Second article on the lessons I’ve learned from the 2025 edition of Lecko’s State of the Art of Internal Transformation. After the hyperconnectivity that was the subject of my first article (Hyperconnectivity in the workplace: digital becomes a burden) I’d like to talk about a subject that interests me as much as it annoys me: sustainable digital and, above all, its environmental impact.
The reason for this? Because, on the one hand, I see companies communicating massively about their environmental ambitions, proudly announcing that their employees no longer fly, but on the other hand not only turning a blind eye to the impact of their digital work environments, but also boasting about trying to use AI for anything and everything, when in fact we’re reaching new heights in terms of environmental impact.
As a reminder ([FR]Hyperbolic discounting and the environmental cost of AI):
– In the United States, the data centers that host this data already consume 2 to 3% of the country’s total electricity, and this consumption is set to triple in the next 5 years.
– 40-50% of the energy consumed is used to produce data, while 30-40% is used to coolthese centers.
– Nearly 40% of the world’s data centers are likely to suffer power outages, forcing them to curb their energy consumption as early as 2027, and therefore unable to meet demand estimated to grow by 160% over the next two years.
Already in 2019, a Green IT study estimated the digital industry’s primary energy consumption at4.2% of global consumption ([FR]Management’s overwhelming responsibility for the impact of lA).
At the time, the digital industry generated around 3.5% of global emissions, growing at 8% per (+8%/year from 2014 to 2019), which should take us to 7% by 2025. But that was before AI.
By comparison, commercial aviation accounts for around 2.5% to 2.9% of global CO2 emissions, and if we take into account the sector’s entire impact, we reach 6% ([FR] Is it true that air travel is responsible for only 3% of global CO2 emissions, i.e. less than digital?).
Knowing that one of the two sectors is under the spotlight and constantly improving its impact, while the other gives the impression that it can do whatever it wants with impunity because no one cares, and that it also conveys an image of modernity, there is cause for concern.
In fact, a major player in the sector told me a few years ago , “we’re doing sh…on a large scale, but cool sh… and as long as everyone’s bashing aviation because it’s a symbol, we’re fine“.
Digital technology has thus become a central pillar of our professional and personal lives. Although it is presented as a factor of simplification and modernity, it skilfully masks a massive and often underestimated environmental cost. Behind dematerialization and so-called efficiency lies an exponential consumption of energy and resources.
But let’s see what the study has to say on the subject…
The triply alarming environmental impact of digital technology
The environmental impact of digital technology has three main causes.
The first is greenhouse gases. Our digital devices, servers and infrastructures are powered by energy which, in the majority of cases, relies on fossil fuels, at least worldwide.
Then there’s the fact that our digital equipment is built using rare materials that are difficult to recycle. Extracting them is energy-intensive and polluting, and recycling them remains a pipe dream.
Finally, there is the invisible consumption of water. Data centers, necessary for data management, use massive quantities of water for cooling, exacerbating water stress in certain regions.
I’d like to add a few figures here, to give you a better idea of what’s going on, with a case that stirred public opinion at the time. Microsoft’s data center in Middenmeer, North Holland, consumed 84 million liters of drinking water in 2021, compared with initial forecasts of 12 to 20 million liters per year (water in, data out: microsoft underestimates dutch center’s thirst). At the time, the Netherlands officially declared a drought due to a particularly dry summer.
And if you think ChatGPT is “cool”, consider that “while training a large language model costs 126,000 liters of water, a conversation with ChatGPT uses about half a liter” ([FR]The Three Shortages of Generative Artificial Intelligence and ChatGPT And Generative AI Innovations Are Creating Sustainability Havoc).
As a well-informed person in a major French business told me, “when you see the financial and environmental cost of each AI query, it’s obvious that we won’t be able to give everyone unlimited access to this technology, or else we’ll have to explain to people that we’ve been kidding them with our environmental engagements”.
Some key figures
The carbon footprint of digital technology is not limited to the use of devices, but extends to every stage of the chain, and the study gives us some fairly precise figures on the subject.
- Data transmission: 1 GB of data transmitted in France generates around 10 gCO2e.
- Storage: keeping 1 GB of data for a year emits an average of 400 gCO2e.
- E-mails: a standard e-mail generates around 0.41 gCO2 per MB transmitted.
- Cloud & team messaging: storing 1 GB of data in the cloud consumes 410 gCO2e/year.
- Videoconferencing: 0.36 gCO2e/min/participant.
These figures may seem low at first glance, but they can be staggering if you put them on the scale of the number of users and their uses.
The problem also lies in trends. You might think that if the figures are bad, there’s an awareness and that things are gradually improving, but that’s far from the case. The study shows us that not only have 80% of the users observed increased their emissions , but also that on an annual basis:
? 15% are increasing their emissions by between 0 and 30%.
? 20% increase their emissions by 30 to 100%.
? 40% double or more their emissions.
