Global server shipments to see CAGR of 14% between 2018-2023, says Digitimes Research

Joyce Chen and Betty Shyu, DIGITIMES Research, Taipei

Worldwide server shipments will grow at a CAGR of 14% in the forecast period 2018 to 2023 as demand from social media, e-commerce and digital enterprises applications continues growing plus the emerging technologies including artificial intelligence (AI), Internet of Things (IoT) and 5G network.

Of the shipments, cloud computing service providers' demand for servers is expected to see a CAGR of over 20% during the period, higher than the average.

As for Intel's next-generation Whitley server platform, it is unlikely to be available until 2020 due to the delay in Intel's new manufacturing technology.

In the past few years, software-defined datacenter (SDDC) has been treated as the key for enterprises to perform the transformation of digitalization. Among the available platform architectures, hyper-converged infrastructure (HCI) that is able to enable a large volume of virtualization services to replace traditional physical equipment, is the latest development trend for servers because of its low hardware costs and strong usage efficiency.

As AI-related applications have become popular, demand for cloud-computing and local-end AI development has been growing rapidly, helping to stimulate the birth of AI-specific servers from vendors. Most of the AI-specific servers are equipped with calculation accelerators and are paired with big data processing software to help manage data and train AI systems.

To reduce clients' process of development and improve their time-to-market, most server vendors are offering server solutions with pre-installed AI systems to their customers.

As more enterprises are adopting microservices such as Amazon Web Services (AWS) and Microsoft Azure's container and serverless applications, to speed up their product development, Digitimes Research expects the transition to boost the development of next-generation dynamic IT architecture, especially features including low I/O access latency and high computing performance, making NVMe and SoC with specific functions key development directions for server vendors.

Currently, most of the available servers in the market can support network transmission speed up to 10Gbps, but vendors and network players have been gradually upgrading to 25Gbps as it is the baseline technology for next-generation 100-400Gbps transmissions.

Digitimes Research estimates that shipments for servers supporting 25Gbps transmission technology will surpass those with 10Gbps in 2020. Since AI-specific servers require a higher bandwidth than regular servers, demand for 100Gbps servers is also expected to pick up steadily.