It is a delicate exercise which Ademe and Arcep have tackled to assess the environmental impact of digital technology. We remember the recent criticisms on the implementation of the CO² indicator of fixed and mobile data on invoices and on the chosen methodology. For their study, the two organizations relied on the methodology of life cycle analysis (multi-criteria (11 environmental indicators), multi-stages (manufacture, distribution, use, recycling) and multi-components).

Before drawing lessons with this method, Arcep and Ademe start from a more general observation of the environmental footprint of digital services in France. On their electricity consumption, for example (48.7 TWh), they represented 10% of overall consumption. With regard to CO2 emissions, they show a score of 16.9 million tonnes of carbon equivalent and would contribute up to 2.5% of the total carbon footprint in France. Finally, these services generated 20 million tons of waste per year over the entire life cycle of the equipment.

Terminals, first responsible for environmental impacts

On this last point, the first responsible for the majority of impacts are “user” and IoT terminals (between 64 and 92% depending on the device). Among them, the pet peeve is the television screen, especially due to the large number of materials and equipment needed to manufacture them. The list also includes laptops, tablets, smartphones, with the same issue on the manufacture and end of life of terminals.

Of the three segments studied, the impact of terminals comes out well ahead. (Photo credit: Ademe/Arcep)

On the IoT, extrapolation work was carried out to find out the number of connected objects in circulation in France. This figure reaches nearly 245 million in 2020. The report finds that power consumption is relatively low, but the volume of devices has a significant environmental impact. Associated with the terminals, the question of the network ranks third in the study. It finds that the fixed network (xDSL and FTTx) is the one that has the most impact compared to the mobile network (2G, 3G, 4G, 5G). Indeed, the former consume more electricity in the use phase, and require more equipment, in particular because of the boxes installed at the users.

A strong disparity for data centers

In second place: data centers. But here again the impacts vary according to the type of data center identified. Thus colocation centers have a strong impact (35 to 50%) compared to those of companies (30 to 40%). The state and local part contributes less (5 to 15% of the impacts) and the HPC supercomputers (between 0.1 and 5%). Figures to be put into perspective, because as the report states, “the difficulty of accessing data requires the use of approximations which represent a limit in the precision of the results of the evaluation of the environmental impacts of the data center park”. It refers to information and documentation on internal data centers (public or private) that are very rarely published.

Colocation and corporate data centers account for 82% of the surface area of ​​computer rooms in France. (Photo credit: Ademe/Arcep)

Based on various studies and analyses, the study indicates that the surface area of ​​computer rooms in France amounts to 883,165 m2. The whole has a PUE of 1.69 (with variations of 1.93 for the public and businesses, 1.55 for colocation and 1.17 for HPC) and an electricity consumption of 11.59 TWh, i.e. 2.4% of national electricity consumption. It should be noted that the data is for 2020 and that since then several datacenters have emerged, particularly for colocation pure players with optimized cooling systems and redistribution of heat for the benefit of the community (urban heating for example).

Avenues for reflection to minimize the impact

Faced with the various results of the study, Ademe and Arcep note that it is the manufacturing phase which is the main source of impact (78% of the carbon footprint), followed by the use phase. (21% of the carbon footprint). On this last point, it is possible to minimize this impact by focusing on policies aimed at extending the useful life of digital equipment through product durability, reuse, reconditioning, economy of functionality or repair.

In addition, the two organizations are campaigning for an improvement in the methodology. They therefore ask for more data (often blocked in the name of business secrecy) to refine their analysis. They plan to publish next April a prospective analysis of the impacts of digital technology in 2030 and 2050 based on 4 scenarios. Appointment is made…