We tried this service out and it claims that only 0.05 g of CO2 is produced every time someone visits this web page:
This result is only an approximation and for the page in question.
How does it work?
Website Carbon™ provides an estimate of the carbon emissions associated with a single load of a webpage.
When you add a URL to the calculator and click calculate a number of things happen.
Firstly it loads the webpage in the same way a human using a browser would. This gives the calculator it’s first data point, the data transfer in KB. This is the amount of data it takes for a page to load.
The calculator uses this amount of data transfer and estimates the amount of emissions based on V4 of the Sustainable Web Design Model.
The model uses the global average grid intensity to estimate the emissions associated with hosting. It checks the Green Web Foundation Directory to verify if hosting uses renewable energy.
With those checks done, Website Carbon™ then adds in an amount of emissions associated with the transmission network. This is based on the amount of electricity required to transmit the data from the host to your device.
Finally, the calculator adds the emissions associated with you loading the page. As with hosting, this calculation uses a global average grid intensity. You should note that the type of device is not taken into consideration.
Based on the above factors, the webpage will be given an amount of CO2e emissions in grams and an associated rating.
But not all pages are equal. The Sustainable Web Design by Tom Greenwood (an authority on such matters, so he should know better) page on Amazon scores less well with tests on image heavy pages or leaning on third party services consistently getting bad ranking1! We should take note also that pages are 7 times heavier than a decade ago2, so it still makes sense to reduce image size whenever possible. We can all save the planet if we stop dumping digital assets in cyberspace3. Cyberspace is a data facility consuming energy and water and land somewhere near you4.
Data centres are not a land of milk and honey. While there’s no single, standard UK assessment that jointly totals up land take + biodiversity loss + (low) permanent jobs for data centres, here’s the picture:
- Policy tests in Scotland (NPF4):
• Policy 3 forces “demonstrably better” biodiversity for national/major/EIA developments—effectively a biodiversity gain test.
• Policy 5(b) protects prime agricultural land unless there’s a specific locational need and no alternative.
These bite on data centres just like any other major development. (NatureScot) - Employment reality check: Commons Library summarises UK evidence—typical direct on-site jobs are small (≈20–88 FTE per site; midpoint 54), with construction and indirect jobs claimed separately. Planning statements often cite the latter, but decision-makers can and do down-weight them.
- Forthcoming national lens (England): the Government plans a National Policy Statement for data centres; like other NPSs, it should come with an Appraisal of Sustainability/Habitats that considers land use and biodiversity. Until that lands, case-by-case planning balance applies.
Bottom line: right now if we want an integrated “public-interest test,” we have to build it yourself for each case—combine the EIA biodiversity/soil findings, NPF4 Policy 3 & prime-land tests, and direct jobs evidence to show whether a site actually stacks up locally. (NatureScot)
Data Centres – Our Burns are already Dry!
There is growing evidence that data centres (especially those supporting AI) are consuming land, water, energy, and biodiversity in ways that are not often fully accounted for. Below are some of the key facts + caveats (because some figures are estimates, and transparency is uneven):
Key Facts
- Water use & cooling demand
- Globally data centres consume ~560 billion litres of water per year. (EthicalGEO)
- In places like Sydney, Australia, recently approved data centre projects are expected to consume up to 9.6 gigalitres of clean water annually, ~2% of Sydney’s maximum water supply. (Reuters)
- New AI workloads are pushing water demand even higher; projections estimate global AI-infrastructure demand could require 4.2–6.6 billion m³ of water withdrawal annually by 2027. (OECD AI)
- Some new data centres are being built in areas already suffering water stress, which exacerbates pressures on local supplies. (EthicalGEO)
- Energy & electricity + indirect land use
- Data centre electricity consumption is growing quickly. One policy paper estimates that data centres used ~415 TWh in 2024, and that could more than double by 2030 under current AI growth scenarios. (Federation of American Scientists)
- Because electricity generation is often based on fossil fuels (or uses water for cooling at power plants), the indirect footprint (carbon, water, land) expands. (EESI)
- Land use & location issues
- Big sites take up large areas (e.g. the Google hyperscale data centre planned in Thurrock, UK will cover 52 hectares) which has implications for land use (agriculture, biodiversity). (The Guardian)
- Two-thirds of new data centres built or in development since 2022 are reportedly located in places already experiencing water stress. So these land + water-use choices often compound other environmental vulnerabilities. (EthicalGEO)
- Biodiversity, materials, and mining
- Electronics & chip production depend on rare/critical minerals, whose extraction often causes habitat destruction, pollution, etc. (UNEP – UN Environment Programme)
- Land clearing for large data centre campuses and associated infrastructure (power lines, cooling ponds/towers) may fragment habitat. Also risk from water draw downs affecting ecosystems. Less data on quantifying this globally in consistent units.
- Employment & local economic benefit often modest relative to environmental/ infrastructural cost
- Many data centres don’t deliver large permanent employment: many workers are needed during construction, but operational staff numbers tend to be small. (You asked earlier about this re UK.) This applies elsewhere too: large capex, limited long-term labour density.
- Also, local benefits (tax, jobs) are sometimes promised but may be offset by environmental cost (loss of agricultural productivity, water competition, ecosystem services).
Caveats & Gaps
- Many data centre operators do not fully disclose water usage, biodiversity impacts, or land area/vegetation loss. So much of the evidence is partial. E.g. only a minority report onsite water use, much less indirect water use or biodiversity metrics. (EESI)
- Definitions vary: “water usage” vs “water withdrawal” vs “water consumption” (difference: do you count water returned, evaporated, lost?) This matters especially in drought-prone areas. (OECD AI)
- Land use impacts and biodiversity loss are harder to quantify globally (for example, what land was previously agricultural, or forest, or wetland etc.), or what the “opportunity cost” is (lost food production, habitat etc.).
- Some centres do try to mitigate (e.g. siting in cold climates, using reclaimed water, efficient cooling, using renewable energy). But mitigation is patchy and often dependent on regulatory pressure.
Bottom Line
Yes, data centres + AI are not just digital issues: they have real physical/ecological footprints. We risk stretching critical resources (land, water, biodiversity) especially in places already under stress, and the local benefits often don’t match the scale of consumption.
OK, we admit to having used AI to craft this article.
- Image-heavy layouts and third-party widgets (ads, analytics, consent banners, embeds) consistently tank both performance and carbon scores ↩︎
- HTTP Archive shows median page weight has exploded from ~505 KB (mobile, Oct 2014) to ~2.3 MB (Oct 2024), with images still the biggest slice and third-party scripts dragging interactivity. ↩︎
- The fix is boring but effective: ship fewer bytes—compress/resize (AVIF/WebP), lazy-load, and cull third-party tags. ↩︎
- And remember: “the cloud” is just buildings that burn power and draw water—typical data-centre water use is ~1.9 L per kWh (with wide variation)—so every kilobyte you don’t send is energy and water you don’t waste. ↩︎