In 2025, Innodisk launched the “AI beyond the edge” initiative at a forum that also hosted Intel, Nvidia, and Qualcomm, which shared details of their latest developments in edge AI. But what does “AI beyond the edge” really mean?
Don Yu, special assistant to the GM at Innodisk, said that “AI beyond the edge” is about enabling systems that operate autonomously, remain connected, and scale across real-world environments. He also mentioned two complementary domains as part of this initiative.
First, industry AI—built for smart manufacturing, automation, transportation, healthcare, retail, and smart cities—enhances on-site responsiveness through real-time recognition, predictive maintenance, and intelligent workflow optimization.
Second, enterprise AI—designed for data centers, on-premise AI, and advanced models such as large language models (LLMs) and visual language models (VLMs)—supports secure, intelligent decision-making across corporate, financial, medical, and public sectors. “That allows small and mid-size businesses (SMBs) to have their own AI engines locally instead of relying on the cloud,” Yu said.
But despite all the promise, deployment of edge AI has been a challenge so far. So, how are these edge AI initiatives faring so far, EDN asked Yu. And what is Innodisk doing to overcome these challenges in effectively implementing edge AI at scale?
Edge AI deployment challenges
Innodisk chairman Randy Chien acknowledges that the exponential rise of generative AI and LLMs has fundamentally changed the design equation at the edge. More specifically, as AI workloads grow in complexity, companies are facing increasing pressure in system integration, hardware-software coordination, and the ability to scale solutions across diverse deployment environments.
“Anticipating this shift early on, Innodisk has built on its strong hardware foundation by structuring its product portfolio into modular building blocks across memory, storage, camera modules, and a wide range of embedded peripherals,” Yu said. “On this foundation, the company has positioned itself as an AI architect, combining these building blocks to meet diverse industry requirements with tailored edge AI systems.”
So, edge AI developers can implement these solutions as individual modules or as fully integrated systems, depending on their application needs. Take the example of the APEX series of edge AI systems, which brings together key building blocks, including AI accelerators, DRAM modules, flash storage, industrial MIPI and GMSL camera modules, and embedded peripherals for networking and industrial I/O.
“The platform enables flexible system configuration based on specific use cases, while supporting customization to meet diverse deployment requirements,” Yu said.
Figure 1 Individual modules are fully integrated systems tailored according to edge AI application needs. Source: Innodisk
Yu added that Innodisk is heavily investing in firmware and software development to bolster its design ecosystem. Take vision-related AI, for instance, where Innodisk provides fully ported drivers for industrial camera modules, supporting both VLMs and computer-vision applications to streamline deployment and minimize integration friction.
Innodisk also provides specialized software toolkits to accelerate system integration. For example, it has introduced IQ Studio to support the development of Qualcomm-powered edge AI systems. IQ Studio is an open-source developer portal that provides essential board support packages (BSPs), reference code, and benchmarking tools.
How modular solutions aid system integrators
These modular solutions—segmented across five layers of compute, memory, storage, sensing and connectivity, and software—are aimed at addressing design challenges before the last mile of AI deployment in vertical markets. This cohesive system-level approach addresses common development challenges for system integrators and solution providers, enabling them to focus on developing their applications rather than managing integration.
Figure 2 Modular solutions handle integration complexity, which allows system integrators to focus on developing their applications. Source: Innodisk
Moreover, there is a wide range of pre-validated solutions that significantly shorten system integration development cycles. Case in point: AI on Arm series of computer-on-modules (COMs) are designed to be deployment-ready. “They can be directly integrated into customer systems with minimal development effort,” Yu said. “Additionally, they can be paired with Innodisk carrier boards and peripherals to support different system configurations.”
Figure 3 COM modules can be paired with carrier boards and peripherals to support different system configurations. Source: Innodisk
These deployment-ready solutions provide system integrators with practical reference points and inspiration for application design when applied in real-world scenarios. Take the APEX-X200 edge AI platform, for instance, which Innodisk showcased at Nvidia GTC 2026. This on-device inference platform analyzes X-ray and CT images in real time, generating draft medical reports and clinical insights through AI-assisted healthcare workflows.
APEX-X200, powered by an Intel Core Ultra 9 processor, also integrates an Nvidia RTX PRO 6000 Blackwell Server Edition GPU with 24,064 CUDA cores and 752 Tensor cores. Furthermore, it supports up to 96 GB of industrial-grade DDR5 memory and a 1 TB PCIe Gen5 x4 NVMe SSD.
Innodisk has also developed perception systems for heavy machinery and large vehicles in collaboration with its subsidiary Aetina. It integrates the Nvidia Jetson AGX Orin platform with up to eight GMSL2 camera modules alongside capture cards and extenders that support cable lengths up to 30 meters.
Figure 4 The edge AI-based perception system facilitates surround-view stitching, blind-spot detection, and driver-monitoring functions. Source: Innodisk
These perception systems enable surround-view stitching, blind-spot detection, and driver-monitoring functions, supporting real-time environmental awareness and helping identify potential risks such as fatigue or distraction under complex operating conditions. “It’s also an example of a modular architecture that supports future system upgrades without requiring major redesign efforts,” Yu said.
Eyeing U.S. and Europe
Innodisk, headquartered in New Taipei City, Taiwan, has global ambitions with more than 1,000 field-proven edge AI deployments worldwide. In Europe and the Unites States, it’s operating in close collaboration with regional distributors and partners in edge AI segments such as industrial automation, healthcare, aviation, and professional workstations.
Innodisk considers industry events a key tool for bolstering its presence in these crucial markets. It has showcased its edge AI solutions at Nvidia GTC 2026 in the United States, ICE Barcelona in Spain, and Embedded World 2026 and CloudFest 2026 in Germany.
Next, to support global deployment requirements, the company ensures its products comply with regional regulations. Its edge AI solutions meet CE and UKCA requirements for Europe and the U.K. and FCC regulations for the United States.
Also, in Europe, where cybersecurity requirements have become increasingly mandatory, Innodisk attained IEC 62443-4-1 certification in late 2025, embedding security throughout the product development lifecycle rather than treating it as a separate feature. It’s critical because the EU Cyber Resilience Act (CRA) is expected to be fully enforced by 2027.
Related Content
- Top 10 edge AI chips
- Speak Up to Shape Next-Gen Edge AI
- Edge AI powers the next wave of industrial intelligence
- How Advanced Packaging is Unleashing Possibilities for Edge AI
- Edge AI Is Forcing a Rethink of Predictive Maintenance Architecture
The post Edge AI deployment made easy for system integrators appeared first on EDN.