The rapid evolution of artificial intelligence (AI) has spurred an unprecedented race for computational power globallyHowever, the expansion of this field comes with an intricate challenge: ensuring resource guarantees, particularly in the context of data centers, which are notorious for their energy consumptionThe effectiveness and stability of the computational industry hinge on a systematic approach that fosters synergy among different resources, elevating their utility significantly.

As AI technologies continue to mature at breakneck speeds, the demand for computational resources is simultaneously escalatingOne of the significant hurdles in this transition is the enormous energy requirementsFor example, consider the OpenAI GPT-3 model, which has 175 billion parameters and consumes approximately 1.287 million kilowatt-hours of electricity per training cycleWith the introduction of GPT-4 — boasting over 1.8 trillion parameters — it is anticipated that the energy consumption will rise dramatically

Advertisements

Estimates suggest that global data center electricity consumption might exceed Japan's total energy usage, estimated at around one trillion kilowatt-hours, by 2026. This alarming trajectory calls for urgent attention to manage the increasing energy consumption alongside the corresponding carbon emissions, improving the efficiency of energy consumption in data centers.

China, as a leader in the global energy supply landscape, has made remarkable strides in establishing robust and extensive clean energy infrastructureThis presents a unique opportunity to support the burgeoning computational industry through the integration of abundant green powerTo optimize this potential, planning that merges computational and energy development is crucialInvesting in innovative models for integrated green power supply for data centers – such as optimizing energy sources, grid networks, and load management – becomes paramount

Advertisements

Moreover, R&D efforts should concentrate on reducing energy loss at various stages within data centers and effectively harnessing waste heat.

Interestingly, while the focus has predominantly been on energy, a recent report from JPMorgan has highlighted another often-overlooked aspect: the colossal water consumption associated with data centersWater is essential for cooling systems, humidity control, and other operational necessitiesAs demand for cooling escalates with advancements in technology such as liquid cooling, water usage is ramping up, showing significant growth rates that compel scrutinyProjections indicate a 6% annual increase in global water usage by data centers from 2017 to 2022, with expectations of maintaining a growth rate of at least 5.5% through 2030.

In China, the supply-demand balance for water is tenuous, particularly in the western regions, where water scarcity is acute

Advertisements

Consequently, developing data centers in these areas necessitates a multifaceted approach that champions water conservationOutreach programs to promote water-saving techniques, stringent regulations, and standards around water use, and incentives for exploring unconventional water sources such as reclaimed water, rainwater, and mine water can enhance water resource managementResearch and development into high-efficiency, low-water-use cooling technologies, particularly in arid zones, is also essential.

Additionally, the critical inputs to computational power development extend beyond water and energy to encompass materials such as copper, gold, and silverCopper, in particular, is indispensable as it is used extensively across various components, chips, servers, and switches in data centers, and plays a central role in electrical systemsCurrent estimates suggest that about 70% of copper usage in the data center sector is geared towards electrical units, with an expected surge in demand — from the current annual consumption of about 100,000 tons to over one million tons by 2027. This stark increase positions the data center sector as a significant contributor to global copper demand growth.

The dynamics of the copper market are influenced by several factors, notably the rising prices catalyzed by the shift toward electric vehicles and other energy transition initiatives

While demand surges, the domestic supply of copper in China remains insufficient, as the country holds less than 5% of the world’s copper reservesThis vulnerability, compounded by the concentration of global copper resources and increasing industry monopolization, suggests that the supply-demand balance will remain tense over the long termTherefore, bolstering domestic production capabilities, enhancing exploration practices, and elevating the efficiency of low-grade copper ore utilization are essential steps in securing a stable supply chain for this key material.

Efforts must also be directed toward establishing a comprehensive recycling system for copper, standardizing management practices, and developing a robust reserve frameworkThis involves establishing fair international trade practices to ensure a stable global supplyThe intricate interdependencies between the various requirements for computational power underscore the necessity of a cohesive strategy

alefox