Written & Interviewed by : LITEON Editorial Team (Corporate Brand Value Development Center)
Technical review :LITEON Center of Core Competence
AI computing is moving beyond centralized data centers and into cameras, robots, industrial equipment, gateways, and other devices that must analyze information and respond close to where data is generated. ETSI’s Multi-access Edge Computing framework similarly brings computing capabilities closer to the network edge to support low-latency, high-bandwidth applications. But moving intelligence closer to the physical world also changes the power problem.
An AI camera installed outdoors has very different requirements from an autonomous mobile robot (AMR), an industrial gateway, or an enterprise edge server. Their power budgets, transient loads, thermal constraints, environmental exposure, and physical space can vary significantly. Edge AI power architecture therefore cannot follow a one-size-fits-all model.
Beyond performance and reliability, energy efficiency is emerging as a critical consideration for Edge AI. Given the vast deployment scale of AI cameras, robots, industrial systems, and intelligent edge devices, even a 2% improvement in power efficiency at the device level can deliver substantial energy savings and carbon reduction globally. High-efficiency power architectures therefore play an increasingly important role, serving not only as an enabler of system performance but also as a foundation for sustainable AI development.
For LITEON, this creates an opportunity to apply decades of power expertise across a much broader spectrum — from compact Device Edge products to industrial systems, AI servers, and Data Center infrastructure. Its value lies not only in supplying individual power products, but in combining customization, thermal and mechanical design, Peak Load support, rugged reliability, and one-stop power coverage from approximately 5W to 5000W.
Edge AI places inference, filtering, and selected decision-making closer to cameras, sensors, machines, robots, or local computing systems rather than sending every task back to a centralized cloud.
This does not mean the cloud disappears. Instead, workloads can be distributed according to latency, bandwidth, computing capacity, and application requirements. NIST research has also described the ongoing transition of IoT processing from cloud environments toward the edge and the networking-efficiency benefits that can result. Learn more from NIST. For power engineers, however, the more important question is what happens when AI computing moves into real-world devices.
Device Edge products are becoming smaller, more integrated, and more computationally capable. Yet reducing enclosure volume also reduces the space available for power conversion, thermal paths, magnetics, and protection components.
This means the power supply must balance:
● compact form factor and high power density;
● conversion efficiency and heat generation;
● component lifetime and thermal margin;
● mechanical structure and heat conduction;
● external surface temperature and user safety.
For LITEON, thermal management is therefore not an isolated cooling problem. It is closely connected with circuit design, material selection, mechanical architecture, component placement, and long-term reliability.
Edge AI is also moving beyond climate-controlled environments. Cameras, sensors, gateways, and industrial equipment may operate outdoors or in facilities without air conditioning.
For selected applications, LITEON can provide rugged adapter designs with capabilities such as IP67 waterproof and dustproof protection and wide-temperature operation, depending on product and customer requirements. Importantly, rugged environmental design should not be confused with redundant power architecture. Rugged design protects equipment against its operating environment; redundancy adds additional power capacity or paths to improve system availability. They address different reliability challenges.
AI workloads are rarely constant. A robot may require substantially more power during acceleration, gripping, or movement. A camera may activate additional illumination or processing functions. Charging and computing loads may overlap. Peak demand can therefore differ significantly from average consumption. This makes Power Budget and Peak Load behavior central considerations in Edge AI power design.
Traditional power selection often begins with rated wattage. Edge AI requires a more detailed understanding of how the device actually operates.
A realistic power profile may need to consider:
● Continuous Power Budget
● Peak Load magnitude and duration
● Input and output requirements
● Standby consumption
● Charging behavior
● Transient response
● Operating environment
● Thermal conditions
● Mechanical constraints
● Reliability targets
This is where customization becomes one of LITEON’s core advantages. Rather than applying the same generic power supply across different Edge AI products, LITEON can adjust power behavior according to the customer’s application, operating sequence, and system requirements.
Physical AI applications such as AMRs and robotic arms illustrate the problem clearly. A robotic system does not consume the same amount of power throughout its operating cycle. Motors, sensors, controllers, cameras, communications modules, and charging systems may create different load profiles depending on what the machine is doing. A power supply designed only around average wattage may therefore be inefficient during normal operation or insufficient during Peak Load events.
LITEON can optimize the power design around actual operating behavior — including Power Budget, Peak Load, thermal conditions, charging requirements, and mechanical constraints — helping improve reliability while reducing unnecessary power and thermal overhead. LITEON has also accumulated power architecture experience for Physical AI applications including AMRs and robotic systems, with DC converter and charging solutions supporting these environments.
Customization is particularly important for SMEs. Many smaller organizations adopting Edge AI do not have large machine rooms, extensive power-engineering resources, or the ability to coordinate different vendors for adapters, chargers, converters, industrial power products, and higher-power infrastructure. Their challenge may not be choosing an AI algorithm. It can be finding a reliable power architecture that matches several devices and operating conditions without creating unnecessary integration work.
LITEON addresses this through a combination of:
● broad wattage coverage;
● application-specific customization;
● Peak Load support;
● rugged options for demanding environments;
● and one-stop power sourcing.
This can reduce the number of supplier interfaces while shortening development and integration cycles.
GaN has become an important technology for compact and high-frequency power conversion. Wide-bandgap semiconductors such as GaN can support greater efficiency and higher-temperature operation than conventional silicon in suitable power electronics designs, while packaging and power-module design remain critical to realizing those benefits. The U.S. Department of Energy discusses these advantages and the importance of packaging and module design in advanced power electronics.
