Taipei, Taiwan - DFI, a global leader in embedded motherboards and industrial computers, has been honored with the 2025 Taiwan Excellence Award for its innovative “Intelligent AI Retailer Kiosk”. This cutting-edge product integrates a high-performance edge AI computing platform with virtualization technology, enabling multiple operating systems to run on a single hardware platform. Capable of managing facial recognition, digital advertising, payment transactions, and on-device LLM intelligent services, the innovation significantly reduces hardware costs and power consumption. Additionally, its built-in Out-of-Band (OOB) management technology enhances digital transformation efforts, meeting the diverse application and energy-saving demands of edge computing.
The Future of AI Computing Is at the Edge.
According to NVIDIA, 64% of major retailers have already deployed AI technologies, with 80% utilizing AI across three or more application scenarios. McKinsey Digital estimates that AI’s potential annual market value in retail and consumer goods ranges from $400 billion to $660 billion. However, the immense computing resources required to train and operate AI models intensify energy consumption and environmental pressure.
DFI’s Product Center General Manager, Jarry Chang, emphasized the growing demand for edge computing amid these challenges: “Real-time decision-making needs and the rapid growth of edge data are driving this transformation. Processing data closer to its source not only reduces latency but also enhances privacy. By integrating virtualization technology, DFI consolidates multiple workloads onto a single industrial computing platform, enabling resource partitioning and remote management. This boosts system efficiency, lowers total cost of ownership (TCO), and tackles the challenges of AI and sustainability.”
Efficient and Energy-Saving Edge AI Innovation.
Traditional architectures require four hardware platforms to support a single operating system, consuming approximately 40 watts of power. DFI’s solution, leveraging a single ATX industrial-grade motherboard RPS630-R680E/Q670E, supports multiple operating systems and functionalities, with tested power consumption of just 32.2 watts.
The “Intelligent AI Retailer Kiosk” is equipped with the Mistral 7B AI language model and Intel® Arc™ GPU for rapid AI computation, the kiosk achieves an 80% language understanding accuracy rate, with response times under 2 seconds. Intel® OpenVINO™ toolkit further simplifies deploying trained AI models onto new systems. Combined with OOB management capabilities, it supports remote operations like BIOS updates, even in a powered-down state, reducing maintenance costs and ensuring uninterrupted operations.
The “Intelligent AI Retailer Kiosk” exemplifies an innovative design that simultaneously achieves energy savings and operational efficiency, injecting new momentum into edge AI applications. Beyond smart retail, its versatile concept is applicable to fields like healthcare, transportation, and advertising, addressing the needs of autonomous and intelligent services.