Nature Positive Role of the Technology Sector 2025

Page 59 of 84 · WEF_Nature_Positive_Role_of_the_Technology_Sector_2025.pdf

Germanium, gallium and arsenic can be used as alternatives to silicon for specific use cases. Germanium is produced as a by-product of zinc smelting and coal burning, with electronics and photovoltaics comprising ~25% of the global market.169 Gallium is a by-product of aluminium and zinc smelting, with semiconductors making up 40- 45% of the global market.170 Arsenic is a by-product of copper, lead and gold smelting, and electronics and technology account for ~8% of the global market.171 Bauxite (aluminium) is primarily surface mined and therefore has similar nature impacts as silicon and copper.172 Zinc, coal, lead and gold are primarily mined underground.173,174,175,176 Underground mining still impacts biodiversity and ecosystems through land use but tends to have more confined impact compared to surface mining, reducing the need for land rehabilitation. Water management is required to avoid the formation of mine water reservoirs that can contain toxic minerals and chemicals. Underground mining creates less airborne dust at the surface level but can release trapped gasses from the ore such as methane, a potent GHG.177 Of course, the process creates different risks from surface mining for the workers involved, which must also be considered. For more details on the impacts and dependencies of metals and mining, please refer to the World Economic Forum’s report: Nature Positive: Role of the Mining and Metals Sector. Data centres Electricity use Data centres require substantial energy to power their servers 24/7 and that requirement is growing rapidly. In 2023, global data centre energy loads totalled ~50-55 GW.178,179 By 2028, predictions suggest global data centre energy loads of ~95- 140 GW,180 equivalent to the total power load of Japan in 2023. Variability in the forecast is driven by supply bottlenecks, potential operational efficiency improvements and a range of potential scenarios for the advancement and rate of adoption of AI. Much of the projected increase is driven by AI. While average server rack power densities increased from 3 kW in 2011 to 12 kW in 2022, rack densities exceeding 20 kW were uncommon.181 Following the widescale introduction of capable generative AI models in 2022, server rack densities have seen increases at the hyperscale level where AI models are trained and operated. Hyperscale server rack densities rarely fall below 30 kW and can be as high as 100 kW or even higher.182 The impact of AI is visible in Figure A3, as hyperscale rack server densities are expected to continue increasing while enterprise/ co-location rack densities remain largely constant. Data centres – server rack density, 2023-2027 FIGURE A3 01020304050 202336 12 202542 13 13 202645(kW/rack) 20271449 20241239+8% Enterprise & colocation Hyperscale Compound annual growth rate (CAGR) Source: Obin, A. et al. (2024). Who Makes the Data Centre. Bank of America (BofA) Securities. Nature Positive: Role of the Technology Sector 59
Ask AI what this page says about a topic: