When I saw this headline, I assumed someone in the UK gov must've suffered heatstroke.
Thanks for being the first to push back against this, but your analysis is still far too charitable:
- Most cloud service put files in a virtual recycle bin for 30+ days. After that, whether they actually delete the files or just permanently hide them from you is anyone's guess. So in the short term, literally nothing happens. In the medium term... who knows?
- Drives don't use less power if they contain less data. At least not unless they were completely "emptied" and spun down, but everything is designed to avoid that ever happening. Since users aren't assigned contiguous physical space, deletion would at most cause myriad tiny "gaps" to appear, ready to be filled up by new data (or not). No actual drive would ever use less power.
- Email, specifically, is usually "snapshotted" at regular intervals to recover from catastrophic failure or ransomware. Corporate email is often never truly deleted at all, for regulatory reasons. For many organizations, cold email storage just grows indefinitely, no matter what you yourself delete.
- Since storage and demand for storage are constantly growing, all that this deletion might accomplish is to slow down the rate at which new drives are spun up a few months from now. That is, assuming they are spun up "as needed", and not just on a schedule, or by the thousand when a threshold is triggered (I have no idea).
- Given how "spiky" CPU and GPU power consumption is, both with peaks two orders of magnitude that of a single drive, with wildly fluctuating day-night usage cycles and presumably safety margins to cushion all these spikes, I seriously doubt that a change in the number of drives would even indirectly impact the water intake of a data center. Yes, water intake is regulated dynamically, but I find it it hard to believe it would respond to a change of a tenth or a hundreth of a percent in total power consumption.
Drives DO actually use less power when they're not actively accessing data. And if your data is ancient, it's quite possible nobody is accessing a given drive at all. So a better - if still completely symbolic - recommendation that could save a few drops of water might be: "Enjoy your summer and don't fret about your old data."
Our data for the UK Archer2 supercomputer storage cluster comes out at 30Wh per year for each GB stored. Higher than yours, but similar order of magnitude. Probably because ours includes RAID setup (for fault tolerance) and ancillary equipment such as access network and archive equipment.
Building on the points XP makes: energy use for storage equipment is primarily determined by access rate, rather than total quantity stored. It is not the storage, but the accessing of data, which uses most of the energy. Hence the act of accessing them to check if you want to delete them will (very slightly) increase electricity use.
Furthermore, if the setup is a tiered-storage setup where old unused data sinks back into an archive state (often on tape) unless accessed, then the act of going through old files will pull them forward onto the more energy intensive storage media, where they will remain if you don’t delete them.
As well as 'not sweating the environmental small stuff', another moral of this story is: understand the dynamics of the system you are intervening in, and assess the impact of your proposed intervention as a *change* to the system, not a naive 'this is the energy footprint, if we stop we save it'. This is very common among reasoning about 'carbon footprint', resulting in naive and incorrect claims about green website interventions too.
And that figure is for "to transmit and store data"
(It is also dated from 2020, and includes the detail that that number has been halving every two years, so this possibly could be extrapolated to be 32.75 kWh per GB-year)
However they also use 0.441 litres of water consumed per kWh vs your 1.8L/kWh
"I think it would be really sad if people deleted lots of pictures that were important to them in the mistaken belief that this will help them conserve water."
The really sad and frustrating thing is that I suspect that there's very little overlap between people who need to hear this, and people who will hear it.
Thanks for this (and analyses like this). IMO, so many "lefties" [sic] get so caught up in the minutia that they lose sight of how to actually make things better.
I also wish people would understand our water cycles.
Yeah that's another question. It does seem like data center storage is relatively region specific? Need to dig more on that. Would be funny if the UK were asking its citizens not to store data in Northern Virginia to help with the drought in London
According to Statista in 2024,361.6 billion emails were sent and recieved everyday and its growing. So based on your results, if this figure can be cut in half, 31.575 million litres of water could be saved globally.
Similarly, according to Photutorial in 2024, 1.9 Trillion photos were taken. If 25% of those photos were stored in the cloud, 34 million litres of water would be saved.
65 million litres of water, according to Google Ai, is roughly enough water for a city of 100,000 people for a year.
I mean all global images everyone on Earth take using as much resources as Woodbridge New Jersey just doesn't actually seem significant at all to me. Like if you drove by a large cluster of data centers using the resources of Woodbridge, and someone said "Every single image anyone alive takes of their lives are stored there" would its resource use seem like a lot or a little? To me that seems shockingly small.
If it were all global images everyone on Earth took I would agree, but it's not, it's just 1 years worth.
Whilst I agree that how this has been presented seems a little bizarre, housekeeping emails, photos, videos and other forms of digital waste is important. At scale, it could slow down the need to build, run and maintain millions of storage arrays. It might only make a small impact on climate change or it might do something meaningful, it really depends on your view of the system.
The issue is less about cooling datacentres, more about how much energy they consume, habits and mindset.
We, as a species, need to become much more energy efficient very quickly, we need to create awareness that everytime we use our digital devices there is an impact, but if we all can build +ve habits through awareness of the need to reduce our digital waste rather than building more and more datacentres, I believe it will make a difference.
Deleting emails you dont need, blurry photos, multiple images of the same thing etc. might not be the biggest impact item on the list for everyone, but is something small we can all do, and it all counts.
And tiny things don't actually add up to mattering much for the climate. We'll be able to completely solve the climate crisis without changing how many photos we have saved. Giving people this kind of advice is a really useless distraction from actually meaningful steps they can take to stop climate change. https://andymasley.substack.com/p/for-the-climate-little-things-dont
Yes, I have read these posts and it made me feel a little depressed if i'm being honest.
Agreed, tech is becoming more and more efficient. We have all the tech and know-how we need to address the climate crisis now and to start the recovery, but the reality is that we are pumping out more and more greenhouse gases, yet doing nothing about it.
All the major cloud providers make billions of dollars profit from customers who use their extremely efficient technology, extremely inefficiently. For example, customers who are being charged on-demand, yet leave compute instances, storage, security and other infrastructure on 24*7, idle for long periods, when they only really need to run in office hours. I've seen this first hand and is recognised behaviour in this report (https://info.flexera.com/CM-REPORT-State-of-the-Cloud?). There is no incentive in the cloud providers to help their customers in fixing this and in many cases these organisations who are responsible for this digital waste make so much money that the financial costs to them go unnoticed, as do the costs to the environment.
The similar is true for digital transformation and technology consolidation, which you correctly identify as work which could give the biggest bang for our buck in reducing many organisations carbon footprint.
Netflix saw a market opportunity to destroy Blockbuster, streaming movies directly to homes. This digital transformation was driven by making money, the digitisation and process efficiency gains was not the primary concern. I have also seen many companies persist with environmentally damaging manual practises even though it would literally take one click of a mouse to fix. For example, I have seen financial institutions send out thousands of paper transaction confirmations daily even when a digital equivalent exists. The manual process involves running a batch job to generate the confirmations, printing, putting them in envelopes and shipping them all over the world only to sit in a dark warehouse somewhere for a regulated number of years before being destroyed. This environmentally damaging process can be stopped by an agreement with the financial institutions counterparty to receive the same information digitally via secure email. So why is this not being done on mass? Because there's no will to do so.
The narrative is not that different for application/technology consolidation, this type of activity often takes years to complete, far longer than the average CTO/CIO spends at an enterprise, so what's in it for them to create one CRM system to replace the 10 the enterprise has acquired over the years?
The only way I believe we can make a meaningful difference is to promote cultural change where our values are based on putting the planet first, before profits. A culture where any energy efficient action is championed and putting the planet first is normalised across the globe. If societies across the world begin to value the planet with action, then people with more power will be more inclined to address some of the bigger environmental challenges rather than perpetuating the "burn baby burn" rhetoric.
This requires a culture shift, which isn't complex, but is difficult. Like you said, there's relatively few people looking into this topic at the moment so the first challenge is to create awareness by raising the profile of the conversation, talking about actions taking place and those that are needed.
Happy to carry on the discussion off-line if you want.
I personally think the main question is just "regardless of intent, what was the outcome of digitizing stuff." Netflix was driven by making money, but I think that's fine, and it's good that incentives aligned to make a much more energy efficient way of getting people stuff they want to watch. In my opinion the question of how bad digitalization is for the environment needs to compare the digital thing to what people would be doing otherwise, not to the most optimized version of the digital activity. I think I'm also just more optimistic in general of profit incentives often aligning with higher efficiency stuff. The main shift I'd like to see is more aggressive policy around energy grids rather than personal lifestyle changes around how much energy we consume.
