to a central party who might not be trusted. Further, by moving computing to the physical location of data, edge computing can enable compliance with data provenance regulations like GDPR.

In the control plane, because of virtualization, small vendors can play a big role in a large production network using AI/ML for orchestration, service automation and analytics. The adoption of cloud-native development also poses existing opportunities for startups on monitoring, security and management of ephemeral microservice sessions. Here we see Enterprise Edge will be early adopters, the successful startups will then find the carriers as huge customers later.

In the application plane, carriers count on 5G to expand beyond just being “dumb pipes” by offering new services on top of 5G like autonomous vehicles, AR/VR, and network gaming. This is the massive consumer adoption success case for both Device Edge and Network Edge, but it will take a few years to materialize after the infrastructure deployment.