News description
    Under the intelligent life trend where everything is connected, the amount of data collected, transmitted, and calculated is increasing enormously. Thus, edge computing at the terminal will play a vital role in reducing the data processing burden of cloud servers.

Impact analysis
    When various kinds of IoT data are uploaded to the cloud for AI computing, training, and analysis, it will tend to delay the overall operation analysis, data response, and machine operation efficiency. So, when shifting part of the machine learning mechanism to the terminal devices from the cloud, instead of initially sending data training, analysis, and prediction work to the cloud server or big data center, edge computing will move them to the edge nodes of various network logic architectures for processing making it is necessary to improve the ability of these edge nodes in processing the data.

 

    According to IDC statistics, it is predicted that more than US$50 billion of terminals and devices will be networked by 2020. Therein, more than 50% of the data need to be analyzed, processed and stored at the network edge. Although On-Line Training is still essential, On-Device Training will become more resplendent. In other words, the terminal devices do not need to be connected at any time. When it is offline, the terminal devices must also have some capabilities in making decisions and responding immediately.

 

    From the rise of edge processors and the demand for hardware devices, more terminals will be equipped with AI chips and algorithms in the future. From Amazon’s acquisition of chipmaker Annapurna Labs, a system-on-a-chip that supports Azure Sphere co-developed by Microsoft and MediaTek, the Edge TPU chip self-developed by Google, etc., it is understood that chips and algorithms will define the development direction of future edge computing software and hardware. Therefore, there will be more demand for carrier boards with AI chips in thinner line width, more layers, and a larger area. The ABF material-based carrier boards have also become a popular product for carrier board manufacturers to compete for investment expansion.