Workflow
自研AI芯片,可行吗?

Core Viewpoint - The article discusses the challenges and complexities of chip design and manufacturing, emphasizing that it is a long and intricate process that differs significantly from the fast-paced nature of the OTT (Over-The-Top) industry [4][5][6]. Group 1: Industry Characteristics - Chip design is portrayed as a manufacturing industry disguised as high-tech, where the final product is a physical entity requiring extensive production resources [5][6]. - The manufacturing chain for chips is lengthy and complex, involving various operational tasks such as ordering, inventory management, and quality inspection [7]. - The unique nature of the chip design industry means that it has not established efficient abstraction and division of labor, making it distinct from the digital products of the OTT sector [6][7]. Group 2: Time and Investment - The time required to design and manufacture a chip is significant, with estimates of 8-10 months from design completion to physical chip availability, and over 36 months for a chip to be publicly released and delivered to customers [10][12]. - The investment required for developing a decent AI chip starts at 2 billion RMB, with production costs per chip being comparable to high-end GPUs, making profitability a challenge [11][12]. - The article highlights that the ROI calculations often overlook the complexities and timeframes involved in chip manufacturing, leading to misconceptions about the feasibility of OTT companies entering this space [8][10]. Group 3: Efficiency and Adaptability - For OTT companies to succeed in chip manufacturing, they must focus on improving efficiency and adapting to the slower, more complex manufacturing processes [12]. - The article suggests that traditional manufacturing processes may need to be re-evaluated in the context of rapid technological changes, where speed and adaptability could be more valuable than reliability [12]. - The potential for innovation in chip design lies in the ability to streamline processes and reduce the time from design to production, which is critical in a fast-evolving tech landscape [11][12].