News description
     Global server shipments showed a quarterly growth rate of 11.2% in 2019 Q3, with shipments in Q4 expected to increase slightly further. Not only is this better than the original expected forecast of a slight decline, but it has resulted in an overall annual growth, which will lead to a lifting effect on affected servo PCBs in the first half of the year.

Impact analysis
    The primary reason for the weak demand in the global server market in 2019 is the slowdown in demand from companies outside the Asia-Pacific region and the hyperscale server market. This decline is especially evident in North America and China, the markets of which account for the largest proportion of global server shipments. If shipment volume in Q4 2019 can be recovered and server specifications improved simultaneously, it is likely that both the price and the volume of server PCBs would rise. 
    In the future, server systems will gradually be developed towards the public cloud using AI computing. Based on the 2018 data volume prediction by Cisco, the global data center flow will reach 20.6ZB (zettabytes) by the end of 2021, which is three times higher than in 2016. The virtualization of data centers will be an essential driving factor for the rapid expansion of cloud computing, which can provide flexible, rapid deployment and efficient services. Based on forecasts, 95% of the data flow (about 19.5ZB) in 2021 will be processed in the cloud.
    From the point of view of PCBs, not only can the construction of hyperscale data centers drive the demand for growth, it is highly critical that the massive flow and high transmission speeds carried by data centers will also increase the number of layers and material requirements, thus boosting the development of high-order PCBs. Take Nvidia for example; in recent years, the company has been actively developing their server platforms. The HGX-2 released in 2018 was the company’s first platform architecture that combined the benefits of HPC (high-performance computing) and AI. The platform uses Nvidia’s fiber-optic interconnect technology (NVSwitch), connecting 16 Tesla V100 GPUs in series to provide 2 petaflops of computing speed - equivalent to 2 trillion floating-point calculations per second. The signal transmission bandwidth between the GPUs is 24 times larger than the previous-generation HGX-1 platform. In summary, the demand for server products will steadily increase matching future generations of computing, and the market for PCBs will also grow, regardless of specifications or region.