Access Official Technical PDF

Share Summary:LinkedInEmail

For industrial automation teams, advances in connectivity and computing matter not only as product announcements but as signals of how connected systems are evolving. Qualcomm’s 2020 technology announcements brought together 5G platforms, AI inference hardware and automotive computing. Taken together, these developments illustrate important themes for 5G edge AI: moving data between devices and computing resources, processing AI workloads efficiently, and building platforms that can scale across different applications.

The announcements were not industrial automation products. However, their emphasis on connectivity, edge processing, software and system integration offers useful context for engineers, integrators and operations leaders evaluating the broader technology landscape.

5G edge AI depends on both connectivity and compute

5G is often discussed in terms of faster wireless communication, but connected systems also need processing resources capable of handling the information they generate. The presentation addressed both sides of that equation: mobile and fixed wireless platforms for connectivity, alongside a dedicated accelerator for AI inference in data centers and at the intelligent edge.

Qualcomm reported that, as of late 2019, 45 or more operators had commercially deployed 5G and more than 340 operators were investing in the technology. It also cited projections for 5G smartphone shipments and connections in the years ahead. Those figures were forecasts available at the time, rather than current market measurements. For automation decision-makers, the lasting point is the expected expansion of connected endpoints and the associated need to plan for communications and computing together.

The presentation also described strong interest in 5G and Wi-Fi 6 solutions for fixed wireless access customer premises equipment. It cited 34 CPE original equipment manufacturers using the Snapdragon X55 5G Modem-RF system, with 11 also using Qualcomm’s Networking Pro 1200 platform. This illustrates a platform approach that combines cellular connectivity with local networking, rather than treating wireless access as a standalone component.

AI inference moves into the infrastructure conversation

AI training and AI inference are different workloads. Training develops a model; inference applies a trained model to new inputs. In operational settings, inference is the processing step that can turn sensor, image or other input data into a classification or prediction. Where that processing takes place—in a data center, an edge system or another computing layer—is an architectural decision shaped by the application.

Qualcomm announced its Cloud AI 100 inference accelerator, built using 7 nm process technology. The company specified peak AI performance above 350 TOPS and claimed more than a tenfold increase in performance per watt over the most advanced AI inference solutions then deployed. These are company-reported figures from the announcement, not a general benchmark for all AI systems.

The presentation positioned the accelerator for data center and intelligent-edge applications, including a 5G intelligent edge box. For automation planners, the relevant consideration is the relationship between inference capacity, power consumption and deployment location. Evaluating an edge AI design requires understanding the workload and system requirements; a peak-throughput figure alone does not establish suitability for a particular plant or control application.

Automotive computing offers lessons in system integration

The automotive section showed how connected and intelligent systems combine multiple technologies. The capabilities presented included telematics, digital cockpit systems, advanced driver-assistance systems (ADAS), precise positioning, computer vision, radar, C-V2X and cloud-based device management. C-V2X enables communication involving vehicles and other road users or infrastructure; in the presentation, it appeared alongside cellular connectivity and automotive processing technologies.

This systems view is relevant beyond vehicles. Complex automation architectures also bring together sensors, compute, communications, software and lifecycle management. The automotive examples underscore why integration decisions—such as how components share data and how software is managed—are central to connected-system design. They do not, by themselves, establish that an automotive platform is appropriate for industrial control.

Qualcomm also introduced Snapdragon Ride, a scalable platform concept for ADAS and autonomous driving. The presentation described a family of processors, an autonomous-driving accelerator and a software stack spanning functions such as perception, localization and planning. It stated that the platform was designed for ASIL-D systems, and presented a planned path to deployment on the road in 2023. Those statements describe the company’s announcement and roadmap at the time, not a claim about present availability or certification status.

What automation decision-makers can take from the announcements

For engineers and technical leaders assessing connected automation, several questions follow from these platform trends:

  • Where should inference run? Compare the needs of the application with the processing resources available at the edge and in centralized infrastructure.
  • How will connectivity fit the architecture? Consider the roles of cellular and local wireless networking, and how each connects devices to the systems that use their data.
  • What must be integrated? Account for sensors, compute, software and lifecycle functions as parts of a system, rather than evaluating a processor or network link in isolation.
  • How will requirements be validated? Treat vendor performance figures and roadmaps as inputs to evaluation, not as substitutes for application-specific requirements and testing.

Conclusion: evaluate the whole connected system

Qualcomm’s 2020 announcements highlighted the convergence of 5G connectivity, AI inference hardware and scalable automotive computing. For industrial automation, the practical implication is not that these consumer and automotive platforms can be transferred directly to a plant. It is that connected-system planning increasingly involves the combined design of communication, processing and software. Engineers and decision-makers can use these trends to frame architecture discussions, while grounding selections in their own operational requirements.

Access Official Technical PDF

About The Author