And that figure is for "to transmit and store data"
(It is also dated from 2020, and includes the detail that that number has been halving every two years, so this possibly could be extrapolated to be 32.75 kWh per GB-year)
However they also use 0.441 litres of water consumed per kWh vs your 1.8L/kWh
Wouldn’t we also want to account for water evaporated to spin turbines for electricity generation? My pocket math says this is a much higher contributor then water use for cooling:
-> about 12,000 gallons per MWh or 12 gallons per kWh.
Converting to liters: 12 * 3.785 -> 45 liter per kWh.
Admittedly the UK should have a much better water usage per kWh since their electricity portfolio has a much higher share of renewables (21.4% vs 50.8%). But still going to be an order of magnitude higher! And we are believing that the data centers are in the uk for that cut and not elsewhere.
Right I agree, the thing is the recommendation specifically said "Because data centers use so much water" so I'm ignoring the water cost of electricity, because the report didn't say "Use less electricity because it uses water to generate." I'm responding to the specific framing that it's the data centers that are using the most water here
I aspire to one day write posts this simple yet powerful 🙏🙏🙏
When I saw this headline, I assumed someone in the UK gov must've suffered heatstroke.
Thanks for being the first to push back against this, but your analysis is still far too charitable:
- Most cloud service put files in a virtual recycle bin for 30+ days. After that, whether they actually delete the files or just permanently hide them from you is anyone's guess. So in the short term, literally nothing happens. In the medium term... who knows?
- Drives don't use less power if they contain less data. At least not unless they were completely "emptied" and spun down, but everything is designed to avoid that ever happening. Since users aren't assigned contiguous physical space, deletion would at most cause myriad tiny "gaps" to appear, ready to be filled up by new data (or not). No actual drive would ever use less power.
- Email, specifically, is usually "snapshotted" at regular intervals to recover from catastrophic failure or ransomware. Corporate email is often never truly deleted at all, for regulatory reasons. For many organizations, cold email storage just grows indefinitely, no matter what you yourself delete.
- Since storage and demand for storage are constantly growing, all that this deletion might accomplish is to slow down the rate at which new drives are spun up a few months from now. That is, assuming they are spun up "as needed", and not just on a schedule, or by the thousand when a threshold is triggered (I have no idea).
- Given how "spiky" CPU and GPU power consumption is, both with peaks two orders of magnitude that of a single drive, with wildly fluctuating day-night usage cycles and presumably safety margins to cushion all these spikes, I seriously doubt that a change in the number of drives would even indirectly impact the water intake of a data center. Yes, water intake is regulated dynamically, but I find it it hard to believe it would respond to a change of a tenth or a hundreth of a percent in total power consumption.
Drives DO actually use less power when they're not actively accessing data. And if your data is ancient, it's quite possible nobody is accessing a given drive at all. So a better - if still completely symbolic - recommendation that could save a few drops of water might be: "Enjoy your summer and don't fret about your old data."
I ended up sharing these points as part of the post, thank you!
Happy to have helped!
Amazing yes I can add these all as notes, hammered this out quickly so I'm probably missing a lot of stuff like this
Our data for the UK Archer2 supercomputer storage cluster comes out at 30Wh per year for each GB stored. Higher than yours, but similar order of magnitude. Probably because ours includes RAID setup (for fault tolerance) and ancillary equipment such as access network and archive equipment.
Building on the points XP makes: energy use for storage equipment is primarily determined by access rate, rather than total quantity stored. It is not the storage, but the accessing of data, which uses most of the energy. Hence the act of accessing them to check if you want to delete them will (very slightly) increase electricity use.
Furthermore, if the setup is a tiered-storage setup where old unused data sinks back into an archive state (often on tape) unless accessed, then the act of going through old files will pull them forward onto the more energy intensive storage media, where they will remain if you don’t delete them.
As well as 'not sweating the environmental small stuff', another moral of this story is: understand the dynamics of the system you are intervening in, and assess the impact of your proposed intervention as a *change* to the system, not a naive 'this is the energy footprint, if we stop we save it'. This is very common among reasoning about 'carbon footprint', resulting in naive and incorrect claims about green website interventions too.
