How bad is our data for the planet?

Episode Summary

Executive Summary: The episode examines the environmental footprint of data centers—especially their electricity and water use—through visits to efficient facilities in Norway and South Africa. It finds individual email/streaming habits have tiny emissions, but global data-center demand is already enormous and rising, driven partly by AI. The key takeaway: efficiency helps, but real climate progress requires low-carbon power and careful constraint on fossil-fuel use.

Main Topics: Global data-center energy demand (Priority: 5/5): The show opens by framing the central question: how much electricity do data centers consume worldwide, and how does that compare to national usage? It emphasizes that the cloud depends on physical infrastructure with substantial energy needs. Water use and cooling (Priority: 5/5): The episode explores how server cooling drives water consumption, including estimates of worldwide daily use and the importance of climate and location in determining sustainability. Norway’s mountain data center model (Priority: 4/5): A high-security underground facility in Norway demonstrates an ultra-efficient approach: renewable electricity, cold fjord water for cooling, and heat-reuse potential. Limits of “green” claims (Priority: 5/5): Mike Berners-Lee argues that even renewable-powered data centers are not automatically climate-neutral if they divert scarce clean electricity from other uses. African data-center growth and renewables (Priority: 4/5): A South African operator shows how regions with coal-heavy grids and power instability are using solar, free-air cooling, and power purchase agreements to expand more sustainably. Personal data habits vs system-scale impact (Priority: 4/5): The episode tests the environmental impact of everyday email and streaming use, concluding that individual footprints are relatively small compared with the much larger systemic footprint of data infrastructure. AI and future demand (Priority: 5/5): The discussion ends with concern that AI will accelerate data-center growth, making efficiency, renewable supply, and broader constraints on fossil fuel use increasingly important.

Key Arguments: Data centers are not abstract “cloud” systems; they are energy- and water-intensive physical facilities that already consume a significant share of national and global electricity. Global data-center electricity use is estimated at 240 to 340 TWh per year, excluding cryptocurrencies, and is likely rising as digital demand increases. Data-center energy demand can grow even when hardware becomes more efficient, because efficiency lowers costs and encourages more use. Water use for cooling is a major concern; globally, data centers may use around 6 billion litres per day, but local scarcity determines how damaging that is. Location matters: cold climates, renewable electricity, and access to water make it easier to run low-impact data centers. A data center powered by renewables is not automatically “green” if it simply takes renewable electricity away from other users instead of adding new clean capacity. In regions like South Africa, renewable energy can serve both sustainability and reliability goals by reducing dependence on unstable fossil-heavy grids. Individual email storage and streaming have modest carbon footprints; the larger climate issue lies in the overall scale of ICT infrastructure and device manufacturing/use. The best long-term solution is not efficiency alone, but constraining fossil-fuel use so efficiency gains actually reduce emissions. AI will likely increase data-center demand sharply, so future planning must integrate renewable supply, heat reuse, and resource-aware siting.

Data Points: Estimated global data-center electricity use: 240 to 340 terawatt hours per year - Mike Berners-Lee’s rough estimate for data centers worldwide, excluding cryptocurrencies. Share of global electricity use: 1.3% to 1.7% - Approximate proportion of worldwide electricity consumed by data centers. Cryptocurrency add-on: about a third again - Berners-Lee said cryptocurrencies likely add roughly one-third more to the data-center electricity total. UK-equivalent electricity use: about the same as the United Kingdom each year - Comparison used to contextualize global data-center electricity demand. Data-center workload growth since 2015: tripled - Berners-Lee described how demand for data-center computing has expanded since 2015. Resulting energy increase since 2015: 20% to 70% - Despite much higher workloads, efficiency gains limited growth in total energy use. Estimated global water use: 6 billion litres per day - Rough estimate of water used by the world’s data centers for cooling. Per-person water equivalent: about three quarters of a litre per person per day - Converted from the global daily water estimate. Irish data-center electricity share: almost 20% - Data centers’ share of electricity use in Ireland, where the listener lives. Norwegian facility size: 25 megawatts - Green Mountain’s mountain-based data center capacity. Number of servers in Norway facility: 50,000 to 60,000 - Approximate server count inside the 25 MW facility. Cooling efficiency claim: 3 kilowatts of power for 1,000 kilowatts of cooling - Green Mountain’s description of its cooling system efficiency. Norwegian renewable mix: 100% renewable; 96% hydro, 3% wind, 1% other - Power sources supplying the Norwegian grid used by the facility. Water depth for cooling intake: 100 meters - Depth of pipes drawing fjord water for cooling at the Norwegian site. Heat produced at Norwegian site: 80 gigawatt hours - Potential reuseable heat output mentioned by Svein Hagaseth. South Africa on-site solar: 814 kilowatts - Solar installed at the Johannesburg data center roof. Planned additional solar: 110 kilowatts - A further rooftop solar project mentioned by Tulani Ngube. Power purchase agreement: 20 megawatts - Off-site renewable energy contract intended to supply South African operations. Load shedding: minimum two hours a day - Scheduled electricity blackouts in South Africa described as a daily norm. Africa data-center capacity growth: five times by 2030 - Estimate of capacity expansion needed to meet demand across Africa. Africa data users vs server space: more data users than the USA; server space equal to Switzerland - Comparison used to show the imbalance between demand and infrastructure. Inbox size: over 59,000 emails; 10.75 GB - The presenter’s own email storage footprint used for comparison. Email storage carbon footprint: 0.2 kilograms per year - Berners-Lee’s estimate for the carbon footprint of storing all those emails. Streaming footprint: 20 grams per viewing hour - Estimated carbon footprint per hour of video streaming. ICT share of global emissions: about 4% - Berners-Lee’s estimate of the carbon footprint of ICT overall.

Pivotal Quotes: "We think it's something like six billion litres of water per day used by the world's data centers." — Mike Berners-Lee: He explains the scale of water consumption tied to cooling global data-center infrastructure. "The climate doesn't care how much renewable energy we've got. What the climate cares about is how little fossil fuel we're burning." — Mike Berners-Lee: He challenges the idea that renewable power alone makes data-center growth environmentally harmless. "Every kilowatt hour that we bring as renewable energy to our data center is a kilowatt hour that we are releasing to the communities or to the utility to use as they need it." — Tulani Ngube: He describes how Africa Data Centers’ solar procurement adds clean power rather than just reallocating existing supply.

Implications: Listeners should not panic about personal email or streaming, but they should recognize that data infrastructure is a major and growing climate issue. The industry’s future depends on adding clean power, improving efficiency, reusing waste heat, and curbing fossil-fuel dependence—especially as AI expands demand.