Toward A 5G, AI-Centric World

A look at the technology that will underpin the next big advances.


The market environment over the coming years will continue to experience an explosion of data generation from a variety of new sources such as smart cars, smart factories, smart hospitals and smart network infrastructure. Data explosion combined with artificial intelligence (AI) will create a renaissance of computing and storage hardware. Mobile World Congress 2018 (MWC) reinforced this market thesis through a clear emphasis on the 5G network, both as a key enabler of data generation as well as a catalyst for transformation of the traditional network infrastructure to a smart network infrastructure. MWC also highlighted progress in mobile AI and augmented reality/virtual reality (AR/VR).

5G is a Key Enabler of Data Generation. Higher communication speeds as well as lower latencies of access offered by 5G will be increasingly important in a world with a growing number of smart Internet of Things (IoT) devices. Each smart car for instance is expected to generate 4TB of data per day. With today’s communication infrastructure it is difficult to imagine even a fraction of such data being transmitted to the Cloud to support learning and analysis.

Major U.S.-centric carriers as well as international carriers announced or reiterated 5G trials and expansion plans at MWC. Here are a few key updates:

  • AT&T expects to provide 5G services in 12 cities by the end of CY18
  • Sprint is preparing six cities for 5G deployments by 2019
  • T-Mobile plans to build out 5G in 30 cities by the end of CY18
  • Verizon is planning to launch 5G in up to five cities by the end of CY18
  • China Mobile expects to start running 5G trials in five cities as early as 2Q18
  • Swisscom is preparing to introduce 5G services at selected sites by the end of 2018

Meanwhile, major international chipmakers including Qualcomm, Intel, HiSilicon and Mediatek reported progress on their 5G chipsets, and network equipment makers Nokia and Ericsson emphasized their readiness for 5G. While industry watchers are mixed in their enthusiasm for 5G, we expect to see meaningful deployments of 5G around the world by 2020, and 5G rollouts to accelerate from there with validation of new use cases.

5G is a Catalyst for Smart Network Infrastructure. High-frequency 5G signals transmit at a shorter range, so they require a distributed network architecture with a dramatically higher number of wireless endpoints than 3G/4G to enable higher bandwidth accessibility for mobile devices. This could accelerate transformation of wireless infrastructure from its traditional heritage in macro base stations to deployment of small cells and smart antenna solutions, both of which would leverage a Cloud-based Radio Access Network (C-RAN). In addition to typical scale benefits achieved through migration to the Cloud, the application of AI (i.e. machine learning, deep learning) to intelligently manage a vastly more complex network infrastructure could generate additional cost benefits for network operators, as well as potentially unlock new revenue streams.

Progress in Mobile AI and AR/VR. Samsung launched its second-generation mobile assistant (Bixby) and its new AR emojis with its new Galaxy S9 smartphone. Arm showed demos of Project Trillium, which is a mobile AI platform for machine learning in Edge devices. Google announced general availability of ARCore 1.0, which allows developers to write AR apps for smartphones. And, Microsoft recently announced Windows Mixed Reality, which allows developers to mix physical and digital content. These developments point to increased adoption of AR/VR.

At Applied Materials, we believe advanced materials engineering holds the keys to unlocking commercial value from AI and IoT, of which 5G is a key enabler. Materials innovation will make possible a broader range of new processor and memory chips optimized for different types of new workloads across data center, network infrastructure and IoT devices. Such a hardware renaissance is necessary because traditional CPU-, DRAM- and NAND-based computing architectures may not be performance-, energy- or cost-optimal for next-generation workloads.

Leave a Reply

(Note: This name will be displayed publicly)