Delivering ICT’s Climate Impact: Data, Methods and Country Experiences

Ana Gabriela Fernandez presented findings showing a dramatic increase in the proportion of Nationally Determined Contributions (NDCs) that reference digital technologies. Earlier reports from 2023 found around 45–50% of NDCs mentioning digital technologies, whereas the latest cycle of NDCs analysed by ITU showed approximately 90% referencing them. This represents nearly a doubling of the share of countries planning to utilise digital technologies in their climate commitments.

A 2023 UNFCCC and TEC report found that around 45% of NDCs included a reference to digital technologies , while a World Bank study in the same year concluded it was approximately 50% . ITU's policy brief for COP30, analysing 53 NDCs from the new cycle, found that around 90% were already referencing digital technologies , with 47 countries planning to utilise them , nearly double the previous baseline .

The majority of digital technology references found in NDCs are oriented towards relatively basic applications such as monitoring systems, data collection tools, and early warning systems. These are practical, operational solutions rather than advanced or cutting-edge technologies. This suggests that while digital technology adoption in NDCs is widespread, the ambition level in terms of technological sophistication remains moderate for most countries.

Ana Gabriela Fernandez noted that most references in NDCs were looking at how to utilise digital solutions around monitoring, data systems, and efficiency solutions, with early warning systems being a very popular data solution referenced in NDCs . More advanced solutions were categorised under 'others' .

While AI is beginning to appear in NDCs, it remains a minority reference, cited in roughly 10–12% of the NDCs analysed. This indicates that advanced technology planning is still at an early stage within national climate commitments. The figure is consistent across both the earlier policy brief and the ongoing full report analysis.

Artificial intelligence was referenced in around 10% of NDCs in the policy brief analysis , and in the ongoing full report analysis, AI is being referenced in roughly 10-12% of the NDCs analysed so far .

Despite the growing role of ICT in climate action, none of the NDCs reviewed were attempting to capture the greenhouse gas emissions of the ICT sector itself within their mitigation or adaptation targets. This represents a significant policy gap, particularly given the rising emissions associated with technologies such as AI. The finding raises the question of how countries should begin to account for ICT sector emissions in their national climate plans.

Ana Gabriela Fernandez reported that none of the entities analysed were trying to capture GHG emissions from the ICT sector in terms of mitigation or adaptation targets , and noted that this raised the question of how to start thinking about this gap, especially as emissions from AI are expected to rise .

Reyna Ubeda emphasised that while digital technologies are increasingly referenced in NDCs, having concrete methodologies and standards is critical to generating reliable data that can substantiate these commitments. Standards provide the requirements and methodologies needed to move from general claims about digital technology's climate benefits to verifiable, comparable data. Without such standards, the assertion that digital solutions reduce GHG emissions remains unsubstantiated.

Reyna Ubeda highlighted that a key recommendation from the ITU policy brief was the implementation of standards and having a concrete methodology to have data, noting that 'standards provide requirements, standards provide methodology' . She introduced the standard L.1480 as the tool to measure the enablement effect of ICT in other sectors .

Reyna Ubeda stressed that the standard L.1480 is not limited to the ICT sector but can be applied by any sector that deploys a digital solution and wishes to assess its environmental impact. This broad applicability makes it a versatile tool for cross-sectoral climate action. She gave examples ranging from steel and cement production to education and e-learning.

Reyna Ubeda explicitly stated that the standard can be applied for any sector, including steel, cement, manufacturing, schools, and education, giving the example of an e-learning tool being assessed for its environmental impact .

Jean Manuel Canet explained that L.1480 was developed collectively within ITU Study Group 5 to provide a standardised, robust way to assess the environmental impact of ICT and digital solutions, including AI. The standard addresses the dual nature of ICT — its own footprint as well as the benefits it can enable in other sectors. Having a common methodology is essential for credible accounting of both the positive and negative environmental consequences of digital solutions.

Jean Manuel Canet noted the importance of having a common methodology to assess how digital services can bring benefits in terms of energy savings and GHG emissions savings , and explained that Study Group 5 collectively decided to develop a standard to give guidance on how to do this in a robust manner . He described L.1480 as covering multiple perspectives including ICT service providers, end users, and organisations contributing to ICT solutions such as telecom operators and data centre operators .

L.1480 accommodates different levels of analytical rigour through a three-tier framework, recognising that practitioners have varying amounts of time and data available. Tier 3 is the simplest and is used before implementation when data is not yet available, providing an order-of-magnitude estimate. Tier 1 is the most robust, relying on actual data collected after implementation, while Tier 2 sits in between.

