In 2025, server spending will account for nearly half of data center capital spending


New research report, 5-year Forecast: Server CPU Updates, Accelerated Computing, and Edge Computing drive Future Data Center spending. Server spending is expected to grow at a compound annual growth rate of 11% over five years and will account for nearly half of data center capital expenditures through 2025.

Changes in corporate and consumer behavior have generated strong demand for computing and digital technology, and current semiconductor contract manufacturing capacity is insufficient to meet the recent surge in global demand.

So the next period is likely to be one of demand exceeding supply. Average server price (ASP) increases are likely to be close to the double-digit levels observed in 2018. The main reason is a global semiconductor shortage. This could lead to a sharp rise in the cost of services and other data center equipment in the short term.

The rise in server prices in 2018 is due to the oligopoly effect of the Internet and cloud computing, which has created demand for large and very large data centers. Many midsize businesses are also located differently from large data centers. In 2018, the United States, together with amazon, Google, Microsoft, Rackspace, IBM and other giants, built a super large data center, accounting for 44% of the world.

Supply and demand dynamics are expected to balance over the long term, and eventually technological changes will drive market growth. The five-year technology trend is expected to look like this

● CPU update cycle: Expected in 2022, both Intel Sapphire Rapids and AMD EPYC Genoa will include more processor cores and memory channels, and support for the latest ports such as CXL, DDR5 and PCIe Gen 5, enabling a denser server shape and new architecture. This is largely due to the heating up of processor competition, with both Intel and AMD having aggressive road maps to introduce new platform updates.

● Accelerated computing: The connection rate for servers with accelerators will grow to 13 percent by 2025. A new type of accelerated server, which can be densely packed with application-specific workloads (such as artificial intelligence and machine learning) optimized for coprocessors, is further upgraded.

Some cloud providers, such as Amazon and Google, have deployed accelerated servers that use their own AI chips, while other cloud providers and companies are widely deploying Gpus and FPGA-based solutions.

● Edge computing: Certain applications (such as cloud gaming, autonomous driving, and industrial automation) are latency sensitive and require multiple access edge computing (MEC) nodes to be placed at the edge of the network, where the sensors are located. Unlike cloud computing, which is replacing enterprise data centers, edge computing creates new market opportunities for new use cases.

As CPU platforms evolve and high-speed computing spreads, fewer servers are available to handle application-specific workloads, but they are more powerful, more dense, more data-centric, and so on. The higher the server ASP, the more applications will grow. Edge Computing in the market We will also expand the available market by deploying servers in greenfields at distributed edge locations.