AI 반도체 산업의 기술발전 방향과 후발자의 추격 가능성에 대한 고찰 : 미국 특허를 이용한 기술수명주기 분석을 중심으로

With the advent of the 4th industrial revolution, new technologies is receiving attention. These are technologies that can change our lives, such as big data, AR/XR, IoT, and autonomous vehicles. These technologies are used more efficiently when combined with AI. Therefore, based on the existing literature on AI semiconductor, which is the core for the AI technology, we have classified AI semiconductors as three types such as the 1st generation(CPU and GPU), 2nd generation(FPGA, APIC and Neural), and 3rd generation(Neuromorphic). With the classification, we collected the patent data and analyzed that whether a window of opportunity exists for latecomers. The results are as follows. First, the amount of the AI semiconductor patents increases steadily, and the growth rate has risen sharply after 2018. Second, the relative technological cycle time of the 3rd generation is shorter than that of the 1st and 2nd generations, and the gap is high in recent years. Third, from the perspective of the window of opportunity, a shorter cycle time means that it is less reliance on existing technologies and using more new technologies. Using less existing technology means that the incumbents and the latecomers are on the same starting line at the time of technological paradigm shift. The technical field of 1st and 2nd generation AI semiconductors on the extension line of existing technologies have difficulties in entering the market for latecomers. However, since the 3rd generation presents a new technological paradigm that utilizes a technology similar to the human brain, the market potentials is high even if the technological level is still insufficient. In other words, in the case of 3rd generation, as a competitive market where there is no dominant incumbents yet, the window of opportunity for latecomers is wide opened.