Over the last several years, the exponential growth of IoT devices, the development and rollout of 5G networks, and advancements in AI have all come together to enable a new frontier in computing that we call “the edge.” Though Intel, for its part, has been flying underneath the radar in this sector, through a series of strategic moves, it has stealthily become a player to watch in the edge market. Last month, though, the company turned heads with an aggressive update to its 2nd Xeon Scalable Platform, bringing higher core counts, larger cache sizes, and higher processor frequency to the table. At the same time, the company unveiled a significant expansion of its 5G networking infrastructure offerings, including the new Atom P5900 Intel architecture-based 10nm SoC for wireless base stations.
Today I want to do a deeper dive on Intel’s rise in the network and on-premises edge, and what exactly makes it a force to be reckoned with.
What’s the edge, and what’s fueling it?
The edge, to put simply, is a broad term to refer to all the data endpoints that exist outside of the datacenter. This includes devices such as smartphones, voice assistants and PCs, but also encompasses the vast, and rapidly expanding number of IoT devices and sensors gathering information out in the world. Intel characterizes “the edge” as data centric workload placements outside of the data center versus down to the specific smartphone and end device.
Traditionally, the data gathered from these “edge” devices had to be routed back to the cloud and/or datacenter in order to be mined for insights, but organizations are increasingly looking to perform these functions on physical compute structure, at or near the data source itself—a concept called “edge computing.” Within edge computing, there are a few subcategories, such as the on-premise edge (when a network operator runs virtualized environments across an on-prem, distributed edge architecture), and the network edge (the point where networks connect to the Internet and other third-party networks). Intel views the network Edge/Regional Cloud as a subcategory of the edge.
I believe by eliminating the round trip of data to the cloud and back, edge computing reduces latency and time to actionable insights which can enhance existing services and opening up new business opportunities for customers. Intel views latency and data sovereignty as key drivers to justify placement of workloads at point of highest return, not necessarily.. The growth in edge computing is being fueled by a number of factors, including the ongoing proliferation of connected devices, advances in AI technology capable of edge and on-device processing, and the development and deployment of 5G networks. Edge and 5G, for that matter, work to enable each other—by combining with 4G technology, edge computing can pick up the slack and give early 5G networks a fallback option. Further, it’s the promise of a world of new 5G-enabled use cases that best makes one of the best financial cases for investing in the next-generation network.
Let’s take a look at Intel’s portfolio and how the company continues to build it out to take advantage of this rapidly growing sector. One of Intel’s undeniable strengths is its release cadence—the company is now committed to annual product launches across its portfolio, including the technologies that are key to edge. The company’s 2nd Gen Xeon Scalable Platform, announced last April, was specifically designed for AI, 5G and data processing. Last month, Intel updated Xeon with several new processors optimized for performance and performance-per-dollar for customers spanning the cloud, network and edge.
On networking edge, the company recently launched the Atom P5900, an Intel architecture-based 10nm SoC built for the high-bandwidth and low latency demands of 5G wireless base stations. For that matter, Intel has set the ambitious goal of being the number one provider of silicon for 5G base stations come the year 2021. The company’s Xeon D SoC is also worth noting, designed for the space and power-constrained environments commonly found at the edge.
In terms of AI accelerators, it has the powerful spoils from its recent acquisition of Habana Labs, which it recently announced would be replacing its ill-fated Nervana chips (see my colleague Karl Freund’s take on the news here). In the realm of computer vision, Intel acquired Movidius in 2016. The company’s Myriad VPUs seek to enable on-device, low power deep learning and visual data processing for applications such as security cameras other IoT devices. Intel also has its Arria line of FPGAs and SoCs, designed to deliver high performance at lower power for applications spanning automotive, datacenter, military, communications and more. It’s important pointing out that integrated AI acceleration is available today on Intel Xeon Scalable and that platform is the defacto platform foundation for AI inference today for data centric applications, particularly those that are latency-dependent.
Taken all together, I believe Intel one of the broadest, most holistic portfolio around for enabling edge compute. Out of all the companies that play in the sector, you’d be hard-pressed to find any others who cover as many bases. The only kind of chip it doesn’t offer is a DSP.
Intel & developers
I believe another strength of Intel’s is its developer play, which cuts across visual, retail, industrial, robotics and network, and has committed to quarterly feature releases. Its free OpenVINO developer toolkit focuses on edge AI acceleration. Its OpenNESS toolkit focuses on fostering open collaboration and innovation in the realm of the network and enterprise edge, with the specific goal of enabling 5G development and deployment. Intel’s Open Visual Cloud software stack seeks to do the same in realm of media, graphics, analytics, and immersive media, featuring end-to-end sample pipelines for developers. Intel says Open Visual Cloud is optimized for cloud native deployment and designed for commercial off-the-shelf x86 architecture as well as Intel’s many accelerators.
I also think it’s worth noting that Intel’s software is open-sourced. As I’ve written previously, Intel has invested in many open source initiatives to bring the ecosystem together to accelerate service/app development for edge.
Intel’s Smart Edge MEC acquisition, which flew pretty much under the radar, is a unique piece of software, particularly for enterprises and businesses who are looking to enable edge applications. It’ll be interesting to see where Intel takes this, particularly related to carriers and their services.
Intel’s edge ecosystem
Beyond its broad developer ecosystem working to accelerate edge innovation, Intel has also amassed an impressive roster of customers and partners for its edge solutions. The company says it is doubling its edge deployments year over year. For the network edge, partners include Lenovo, Red Hat, Advantech, Caswell, Inventec, Lanner, Nexcom, World Wide Technology, Ericsson, ZTE, Verizon, Vodafone, Nokia, China Telecom and many more. For on-prem edge computing, Intel is working with the likes of Foxconn and QNAP. There are some big names in that list, and I expect there will be more come this time next year.
It’s important to note that Intel has also established a builders programs with hundreds of ecosystem partners. You can find more in that here.
All of this to say that Intel is now one of the best-positioned vendors for the on-premise and network edge. Through strategic acquisitions and internal development, the company has amassed an impressive portfolio with processors, accelerators, SoCs, VPUs, and FPGAs geared specifically towards compute at the networking and on-prem edge. Furthermore, through its open source toolkits, the company is fostering an impressive developer and partner community to leverage this technology.
I believe Intel has a lot of market opportunity in these areas, and it seems to taking the right steps to take advantage of it. That market opportunity involves not only enterprises, but CSPs (Cloud Service Providers) that are driving disruption in the market by approaching edge with new business models, and some are doing it in collaboration with CoSPs. AWS Wavelength and Verizon are good examples.
The days of Intel flying under the radar in edge compute are likely over.
Note: Moor Insights & Strategy writers and editors may have contributed to this article.