How Sustainable is AI?
Do generative AI's benefits for the planet outweigh its impacts? Let's try to find out...
At Scaleway, sustainability is a core concern. This why we decided to create a unique environmental footprint calculator, designed to provide our users with a
users a clear and comprehensive view of the environmental impact of our products.
of our products. Unlike other cloud calculators, which only measure carbon for now, this tool measures not only carbon emissions (scopes 1, 2 and 3) - including hardware-specific energy consumption - but also water consumption.
Based on as a robust a methodology as the PCR of French national ecological agency ADEME, the calculator provides transparent, reliable estimates, both at the time of purchase and in detailed and in detailed monthly reports. The aim is simple: to enable Scaleway clients to make informed decisions and optimize their infrastructures, in order to reduce their environmental impact.
A detailed and comprehensive methodology is essential to guarantee the reliability of environmental impact data and to make it usable in infrastructure choices. By providing an accurate, multi-criteria view, we enable companies to understand the full ecological footprint of their operations.
This data, transparent and aligned with recognized standards, serves as a solid basis for comparing different options, identifying levers for improvement, and making informed decisions. In this way, a rigorous methodology does more than simply measure impact: it becomes a genuine strategic tool for effectively reducing one's footprint while supporting a more responsible digital transition.
Yet this journey has not been without its pitfalls. Accurately measuring environmental footprint, particularly scope 3 (which encompasses the indirect emissions of suppliers and purchased products), is a complex challenge.
Scope 3 represents the CO2 emissions of the suppliers and equipment we use. This category is often the most difficult to assess, as it depends on the quality and availability of data supplied by equipment manufacturers and other partners. Nonetheless, it is indispensable.
As a cloud provider, scope 3 represents over 80% of the carbon footprint of our services. To ignore this category would be to miss out on a substantial part of our overall footprint, which makes our commitment to its measurement all the more crucial for comprehensive and relevant environmental impact monitoring.
At Scaleway, we have a wide variety of hardware equipment from multiple suppliers, some of which has a lifespan in excess of 10 years. We also practice intensive recycling, reusing components to integrate them into other servers, which extends the lifespan of our hardware.
However, this approach adds a further layer of complexity to the precise calculation of the environmental impact of our servers, due to the traceability and (lack of) heterogeneity of the equipment used. Collecting accurate impact data for these machines, particularly the older ones, proved a major challenge. Indeed, some of our equipment suppliers did not have sufficiently detailed impact data, or such data was not always available for older equipment.
After encountering a number of obstacles on this front, we decided to capitalize on the excellent work carried out by French association Boavizta, which has developed impact models based on emission factors and the average life cycle analysis (LCA) of hundreds of types of hardware, including servers. This enabled us to guarantee more robust estimates, although some uncertainty remains.
It is important to point out that our methodology is still evolving, and we are ready to integrate data from our suppliers as soon as they become available, in order to refine our calculations and guarantee even greater accuracy.
Another major challenge concerned our infrastructure servers used for in-house tools. The lack of comprehensive, automatable data on these servers, such as their specifications (CPU, RAM, etc.), made it difficult to assess their environmental impact linked to manufacturing. This time, the difficulty was linked to our internal organization. To overcome this problem, we implemented a progressive data collection method, automating as far as possible the retrieval of hardware specifications. At the same time, we applied generic impact coefficients based on standard specifications for servers for which data was not available.
The environmental impact of data center construction is a particularly sensitive issue. In the absence of specific data on the construction of each data center, we had to rely on estimates based on generic emission factors provided by ADEME. However, the level of uncertainty surrounding these estimates remains relatively high.
We have chosen to precisely document this uncertainty in our methodology, while incorporating the best available estimates. The use of emission factors per m² provided by ADEME enables us to calculate a carbon footprint in relation to the size of our infrastructures, although we continue to work on improving the accuracy of the data as new information becomes available.
Non-IT impacts (offices, travel, etc.) represent a more indirect part of our environmental footprint, but are nonetheless crucial to measure. We had to define how and how often to update these data: should we base ourselves on annual averages, monthly data or choose another approach?
We opted for an annual update of non-IT data, based on personnel stats in our Impact Report, in order to offer our users a consistent and comparable overview. This decision enabled us to erase seasonality concerns, and to maintain a good balance between accuracy and practicality in the ongoing monitoring of our global footprint.
Measuring environmental impact accurately and comprehensively is a complex challenge, particularly when it comes to scope 3 emissions. However, thanks to our commitment to transparency and the continuous improvement of our methods, we are able to provide our users with reliable information to help them reduce their carbon footprint. Our collaboration with experts and our participation in working groups such as ADEME's PCR group reinforce our ability to innovate in this crucial area.
We are pursuing our optimization mission, and will continue to listen to feedback from our users as we refine our tools and calculation methods.
Find out more about the environmental footprint calculator here; and check out our latest presentation on this topic, from Green IO Paris, here!
Do generative AI's benefits for the planet outweigh its impacts? Let's try to find out...
What is next for Green IT? Regulation, and 2050 carbon neutrality targets, will drive a shift from nice-to-have to must-have. And then there's AI...
Green IT can reduce energy consumption by up to 30%... but that's not all! It can also cut server requirements in half, and save huge amounts of CO2 and water. Read on for inspiration!
You can't improve what you can't measure. The good news: you can, when it comes to cloud, hardware, software and website energy consumption. Let's go!
The digital sector generates 4% of global greenhouse emissions. Data centers and hardware are the main contributors; but they all run on code, written by developers. How can that impact be reduced?