Home/Glossary/Edge AI and Edge Computing ASICs
Semiconductor knowledge

Edge AI and Edge Computing ASICs

Custom chips that process data near the place it is generated, such as a camera, sensor or industrial device. They can reduce the amount of data that must be transmitted to a remote system.

← All glossary termsTechnical guide · 5 min read
In this article

Custom chips that process data near the place it is generated, such as a camera, sensor or industrial device. They can reduce the amount of data that must be transmitted to a remote system.

A compute and memory design balance
Conceptual architecture concerns. Block size does not represent area, bandwidth or measured performance.

Processing near the source

An edge computing ASIC performs selected work near the device or data source rather than relying entirely on a remote computing service. Edge AI adds local inference on sensor, audio, image or other inputs. The motivation can include response time, connectivity limits, energy or data-handling requirements. Start with the complete local pipeline: acquisition, preprocessing, inference, decision and communication. A fast neural-network engine alone may not improve the product if another stage dominates latency or if moving data into the engine consumes excessive energy.

Fit the model to the device

Local memory, available power and thermal conditions constrain the supported workload. Model size, numerical precision and operator coverage must be evaluated together. Some devices operate intermittently and spend much of their time asleep; others need continuous low-latency processing. Test the model on representative inputs and difficult environmental conditions. Accuracy or quality requirements should remain visible when reducing precision or changing the network. A smaller model is useful only if it still supports the intended product behavior within the defined limits.

Plan updates and fallback

The deployed model and surrounding software may change after the chip is manufactured. Define the supported model formats, memory limits and update mechanism. Decide what happens when local confidence is low, input quality is poor or connectivity is unavailable. A cloud fallback can be useful in some architectures but should not be assumed in a product that must operate offline. Security requirements also apply to model and firmware updates. The ASIC's fixed capabilities should match a realistic range of future software changes rather than an undefined promise of supporting any model.

Measure the full decision path

Evaluate energy per useful decision, end-to-end latency and sustained behavior in the actual operating modes. Include sensors, memory, host processing and communications where they belong in the product boundary. A benchmark that excludes preprocessing or assumes an always-full batch may not represent an interactive edge device. State whether results are measured or estimated and preserve the input conditions. This gives architecture teams a credible basis for comparing a custom ASIC with a processor, FPGA or other available accelerator.

Questions about edge AI

Does edge processing eliminate all cloud communication?

No. Systems can combine local processing with remote services in different ways. The architecture should state which functions remain available offline and which depend on a connection.

Is peak accelerator throughput the main edge metric?

Not always. Energy, response latency, memory footprint, model quality and sleep behavior can be more important for a particular device than maximum arithmetic throughput.

Define a fair workload boundary

Specify the input size, data format, numerical quality and complete work being measured. Include transfers, preprocessing or host activity when they are part of the product outcome. Distinguish single-request latency from sustained throughput and explain batching assumptions. A comparison is useful only when the implementations provide equivalent results under comparable conditions. Preserve representative inputs and include difficult cases that challenge memory or control behavior. This prevents a favorable arithmetic peak or a carefully selected kernel from being mistaken for a general statement about every application the chip might encounter.

Account for data movement

Trace where operands, intermediate results and persistent state reside. Estimate the traffic between processing elements, on-chip storage and external memory. Consider reuse, contention and synchronization. More compute capacity helps only when supporting resources can keep it supplied with useful work. Evaluate the effect of changing model size or problem shape, because a design that fits one workload in local memory may behave differently when that capacity is exceeded. Include the relevant interconnect and package constraints when the architecture relies on high external bandwidth.

Connect hardware with software

Review how compilers, drivers, runtime scheduling and application code use the proposed architecture. Identify unsupported operations and the cost of fallback paths. Provide profiling and diagnostics that distinguish a software mapping problem from a hardware capacity limit. Keep the software configuration tied to benchmark results and hardware revisions. Planned flexibility can extend a chip's useful life, but it has finite limits in memory, interfaces and supported operations. Explain those limits so product teams can evaluate a realistic roadmap rather than relying on a broad claim of future compatibility.

Validate sustained operation

Power delivery, cooling, operating voltage and environmental conditions affect the performance a product can sustain. State whether evidence comes from a model, implementation estimate or silicon measurement. Evaluate the duration and operating modes relevant to the intended application rather than only a short peak test. Report useful work alongside energy or power with a clear measurement boundary. When one design changes numerical precision or output quality, show that difference explicitly. A strong engineering comparison explains why a result occurs and where it stops applying, allowing the team to make an informed architecture or deployment decision.

A practical cross-discipline review

Before committing the next stage, bring the design, verification and product-engineering owners together around the same revision. Confirm the assumptions that pass between them: operating modes, interfaces, test access and the evidence expected at release. List unresolved issues with a responsible owner and a concrete way to close them. An open question is easier to manage when its consequence is understood than when it is hidden inside a general progress estimate. Preserve the decision record so a later change can be evaluated against the original reasoning. This review is useful even for a narrow specialist task, because the surrounding product still determines whether the delivered result is usable.

Include the receiving team's acceptance question in the review: what must be demonstrated before this result can be used safely in the next engineering stage? Agree how that demonstration will be recorded and how a subsequent change triggers a focused reassessment of the affected evidence.

Explore the related Avnet offering or discuss your project requirements.

Let's talk about your next project.

Explore design, technology enablement and manufacturing with Avnet ASIC Solutions.

Contact our team →