Someone sent a Freedom of Information request to the relevant government department and got back this response:
https://www.whatdotheyknow.com/request/deleting_old_emails_and_pictures_2#incoming-3156980
It looks like they use 0.015 kWh per GB per hour (131 kWh per GB-year) which they get from:
https://www.cmswire.com/digital-experience/calculating-the-pollution-effect-of-data/
And that figure is for "to transmit and store data"
(It is also dated from 2020, and includes the detail that that number has been halving every two years, so this possibly could be extrapolated to be 32.75 kWh per GB-year)
However they also use 0.441 litres of water consumed per kWh vs your 1.8L/kWh
They also use an email size of around 1-1.5MB
I don't know what the fuck we're doing either, and I live here.
Love this weird, thanks for doing the almost unimaginable math.
Sheila Lewis
According to Google's 2025 sustainability report they only use about 1l per MWh, which is about 1000x better than even this article said.
Amazing
"I think it would be really sad if people deleted lots of pictures that were important to them in the mistaken belief that this will help them conserve water."
The really sad and frustrating thing is that I suspect that there's very little overlap between people who need to hear this, and people who will hear it.
Thanks for this (and analyses like this). IMO, so many "lefties" [sic] get so caught up in the minutia that they lose sight of how to actually make things better.
I also wish people would understand our water cycles.
Imagine if people listened to them. That would really help the productivity growth figures!
Do we even know if Google & co generally store UK users' emails within the UK?
Yeah that's another question. It does seem like data center storage is relatively region specific? Need to dig more on that. Would be funny if the UK were asking its citizens not to store data in Northern Virginia to help with the drought in London
You made an error in your opening para.
5.5 Watts * 24 hours in a day = 132 Watt-hours per day. 132 * 365 = 48.2 kWh per year for 18 TB, or 1 kWh per year for every 370 MB of data stored.
That should be 370 GB. Your estimates of how efficient data deletion is are therefore three orders of magnitude too generous.
Great catch! Dumb to stumble over that. I’ll go back and fix
Came here to say this
Fixed, thanks both!
Interesting read.
According to Statista in 2024,361.6 billion emails were sent and recieved everyday and its growing. So based on your results, if this figure can be cut in half, 31.575 million litres of water could be saved globally.
Similarly, according to Photutorial in 2024, 1.9 Trillion photos were taken. If 25% of those photos were stored in the cloud, 34 million litres of water would be saved.
65 million litres of water, according to Google Ai, is roughly enough water for a city of 100,000 people for a year.
I think that is quite significant.
I mean all global images everyone on Earth take using as much resources as Woodbridge New Jersey just doesn't actually seem significant at all to me. Like if you drove by a large cluster of data centers using the resources of Woodbridge, and someone said "Every single image anyone alive takes of their lives are stored there" would its resource use seem like a lot or a little? To me that seems shockingly small.
If it were all global images everyone on Earth took I would agree, but it's not, it's just 1 years worth.
Whilst I agree that how this has been presented seems a little bizarre, housekeeping emails, photos, videos and other forms of digital waste is important. At scale, it could slow down the need to build, run and maintain millions of storage arrays. It might only make a small impact on climate change or it might do something meaningful, it really depends on your view of the system.
The issue is less about cooling datacentres, more about how much energy they consume, habits and mindset.
We, as a species, need to become much more energy efficient very quickly, we need to create awareness that everytime we use our digital devices there is an impact, but if we all can build +ve habits through awareness of the need to reduce our digital waste rather than building more and more datacentres, I believe it will make a difference.
Deleting emails you dont need, blurry photos, multiple images of the same thing etc. might not be the biggest impact item on the list for everyone, but is something small we can all do, and it all counts.
I actually completely disagree and have written about this. Digital infrastructure is often replacing much more wasteful activities. https://andymasley.substack.com/p/computing-is-efficient
And tiny things don't actually add up to mattering much for the climate. We'll be able to completely solve the climate crisis without changing how many photos we have saved. Giving people this kind of advice is a really useless distraction from actually meaningful steps they can take to stop climate change. https://andymasley.substack.com/p/for-the-climate-little-things-dont
Yes, I have read these posts and it made me feel a little depressed if i'm being honest.
Agreed, tech is becoming more and more efficient. We have all the tech and know-how we need to address the climate crisis now and to start the recovery, but the reality is that we are pumping out more and more greenhouse gases, yet doing nothing about it.
All the major cloud providers make billions of dollars profit from customers who use their extremely efficient technology, extremely inefficiently. For example, customers who are being charged on-demand, yet leave compute instances, storage, security and other infrastructure on 24*7, idle for long periods, when they only really need to run in office hours. I've seen this first hand and is recognised behaviour in this report (https://info.flexera.com/CM-REPORT-State-of-the-Cloud?). There is no incentive in the cloud providers to help their customers in fixing this and in many cases these organisations who are responsible for this digital waste make so much money that the financial costs to them go unnoticed, as do the costs to the environment.
