[Shenzhen, China, March 23, 2023] Huawei held the Top 10 Trends of Data Center Facilities conference with the theme of "Smart DC, Building the Green Future" and released a white paper. At the conference, Fei Zhenfu, President of Huawei Data Center Facility Business, shared Huawei's explorations and insights on the trends of data center facilities, such as low carbon, rapid deployment, high density, storage and backup integration, and distributed cooling, from the perspectives of data center industry development and technological innovations.
Fei Zhenfu, President of Huawei Data Center Facility Business, shared Huawei's insights on the top 10 trends of data center facilities.
According to Mr. Fei, in the era of digital intelligence, global computing requirements will increase exponentially, driving a boom in data center construction. To achieve carbon neutrality, the data center industry must accelerate energy conservation, carbon emission reduction, and green transformation. This presents challenges and opportunities. Based on its practices and industry insights, Huawei identified the top 10 trends of data center facilities for the next five years, guiding low-carbon, sustainable development of the data center industry.
Trend 1: Low Carbon
With carbon neutrality as a global mission, data centers will speed up a green, low-carbon transition. Clean energy, such as PV, wind, and hydropower, will be more widely used in data centers. It is estimated that, by 2027, the utilization rate of green power will exceed 50% thanks to the use of green and low-carbon energy sources. Moreover, advanced energy conservation solutions and technologies such as waste heat recovery will be adopted to improve the energy efficiency and recovery rate, continuously advancing low-carbon data centers.
Trend 2: Sustainability
Whether a data center is sustainable is evaluated by the utilization efficiency of energy and resources and the impact on the environment. In addition to the power usage effectiveness (PUE), data centers will also be measured by more comprehensive indicators, such as renewable energy utilization, water usage effectiveness (WUE), carbon usage effectiveness (CUE), space usage effectiveness (SUE), grid usage effectiveness (GUE), material recovery rate, and lifetime contaminant emissions.
Trend 3: Fast Deployment
Thanks to the accelerated development of AI and HPC, global computing requirements are snowballing with a compound annual growth rate (CAGR) of over 50% in the next five years. Computing services need to meet skyrocketing demands within a short period. For instance, ChatGPT attracted 100 million monthly active users just two months after its launch. To cope with the skyrocketing demands, data centers must be rapidly deployed, and their construction period will drop from 12 months to 6 months or even shorter.
Trend 4: High Density
According to Moore's Law, chips are upgraded every two years. Their performance and power consumption will significantly improve, driving data centers from low to high density. With the continuous increase in computing performance and power density, cloud data centers will become the mainstream, and diversified computing collaboration will become the norm. The standard power density per IT rack will increase from 6–8 kW to 12–15 kW in 2027. The power density in supercomputing and intelligent computing centers will increase to 30 kW or higher.
Trend 5: Elasticity
A conventional data center requires a one-off construction, resulting in a high capital expenditure (CAPEX), complex capacity expansion, and subsequent mismatch between facilities and IT equipment. The lifetime of IT equipment is generally four to five years, with its power density doubling every five years, while that of data center facilities is 10 to 15 years. In the future, data centers will be more flexible regarding upgrades. One generation of facilities will support the power evolution over two to three generations of IT equipment. A standard design will enable on-demand deployment and elastic capacity expansion.
Trend 6: Prefabrication
A prefabricated and simple architecture will be used to cope with challenges conventional data centers face, such as long construction periods and complex O&M. For example, a PowerPOD is prefabricated and adopts a modular design. Core components are pre-installed and pre-commissioned in the factory. The onsite delivery time will be shortened from two months to two weeks, accelerating service rollouts. In the future, a data center building will house over 1000 racks, and a campus will accommodate over 10,000 racks. If a prefabricated architecture is used, the construction period will be shortened from 12 months to 6 months for a data center with 1000 racks, enabling quick delivery and on-demand deployment.
Trend 7: Storage and Backup Integration
Data centers are energy-hungry, making it necessary to apply renewable energy and energy storage. As renewable energy is intermittent and random, feeding more renewable energy to power grids will compromise the stability of the power supply. Electricity price differences between peak and off-peak hours widen, and staggering electricity usage is encouraged. All these mean opportunities for energy storage systems (ESSs). ESS deployments in data centers will integrate short-term backup power systems. The systems can take part in frequency regulation and peak shaving to improve the stability of power grids. Peak shaving can also improve the grid usage effectiveness (GUE) so that more IT racks can be deployed to increase revenue.
Trend 8: Distributed Cooling
A conventional centralized cooling system has disadvantages such as complex O&M, a large fault domain, and low reliability. Once a fault occurs, the reliability of the entire data center will be affected. In contrast, a distributed cooling system uses cooling sources for each data hall and adopts a redundancy configuration as required. A single point of failure will not affect overall system operations. Such a system can also be quickly delivered. In the future, distributed cooling systems will gradually replace centralized cooling as the preferred cooling solution for large data centers due to their flexible architecture and high reliability.
Trend 9: Smart O&M
As digital and AI technologies develop rapidly and data centers keep growing in scale, O&M will become increasingly complex. Thus, smart O&M is provided to improve data center resource utilization and reliability. Intelligent operations will automatically determine the data center asset statuses to maximize asset value. Intelligent maintenance will identify faults and optimize energy conservation in real time to improve O&M efficiency, achieving predictive maintenance and optimal energy efficiency.
Trend 10: Security & Trustworthiness
Security and reliability are top priorities for data center facilities. With enhanced digitalization and intelligence, data centers will strengthen their hardware reliability, software security, system resilience, safety, privacy, and availability. AI technologies will implement predictive maintenance on components, devices, and systems at the hardware level. The hierarchical defense will be provided at the software level. A combination of software and hardware measures will bolster the reliability of data centers.
With insights on future trends, we will drive the data center industry toward green transformation and sustainable development. Looking ahead, Huawei will strengthen technological innovation to lead the industry development and offer more competitive products and solutions to build a solid digital foundation for a greener, low-carbon future for the data center industry.
For more information, please download the White Paper on the Top 10 Trends of Data Center Facilities at https://digitalpower.huawei.com/attachments/index/e8c4f4fd705a4d48ad229502e3eb668b.pdf.
