top of page

Year of writing: 2020
Author: Donny

Data Center Data Cente r  Part 1

Comprehensive understanding of the data center

The data center theme will be divided into three comprehensive introductions to data center related industries, especially white-label servers, while introducing several concept stocks, and analyzing related industry chains and future trends.

In the data center, it is necessary to provide sufficient power, effective cooling system, and a huge hinterland, and set up a large number of servers at the same time. These networked devices are responsible for storage, data processing, calculation, analysis and other tasks. Especially in the AI era, there will be more AI and machine learning rely on cloud computing. Applications that are closer to our lives will ask Google if you have any doubts on the Internet. This kind of search engine; Uber, which we go out and ask for a car; Xiaopi and Amazon, which are online shopping; Netflix, which is chasing dramas and watching movies. , IQiyi; Facebook and IG for sharing life; Stadia for game streaming. All of these require the operation of the cloud server in the data center to support our lives. Some people describe data as the power of the AI era. Then I want data The center is probably the power plant, which supplies data for operations around the world.

Types of data centers

The data center is divided into four categories according to the purpose of use:

1. Colocation Data Center  

Hosting means that the user has his own host or server, but there is no room in the computer room, and he does not want to set up electromechanical equipment, cooling system, security, etc., and the user places his host in the hosting data center. The center assists in keeping and turning on and off the machine, but companies still need to build their own systems. This advantage includes reducing the cost of building their own data centers and having independent hosts, but for corporate users, talents with certain technical thresholds are required. Taiwan Many local operators such as Yuanchuan provide this service.

2. Hyperscale Data Center  


This type of data center has flourished in recent years. It is Amazon, Microsoft, Google, Facebook, IBM and other companies that we usually follow in the news media. They not only provide internal use of the company, but also rent out the host (server) and computing. The performance is used by ordinary enterprises, so that ordinary enterprises do not need to build data centers and purchase host servers by themselves, and flexibly schedule the use of each host to improve the use efficiency of each host. General enterprises can also save costs, quickly build, and expand in time. Now major manufacturers provide machine learning and AI applications, allowing users to enjoy these functions at a premium as long as they use cloud servers. This drives the general enterprise to be more willing to transfer data to cloud virtual hosts. Very large data centers have at least 5,000 servers, and generally have 50,000 to 100,000 servers. Major manufacturers will also set up data centers around the world according to their needs. At present, the number of ultra-large data centers in the world is about 550. Trendforce and Cisco predict that they will grow at a CAGR (compound annual growth rate) of 14% and 25%, respectively. It can be seen that the rapid growth of data centers and the huge demand for servers will drive the growth of the overall data center and server industry. . The main operators of super-large data centers are concentrated in the two major countries in the United States and China, including Amazon's AWS, Microsoft's Azure, Google's GCP, Tencent, Baidu, etc.

Google 資料中心

When you see the blue, red, yellow and green design, you know that it’s Google’s data center



For companies with huge cloud computing needs, they build corporate data centers for their own internal use. Although they are not as large as Google and Amazon, they are usually classified as second and third line customers of server ODM factories, but in the data center , The server market still has huge influence, such as Netflix, Uber, Twitter, Salesforce, Zoom, etc.

4. Telecom Data Center (TELECOM DATA CENTER)  


A data center built for telecommunications companies. This type of data center is responsible for message transmission and communication services. Because of the promotion of 5G and the popularization of the Internet of Things in the future, it is imperative to transmit large amounts of data. Therefore, telecommunications companies need to build more The data center responds. In the past, the telecommunications equipment used in the telecommunications data center was exclusive equipment, controlled by Nokia, Ericsson, and Huawei; now Lotte Telecom pioneered a virtualized mobile network infrastructure to replace traditional telecommunications equipment. Virtualized architecture no longer requires dedicated hardware and software equipment, only a general-purpose white-brand server. This architecture can make it easier for users to connect to the Internet and upgrade, and telecom companies can effectively reduce the cost of construction by about 40%. Layout for AIOT (Intelligent Internet of Things). Telecom companies including AT&T, Verizon, and Lotte Telecom have all built traditional or virtual telecom data centers. In the future, more telecom companies will invest in building virtualized mobile network infrastructure.