The similar is true for digital transformation and technology consolidation, which you correctly identify as work which could give the biggest bang for our buck in reducing many organisations carbon footprint.
Netflix saw a market opportunity to destroy Blockbuster, streaming movies directly to homes. This digital transformation was driven by making money, the digitisation and process efficiency gains was not the primary concern. I have also seen many companies persist with environmentally damaging manual practises even though it would literally take one click of a mouse to fix. For example, I have seen financial institutions send out thousands of paper transaction confirmations daily even when a digital equivalent exists. The manual process involves running a batch job to generate the confirmations, printing, putting them in envelopes and shipping them all over the world only to sit in a dark warehouse somewhere for a regulated number of years before being destroyed. This environmentally damaging process can be stopped by an agreement with the financial institutions counterparty to receive the same information digitally via secure email. So why is this not being done on mass? Because there's no will to do so.
The narrative is not that different for application/technology consolidation, this type of activity often takes years to complete, far longer than the average CTO/CIO spends at an enterprise, so what's in it for them to create one CRM system to replace the 10 the enterprise has acquired over the years?
The only way I believe we can make a meaningful difference is to promote cultural change where our values are based on putting the planet first, before profits. A culture where any energy efficient action is championed and putting the planet first is normalised across the globe. If societies across the world begin to value the planet with action, then people with more power will be more inclined to address some of the bigger environmental challenges rather than perpetuating the "burn baby burn" rhetoric.
This requires a culture shift, which isn't complex, but is difficult. Like you said, there's relatively few people looking into this topic at the moment so the first challenge is to create awareness by raising the profile of the conversation, talking about actions taking place and those that are needed.
Happy to carry on the discussion off-line if you want.
I personally think the main question is just "regardless of intent, what was the outcome of digitizing stuff." Netflix was driven by making money, but I think that's fine, and it's good that incentives aligned to make a much more energy efficient way of getting people stuff they want to watch. In my opinion the question of how bad digitalization is for the environment needs to compare the digital thing to what people would be doing otherwise, not to the most optimized version of the digital activity. I think I'm also just more optimistic in general of profit incentives often aligning with higher efficiency stuff. The main shift I'd like to see is more aggressive policy around energy grids rather than personal lifestyle changes around how much energy we consume.
Someone sent a Freedom of Information request to the relevant government department and got back this response:
https://www.whatdotheyknow.com/request/deleting_old_emails_and_pictures_2#incoming-3156980
It looks like they use 0.015 kWh per GB per hour (131 kWh per GB-year) which they get from:
https://www.cmswire.com/digital-experience/calculating-the-pollution-effect-of-data/
And that figure is for "to transmit and store data"
(It is also dated from 2020, and includes the detail that that number has been halving every two years, so this possibly could be extrapolated to be 32.75 kWh per GB-year)
However they also use 0.441 litres of water consumed per kWh vs your 1.8L/kWh
They also use an email size of around 1-1.5MB
Wouldn’t we also want to account for water evaporated to spin turbines for electricity generation? My pocket math says this is a much higher contributor then water use for cooling:
https://www.publicpower.org/periodical/article/how-much-water-our-electricity-uses
-> about 12,000 gallons per MWh or 12 gallons per kWh.
Converting to liters: 12 * 3.785 -> 45 liter per kWh.
Admittedly the UK should have a much better water usage per kWh since their electricity portfolio has a much higher share of renewables (21.4% vs 50.8%). But still going to be an order of magnitude higher! And we are believing that the data centers are in the uk for that cut and not elsewhere.
https://www.renewableuk.com/news-and-resources/press-releases/official-stats-show-renewables-generated-over-half-uk-s-electricity-for-the-first-time-in-2024/
https://www.eia.gov/tools/faqs/faq.php?id=427&t=3
Right I agree, the thing is the recommendation specifically said "Because data centers use so much water" so I'm ignoring the water cost of electricity, because the report didn't say "Use less electricity because it uses water to generate." I'm responding to the specific framing that it's the data centers that are using the most water here
They say
“ Delete old emails and pictures as data centres require vast amounts of water to cool their systems.“
Whoops, didn’t catch how specific that was.