Industrial edge computing is part of a broader shift toward systems that process, store, analyze and transfer data across connected environments. Intel’s 2018 annual report describes a transition from a business centered on PCs and servers toward data-centric computing, with products aimed at applications spanning embedded systems, networks, data centers and the cloud. For industrial automation teams, the strategic themes are relevant: computing needs vary by workload, infrastructure must scale across locations, and hardware capabilities depend on effective software and connectivity.
Why industrial edge computing calls for a broader architecture
Manufacturing, energy, retail and other industrial markets generate data that can be used in applications ranging from embedded computing to analytics. The report identifies these sectors among the markets served by Intel’s Internet of Things Group, alongside high-performance computing solutions for targeted verticals. It also frames customer needs around processing, analyzing, storing and transferring data—activities that extend beyond any single processor or computing location.
That perspective supports an edge-to-cloud view of industrial infrastructure. Computing resources may be needed close to equipment and applications, across communications networks, and in centralized data centers. A system integrator evaluating industrial edge computing therefore needs to consider how workloads and data move across these layers, rather than treating the edge device as an isolated endpoint.
The report presents an expanded portfolio intended to address this range: CPUs, memory and storage products, networking and connectivity technologies, programmable devices, and software. It describes solutions that scale from the edge to the network and cloud, while recognizing that customer requirements differ by market and workload.
Match compute architecture to the workload
Industrial applications do not all place the same demands on compute hardware. Intel’s report groups its development around four computing architectures: scalar CPUs, vector GPUs, matrix-oriented AI accelerators and spatial FPGAs. The practical implication is not that one architecture replaces the others, but that different kinds of computation may call for different processing approaches.
For automation engineers, this is a useful framework when assessing applications such as control-adjacent data processing, machine vision, analytics or AI inference. The report describes CPU, graphics, FPGA and AI accelerator products as part of a portfolio intended to span AI training and inference. It also points to programmable solutions, including FPGAs, as a way to manage changing demands in next-generation data centers and accelerate emerging applications. Specific hardware selection still depends on the requirements of the application and system design.
Memory, storage and interconnect affect the data path
Processing capability is only one part of an industrial data system. Data must also be stored and moved between components. The report describes memory and storage technologies intended to address different points in the data hierarchy. Intel Optane DC persistent memory, as characterized in the 2018 document, combined memory-like performance with greater capacity and persistence associated with storage. The stated aim was to bring more data closer to the CPU for workloads involving large data sets, such as AI and large databases.
The report also discusses 3D NAND and solid-state drives as part of its storage portfolio, alongside interconnect technologies. Silicon photonics is presented as a way to integrate lasers into silicon for high-speed optical connections that could help address data-center networking bottlenecks. These examples show why system planning should account for the whole data path: the ability to process information is linked to how quickly it can be accessed, stored and communicated.
In industrial environments, that means considering data movement between equipment, edge computing resources and wider infrastructure as part of architecture design. The source does not prescribe a particular factory network or topology, but it emphasizes connectivity and interconnects as core technology areas alongside compute and memory.
Software and security are part of the platform
Hardware capabilities do not automatically translate into usable application performance. Intel’s report states that software plays a critical role in unlocking hardware performance and describes work toward unified programming tools across CPU, GPU, FPGA and accelerator platforms. For system integrators, software compatibility and development effort are therefore important considerations when combining different compute resources.
The report also identifies embedded security features as one of six engineering pillars, and describes security as integrated into the product design process. Industrial decision-makers should treat security as an architectural concern alongside processing, storage and connectivity. The source does not provide a detailed security framework, but it makes clear that the company viewed security as an ongoing product and technology priority.
Operational discipline matters behind the technology
Industrial computing depends not only on product design but also on the ability to supply and manufacture components. Intel describes itself as an integrated device manufacturer, producing many products in its own facilities while also using third-party foundries and subcontractors for certain components and activities. Its 2018 report notes that stronger-than-expected demand put pressure on factory capacity, requiring production prioritization and additional investment.
This is a reminder for technical decision-makers to include supply and lifecycle considerations in planning. Product roadmaps, availability, manufacturing capacity and sourcing arrangements can affect deployment schedules. The report also links operational excellence with simplifying and automating routine processes, an approach intended to support efficient core operations and investment in new products.
Practical implications for automation teams
The 2018 strategy offers several useful considerations when planning industrial edge computing:
- Define workloads first. Identify what must be processed at the edge and what can be handled elsewhere before selecting compute hardware.
- Plan across the data path. Evaluate processing, memory, storage and connectivity together, including how data moves between edge devices and centralized infrastructure.
- Assess software requirements. Consider programming tools and software support when combining CPUs, FPGAs or other accelerators.
- Include security and supply factors. Treat security as part of system design and account for product availability and manufacturing constraints in deployment planning.
Conclusion
Industrial edge computing is best understood as one layer in a wider data-centric architecture. The ideas in Intel’s 2018 report—workload-specific compute, memory and storage choices, connectivity, software, security and operational discipline—highlight the interdependence of the technologies behind industrial data systems. For automation engineers and technical leaders, the practical task is to align those capabilities with application needs and the realities of deployment, from the edge through the network to the cloud.