Generative AI worsens the situation
At the start of this post, I gave you some figures, specifying that these were pre-AI forecasts, and for good reason: the rise of generative AI is only accelerating this environmental drift.
For example, a query on ChatGPT-4 consumes 10 times more energy than a query on GPT-3.5, data centers should see their electricity consumption double by 2026 and Microsoft, despite its environmental commitments, saw its emissions increase by 30% in 2023.
The gap between rhetoric and reality.
When we see the gap between what businesses say and what they actually do, we have to wonder whether they aren’t greenwashing on a grand scale when they talk about digital issues, or whether, as when they talk about hyperconnectivity, they aren’t thinking “if it’s intangible and you can’t see it, it doesn’t exist”.
The study also highlights a number of aggravating factors.
One of these is the rebound effect. By this we mean that rather than reducing consumption (which would be rationally possible), technological advances simply introduce new uses that stimulate demand and therefore mechanically increase the digital footprint.
In my opinion, there’s also the experimentation effect, particularly with AI. We’re trying out everything and anything, using AI even for use cases that aren’t appropriate, but there’s no guarantee that tomorrow we’ll return to more rational and reasoned usage.
Another factor is the multiplication of data. I think that the older among you, like me, saw the arrival of the cloud as a liberation (I remind the younger ones that in the mid-2000s it was unthinkable to talk about cloud or SaaS to most CIOs…) because when we talked about collaborative digital uses, the question of storage space very quickly came up in conversation. Everything was stored on the business’s servers, which were not infinitely expandable, so whether you were talking about shared documents or even mailboxes, you had to clean them out regularly.
With the cloud, this storage capacity has not become infinite, but almost, and in any case it’s perceived as such, which means that users have absolutely no hygiene in this area.
Worse still, the same documents will be found in different storage spaces, in personal drives, in mailboxes (and in the mailbox of each recipient), with no one bothering to optimize information management.
What are the levers for digital sobriety?
Here too, the study offers some runways that may seem like common sense to you, but experience shows that common sense is perhaps the least common thing in business.
First of all, we need to change the way we use data. This means deleting useless data and archiving those that need to be kept but are not permanently accessible, or using collaborative tools rather than sending multiple e-mails. When we still see, in 2025, people sending a document as an attachment instead of sharing a link to the document, when this should have been the norm for at least 10 years, we can see how far we’ve come.
In the same way, entrusting ChatGPT with a simple search that Google would do so easily is a bit like having a pizza delivered by Airbus.
Ironically (or not), while the very essence of the new tools that have invaded our work environment is sharing rather than multiplying documents, they have produced the opposite effect.
Of course, since the environment is a shared concern , raising employee awareness is not only essential, it should be easy.
But when we’re talking about intangible things, we end up, in my opinion, with the same problem we encountered with personal use: the ecological terrorist is less the one who takes the plane than the bing-watching Netflix and fast-fashion enthusiast, but the idea doesn’t take root in people’s heads.
One point of awareness, in particular, concerns the need to insist on collective efficiency as the driving force behind change. In my opinion, this is part of the problem: everyone thinks that their own impact is negligible and that they can’t influence the practices of others. But it’s when you send a 10 MB PDF to 20 people that the problems start…
You also need to set up tracking indicators to monitor changes in practices. With the advent of solutions like Gr33t, this is almost the easiest part, and in any case, it’s where it all starts.
There are also questions of governance? Coercive policies or empowerment? The answer is probably somewhere in between, and I’ve noticed that businesses have been able to stand firm on similar issues when they needed to. But we were talking about things that had a materiality (travel, printing of documents) and had a financial cost. But as with modern solutions, there’s no such thing as pay-per-use billing….
We also need to raise awareness among managers, by making digital sobriety an area of performance that is evaluated and valued (“Tell me how you measure me, and I will tell you how I will behave”). This may seem basic, but for me one of the basic principles of change is to align evaluations and eliminate paradoxical injunctions.
On the other hand, measuring employees in relation to their digital activity is, for many reasons, totally counter-productive, and those who have thought of it have gone backwards (How to motivate your employees to blow hot air instead of being productive (thank you Microsoft)).
Finally, there’s the possibility of organizing internal events and challenges on the subject or even, very simply, joining existing initiatives like Digital Cleanup Week.
Bottom line
While I think the study has the merit of pointing the finger at the subject, I still find it extremely benevolent, given the scale of the issue and the reluctance of businesses to really take action on the subject of employees’ digital habits.
It’s time to get out of denial. Digital technology, far from being immaterial, has a colossal ecological footprint. Its impact on the environment can no longer be ignored or minimized. It’s not promises that will change the situation, but concrete, measurable action.
Digital work already accounts for half of the CO2 emissions linked to a workstation, and the multiplication of data and the exponential growth of data flows are only making the situation worse.
It’s not a question of technology, but of behavior.
Image: Green IT by DC Studio via Shutterstock.