Self-built server vs. cloud server

With the increasing number of online and networked applications, and the new crown pneumonia (COVID-19) accelerating the online deployment of enterprises, the cloud industry is also booming. In the past, companies used to build traditional servers by themselves, purchase branded servers, and set up the servers in the company for data storage. However, this mode requires all-in-one construction, maintenance, updating, and expansion; when Amazon launched in 2006 Create new business models and lease cloud hosts to companies in need. Now more and more companies tend to build data in cloud data centers, which can effectively reduce costs and maintain stability for companies. This has also greatly changed the entire server industry, making the sales of traditional brand servers whose main application is enterprise-built servers have slowed down in recent years; on the contrary, cloud servers have grown rapidly. According to a research conducted by IDC, a technology research institution, in 2018, enterprises used internal traditional brand servers accounted for about 60%, and cloud data centers accounted for 40%; by 2023, it is expected that more than half of server-related expenditures will be in cloud data centers, and 50% %.

Brand server vs. White brand server

In the past when companies built their own servers, most of them bought branded servers (HPE, DELL). At the same time, the major cloud companies that provided rental server services also purchased servers directly from these branded server manufacturers and built data. Center, large cloud users ( such as Salesforce, Uber, etc.) are also the same; however, due to the increase in usage and the impact of customized demand, first-tier cloud manufacturers and second- and third-tier cloud users are also turning to ODM factories to design and order directly Servers, such as white-label servers, have changed the appearance of the server industry, creating a phenomenon where one goes from one to the other.


The foothold of the Taiwanese factory── ODM

Most of the brand servers and white-brand servers ODM (original factory commissioned design and manufacture) are manufactured by Taiwanese manufacturers. However, the main force of each manufacturer is different. Some mainly focus on brand server foundry and some directly face the cloud manufacturers to assist in the design regulations. Server for drawing and manufacturing data center. Several big brothers in the electronics industry in Taiwan have all entered the server market, and some have even pulled a subsidiary to design and produce and win large orders together, including Hon Hai, Inventec, Quanta and Yunda, Wistron and Wisdom. , Heshuo, ASRock, Mitac.


The main OEM brand server vendors headed by HPE and Dell include Inventec, Wistron, Mitac, and Hon Hai. Inventec has the largest market share in brand server ODM, although these companies are more or less White-brand servers are produced, but brand-name servers still account for the majority. Wistron's subsidiary, Wivy, is responsible for the design of white-label servers and data centers, and then Wistron will manufacture these servers through its huge production capabilities and supply them to data centers. Quanta and its unlisted subsidiary, Yunda, have also won many white-label server orders through the same cooperation model. Yunda is responsible for the design and Quanta is responsible for the production. These large PC/NB foundries in the past are gradually leaning towards high-margin orders such as servers. More detailed white-label/brand server proportions and supply chain division of labor will be analyzed in Part 2.

Three major cloud service categories Iaas / Paas / Saas

To distinguish cloud applications by service, it can be roughly divided into three parts: the bottom-level IaaS (Infrastructure as a Service) infrastructure as a service, the middle-level PaaS (Platform as a Service) platform and services, and the top-level SaaS (Software as a Service) ) Software and services. The IaaS service is to provide users to rent and use their cloud host, which is the service provided by the super-large data center operators such as AWS, Azure, and GCP. PaaS provides a software development platform that allows enterprises to quickly deploy applications or functions they want to develop. SaaS regards software as a service, and it is also the service most frequently encountered by our general users. We can directly start using this online without any server-related settings. The most famous offering of SaaS is the first to create this service. CRM software Salesforce, Google provides a wide range of services such as Gmail, which are all classified as SaaS. These three categories affect the changes in the entire data center industry. Part 2 will fully analyze the data center customers and supply chain, and find relevant concept stocks with high development for readers. If you don't want to miss it, please subscribe for free below.

bottom of page