DifferAI Obtains Korea's First Edge AI Semiconductor Technology Patent in China and the US

Artificial intelligence (AI) fabless semiconductor design company DifferAI is targeting the global edge AI semiconductor market by obtaining patents in China and the United States.

DifferAI announced on the 28th that its edge AI semiconductor-related technology has been registered as a patent in the U.S. and China for the first time in Korea.

The newly registered patents include △ a Chinese patent for a neural network parameter optimization method, neural network calculation method, and device suitable for hardware implementation △ and a U.S. patent for a neural network parameter optimization method, neural network operation method, and device suitable for hardware implementation.

These patents relate to technology for lightweight data processing between neural processing unit (NPU) operations, which is known to significantly reduce power consumption during AI data computations. According to DifferAI, an NPU chip applying this technology can demonstrate approximately 100 times the efficiency compared to AI computations using graphics processing units (GPUs).

Lee Sangheon, CEO of DifferAI, said, “With the recent industrial expansion of artificial intelligence, power consumption issues have become prominent worldwide,” adding, “In this situation, it is necessary to implement services through efficient distribution between edge and servers instead of a centralized AI service structure such as data centers.”

He continued, “This can disperse the concentration of power and communication,” and added, “We hope that the newly registered patent technology will play a key role in solving these issues in the future.”