Low-power artificial intelligence (AI) corporations DeepX said on the 14th that it obtained the government's "green technology certification" for its AI inference technology that accelerates low-power Edge Computing at the edge.
A total of four products, including two DX-M1-based M.2 acceleration modules and two DX-H1-based PCIe acceleration modules that apply the technology, also received "green technology product verification." The certification and verification are valid through Sept. 9, 2030.
According to certified testing by Korea Testing Laboratory (KTL), DeepX's high-performance PCIe lineup reduced power consumption by 21.84% to 26.25% and lowered operating temperatures by more than 20 degrees compared with a reference system. The compact M.2 lineup also cut power consumption by 17.26% to 23.35%.
DeepX said it boosted power efficiency by reducing data movement and memory access through its self-developed Neural Processing Unit (NPU) architecture. By cutting the amount of data shuttling between memory and compute units during processing, the approach lowers power consumption and heat while maintaining AI inference performance.
Green technology certification is a program based on the "Framework Act on Carbon Neutrality and Green Growth for Coping with the Climate Crisis." With this certification, DeepX can receive benefits such as priority purchasing in public procurement, commercialization bonus points, policy funds and financing, and support for overseas expansion.
A DeepX official said, "As AI spreads beyond data centers to industrial sites such as robots, factories and mobility, achieving high performance with less power will become more important," adding, "We will expand the application of low-power AI technology across various industrial sites to help reduce the energy burden from the spread of AI."