But using GaN itself is not LITEON’s core differentiator. GaN components are commercially available to many manufacturers. The harder problem is turning those devices into a compact product that can operate efficiently, dissipate heat effectively, protect internal components, remain comfortable to touch, and be manufactured reliably.
That integration is where LITEON’s accumulated know-how becomes important.
Thermal management begins before heat reaches the enclosure.
LITEON can optimize circuit design and conversion efficiency to reduce unnecessary power loss, decreasing the amount of heat that must ultimately be managed inside a compact system.
Thermal materials determine how efficiently heat moves from power devices into other structures. Material selection therefore needs to consider not only thermal conductivity, but also mechanical design, insulation, reliability, and product lifetime.
The next challenge is physical integration. High-density components must fit into increasingly small spaces while preserving:
● appropriate thermal paths;
● electrical insulation;
● structural reliability;
● component life;
● manufacturability;
● and acceptable enclosure temperature.
This combination of thermal management and mechanical design — rather than GaN alone — represents the deeper technical barrier. LITEON’s advantage is its ability to turn high-density semiconductor technology into a complete power product by coordinating circuit topology, thermal materials, mechanical structure, component placement, and manufacturing experience.
Edge AI does not exist at only one computing layer. At one end are compact sensors and endpoint devices. Moving upward are cameras, gateways, AMRs, industrial equipment, enterprise edge systems, and local computing infrastructure. Beyond those systems are AI servers and centralized Data Centers.
LITEON’s power portfolio spans this continuum, with approximately 5W to 5000W of power coverage. Relevant solution categories include:
● compact adapters and chargers;
● Device Edge power;
● industrial power supplies;
● battery charging and DC conversion solutions;
● enterprise and edge-computing PSUs;
● AI server power;
● and Data Center-level power infrastructure.
For customers, the advantage is not simply having more products in a catalog. A broad portfolio allows one power partner to support different stages of an AI architecture as requirements evolve. A company may begin with a small Edge AI device, expand into industrial edge computing, and later connect that system with higher-performance servers or Data Center infrastructure. Working with a supplier that understands all of these power environments can reduce the need to rebuild power-supply relationships every time the computing architecture changes.
LITEON’s coverage from Device Edge to Data Center also gives it visibility across different power trends. This helps the company respond more flexibly as market requirements shift and evaluate where technologies such as GaN, SiC, digital power, higher-density conversion, or new interface standards are most appropriate.
The broader value is LITEON’s ability to combine technologies according to actual customer requirements rather than promoting one technology for every application. A compact Device Edge product may prioritize size and thermal performance. An industrial system may prioritize ruggedness and Peak Load. A server environment may prioritize power density, efficiency, management, and availability.
Years of power design experience allow LITEON to select and customize the appropriate combination while maintaining reliability as a central design priority. That turns “one-stop solution” from a sourcing convenience into an engineering advantage.
Physical AI will expand the role of Edge AI beyond data processing into machines that perceive and interact with the physical environment. AMRs, robots, smart factories, intelligent transportation systems, and sensor-driven industrial equipment all require stable power because computing results must translate into real-time physical action. For these systems, a power disruption does not simply interrupt software. It can affect sensing, movement, control, charging, production, or machine availability.
The future Edge AI market will not converge on one standard wattage or form factor. Instead, LITEON expects different directions across the edge: Device Edge will increasingly emphasize miniaturization, higher switching frequency, and greater integration among semiconductor devices, power components, and third-generation semiconductor technologies. Enterprise and Industrial Edge will increasingly emphasize digital power, higher conversion efficiency, higher power density, remote management, and more diverse power architectures.
Digital power has already been adopted in cloud and server environments. As more computing occurs locally, similar capabilities are moving closer to Edge AI equipment.
Digital power can add greater visibility and control through capabilities such as:
● real-time power monitoring;
● status and fault information;
● configurable operating parameters;
● remote diagnostics;
● and more intelligent system-level power management.
The significance is not simply digitizing the PSU. It is enabling power infrastructure to become more responsive to the conditions of the AI system it supports. For LITEON, its presence across Device Edge, Enterprise / Industrial Edge, server, and Data Center power provides a foundation for applying these technologies where they create the greatest system value.
The shift toward Edge AI is not only moving computing closer to devices. It is creating a wider and more fragmented set of power requirements. A camera, AMR, industrial gateway, robot, local AI server, and Data Center cannot be optimized through the same generic power architecture. Successful Edge AI deployment therefore requires power design that considers Power Budget, Peak Load, space constraints, thermal performance, environmental conditions, reliability, and future scalability together. This is where LITEON’s differentiation becomes clear.
With deep customization, thermal and mechanical integration, GaN application expertise, SiC technology capabilities, broader advanced semiconductor know-how, rugged power design, Physical AI application experience, and approximately 5W-to-5000W coverage from Device Edge to Data Center, LITEON can help customers build power architectures around how their AI systems actually operate — not simply around rated wattage.
As Edge AI evolves toward smaller Device Edge products, more capable Physical AI systems, and increasingly digital Enterprise Edge infrastructure, the power layer will need to evolve with it.
Explore LITEON’s power solutions to build a more reliable, efficient, and application-specific power foundation for your next Edge AI deployment.