Author: David Ramos, Deputy Regional Manager Iberia, Italy & Greece – Advantech
The demand for implementing artificial intelligence (AI) algorithms and models directly into edge devices is growing among design engineers. The reason is clear: this approach brings AI capabilities closer to the application, enabling faster processing, reduced latency, and greater privacy—all without relying on centralized cloud servers for processing. In this regard, Advantech offers designers an additional competitive advantage: robust, low-TDP AI-on-Module solutions that incorporate state-of-the-art System-on-Computer platforms in compact form factors from leading manufacturers such as Qualcomm®, NVIDIA, Rockchip, and NXP®, with the option to enhance them with AI acceleration modules.
On the front lines: Critical mission
Modular AI solutions will enable the development of the next generation of rapidly growing battery-powered applications, such as drones and autonomous devices for the defense industry. The latest AI technology in more powerful System-on-a-Chip (SoC) processors, with integrated NPUs and enhanced connectivity features, is contributing to the development of greater functionality. The creation and deployment of AI models are supported by a growing number of existing toolsets and example models to get started. Expected AI performance can be even more accurately measured at the beginning of the development cycle.
Next-generation edge AI is an enabling factor that helps designers meet the fundamental goals of this demanding industry.
Long time-to-market and industry-specific requirements also mean that design engineers are already looking to the future, planning to do things with hardware that doesn't yet exist. Therefore, leveraging the near-future availability of pre-configured edge AI modules from Advantech is both attractive and cost-effective.
Commercial applications
It's worth noting that this perspective is by no means limited to the defense sector. Europe is experiencing a growing demand for commercial applications of Edge AI. Many sectors can benefit from this cutting-edge technology. Advantech's latest solutions allow designers to already utilize PoE power on their application boards. The company focuses particularly on its "Computer on Modules" product portfolio and has some of the best resources and infrastructure in the industry available locally to support engineers in developing their systems.
In addition, Advantech is working on launching PoE-powered embedded computers with AI capabilities. These should allow for rapid time-to-market and be well-suited for small and medium-sized businesses.
Another common concern in commercial applications is data protection and the relatively strict EU legal restrictions. In many applications, processing images at the network edge and sending only anonymized data to the server contributes to regulatory compliance for these solutions.
Knowledge is power
Thermal design power (TDP) represents the maximum amount of heat a component is expected to generate under normal operating conditions. In the case of semiconductors, a low TDP indicates that a processor consumes less power and generates less heat than one with a higher TDP. As a result, system design engineers can offer their customers a solution with reduced power consumption, enabling longer battery life in portable devices, lower heat output that eliminates the need for complex cooling systems, a smaller environmental impact, and potentially quieter operation.
Design engineers must also focus on form factor. Edge AI devices often operate at the limits of thermal solutions. A small form factor is crucial due to the need for compact, energy-efficient, and rugged devices that can be deployed in diverse environments and applications. But is it possible to achieve more efficient thermal solutions and better systems overall with OSM or SMARC modules? This can often be a tough nut to crack in the conceptual design phase.
Advantech can meet both low TDP and small footprint requirements with its next-generation edge AI modules, several of which will be available soon. Among the new releases will be solutions based on Qualcomm® and NXP® CPU platforms.
Roadmap to the future
Advantech maintains an intense plan of new AI module (AI-on-Module) releases as innovations progress from the conceptual phase, through development and validation testing, to mass production.
Among the latest solutions to enter production is the Advantech AOM-2721, which offers a compact L-sized OSM (Open Standard Module) form factor of just 45 x 45 mm and is based on the high-performance Qualcomm® QCS6490 platform. This chipset delivers powerful edge AI through its AI accelerator architecture, with a potential peak AI inference performance of up to 13 TOPS and a low TDP.
On-device machine learning facilitates the exponential proliferation of local computing use cases. The Qualcomm® QCS6490 system-on-a-chip (SoC) is specifically designed for high-performance local computing, combining up to an 8-core Qualcomm® CPU, an integrated Qualcomm® Adreno™ GPU, and a powerful AI engine (NPU + digital signal processor). With three AI-based image signal processors (ISPs), scalable computational performance, premium multi-gigabit connectivity options, and expanded I/O, the QCS6490 is the ideal choice.
The Advantech AOM-2721 OSM 1.1 AI module with Qualcomm® QCS6490 offers PCIe, USB 3.2, Gigabit Ethernet, MIPI-CSI camera serial interface, I2C serial communication protocol, serial peripheral interface, general-purpose I/O, PWM (pulse width modulation) and integrated DisplayPort interfaces, as well as DisplayPort and MIPI-D serial interface displays for embedded applications.
Another new AI module, the Advantech AOM-5521, is also now in series production. It is based on the NXP® i.MX 95 System-on-Computer platform and offers a compact 82 x 50 mm SMARC form factor. The i.MX 95 provides secure and energy-efficient edge computing, combining AI-accelerated vision processing with high-performance computing capabilities and immersive 3D graphics powered by Arm® Mali™ technology. Key features include the innovative NXP Neutron NPU accelerator for machine learning, high-speed connectivity, and security and protection features, as well as the integrated EdgeLock® secure enclosure.
The Advantech AOM-5521, with its integrated NXP® i.MX95 and AI NPU, delivers reliable and secure operation. It offers extensive connectivity options, including USB 2.0, USB 3.2 Gen1 by 1, 10 Gigabit Ethernet, a MIPI-C serial interface, PCIe, a dual-channel LVDS signal interface, and a 4-lane MIPI-D serial display interface. These features make the AOM-5521 an excellent choice for a wide range of embedded applications.
High-performance products
In addition to the Qualcomm® QCS6490 and the NXP® i.MX95, Advantech also plans to launch "AI-on-Module" solutions with the Qualcomm® Elite/X Plus and Qualcomm® QCS9100 models.
This is great news for sectors like defense design engineering, as it expands options for key applications, including hovering drones. The availability of AI modules with small form factors and low TDP—as low as 6,5 watts in some cases—enables critical tasks such as autonomous object detection. In fact, any application requiring features like regulatory compliance, cybersecurity, and resilience will benefit. Options include both solderable OSM modules and SMARC modules, providing greater freedom in mechanical designs.
Innovative options for commercial applications, such as facial recognition in intercoms, are also on the horizon. For example, one upcoming innovation is Advantech's AOM-3841 AI module, which incorporates the low-TDP Rockchip RK3576 SoC.
At a time when design engineers are striving to create optimized solutions by balancing multiple attributes—performance, reliability, resilience, form factor, and low TDP—Advantech's next-generation AI modules offer a convenient and efficient path forward. In Eastern Europe, where the growing demand for AI-based solutions is fueled by increased venture capital investment, strong government support, and a growing pool of skilled tech talent, designers should pay close attention to the availability of edge AI platforms for use in a region that continues to make remarkable strides on the global stage.
