Can Legacy Fabs Keep Up with IoT Demand?
The Internet of Things (IoT) is set to drive demand and innovation in the semiconductor market over the next decade. While some consumer IoT applications will require semiconductors manufactured using cutting-edge technologies to deliver fast performance and low power consumption, the vast majority of chips for IoT applications will be utilized in client-side applications. These chips, such as a sensor monitoring room temperature in a connected HVAC system, require processing capabilities that can be met using legacy process (90 and 45nm) technologies manufactured on 200mm wafers.
And herein lies the opportunity and challenge for legacy manufacturing. According to iSuppli, nearly half of the silicon consumed today is devoted to 200mm and smaller wafers, and with the exponential growth of IoT devices, the reliance on legacy systems will significantly increase. As the IoT expands, the number of sensors could reach one trillion by the end of this decade, and eventually the IoT market could consume that many sensors per year. Gartner further estimates it would take nearly 290 200mm fabs to meet this demand for sensors alone. So the question is, are 200mm fabs up to the task of delivering the trillions of devices needed to link everything?
Both leading-edge and 200mm fabs require advances in processes and capabilities to fabricate future designs and operate at peak utilization and reliability so they can be productive and profitable. For 200mm equipment to remain productive, these requirements depend on technical extensions and fab enhancements. Among the priorities is upgrading 200mm tools to run the modern automation software and advanced process control (APC) applications that comprise an emerging and effective industrial IoT (IIoT) approach for maximizing productivity and yield in fab environments.
Implementing an IIoT enables fabs to move to a predictive mode where data gathered from APC technologies and by sensors attached to thousands of machines can be used to detect, classify, diagnose, control, predict and prevent various failure modes. It is the new data that makes predictive maintenance a reality. Existing data retention rates are about 90 days, which is based on just having sufficient data for wafers currently in the fab and to review gross trends. Advanced analytics for a predictive approach will require a much larger data time frame – up to 12 or more months of data.
The increased data from the IIoT, coupled with big data analytics gathered from fab data mining tools, can provide enough granularity to allow chipmakers to identify when machines may experience problems and take appropriate corrective action before production is impacted. In the future, the IIoT and big data analytics will generate automated learning within the fab, enhancing automation systems and correcting issues before they become a problem, without the need for human intervention.
With the need to move faster to the next node, increase yield, and reduce costs, implementing a modern, more intelligent, connected and responsive manufacturing IIoT microcosm is going to be critical in producing the chips that will fuel the IoT age.