Jean Manuel Canet explained that Tier 3 is the simplest possible assessment, used before implementation to get an order of magnitude of the expected impact , Tier 2 is an intermediate assessment , and Tier 1 is a full study relying on existing data collected after implementation, such as interviewing teleworkers and gathering commuting statistics .

L.1480 requires a comprehensive net calculation that considers three types of effects. The first-order effect captures the direct environmental footprint of the ICT system itself. The second-order effect captures the environmental benefits enabled by the ICT solution, such as reduced travel. The rebound effect accounts for unintended behavioural changes that may partially offset the benefits, such as people using saved time and money for additional activities.

Jean Manuel Canet described the first-order effect as the direct impact from all components making an ICT solution feasible, using teleworking equipment as an example . The second-order effect was illustrated by reduced car travel from teleworking leading to GHG savings . The rebound effect was explained as the time and financial savings from not commuting potentially leading to new activities and expenditures .

The consequence tree is a central analytical tool within L.1480 that systematically maps all the consequences — positive and negative — arising from the use of an ICT service. By drawing this tree, practitioners can ensure that no significant impact is overlooked and can arrive at a comprehensive net assessment. The standard provides guidance on how to construct this tree for different types of ICT solutions.

Jean Manuel Canet described the consequence tree as being at the heart of the document, providing guidance on the different types of impact and consequences . He gave the example of a virtual meeting consequence tree being included in the standard , and showed how the consequence tree was drawn for the Carbon Lens project in Senegal .

The Carbon Lens project in Senegal is a case study in which L.1480 was applied to assess the environmental impact of an AI-enabled solution for carbon stock monitoring. The project uses satellite imagery, AI models, and local engagement to automate what was previously a costly and time-consuming manual data collection process. It aims to increase accuracy in carbon stock monitoring and facilitate scalability across geographies.

Jean Manuel Canet described the Carbon Lens project as a tool combining satellite imagery, AI, and local engagement to automate costly and timely manual data collection, increase accuracy and stock monitoring, and facilitate scalability . The project uses Sentinel-1 and -2 satellites, PICA radar data combined with LIDAR, and an AI model trained on field and IOMAS data .

In the Carbon Lens case study, L.1480 was applied at the Tier 3 level, which is appropriate for assessments conducted before or during early implementation when full data is not yet available. The assessment compared a baseline scenario without AI or satellite imagery against the ICT-enabled scenario. The consequence tree was drawn and the results showed that the ICT scenario produced reduced emissions compared to the baseline.

Jean Manuel Canet explained that the impact assessment was conducted using L.1480 at Tier 3 assessment perspective , with GIZ designing the baseline scenario without data and AI and then defining the ICT scenario with all relevant equipment . The consequence tree was drawn and results showed reduced emissions from the baseline scenario with the ICT action scenario .

The overall finding of the Carbon Lens L.1480 assessment was that the AI-enabled ICT solution is less carbon-intensive than the manual baseline system it replaces. This result can be used to support policy decisions and advocacy for the use of AI in environmental monitoring. The study can also be refined in the future as more actual operational data becomes available.

Jean Manuel Canet stated that as a full result, the ICT solution in that case is less carbon-intensive to operate than the system it replaces , and that the study results can be used to support policy action and refined in the future with more actual data .

Jean Manuel Canet highlighted that ITU has developed a new standard, L.1801, specifically designed to assess the direct carbon footprint of AI systems. This standard is already being applied in practice, with two consulting companies conducting an assessment of the Mistral AI system using it. Major technology companies such as Google, which participated in the standard's development, are also expected to adopt it.

Jean Manuel Canet confirmed that ITU has means to assess the footprint of AI with the new standard L.1801 , and noted that Carbon4 and Resilio are currently using L.1801 to assess the Mistral AI system . He also mentioned that Google participated in the preparation and development of the standard and may use it in the coming weeks .

Jean Manuel Canet advocated for a dual approach to managing AI's environmental impact: keeping the direct footprint of AI as low as possible while simultaneously measuring and maximising the enabling benefits AI can deliver across sectors. Standards such as L.1801 are the tools to achieve this balanced approach. This allows for evidence-based decision-making about where and how AI should be deployed for maximum environmental benefit.

Jean Manuel Canet called on all stakeholders to push so that the actual direct footprint of AI is kept as low as possible , and recommended using L.1801 to assess and maximise the benefits of using AI in different scopes and environments .

Philippe Tuzzolino presented Orange's strong net zero commitment and the concrete progress made towards it. Orange uses the ITU methodology L.4070, validated by the Science Based Targets initiative (SBTi), to collect and report emissions data across Scopes 1, 2, and 3. By 2025, Orange had significantly exceeded its planned reduction targets, demonstrating that ambitious climate commitments can be achieved in practice.

Philippe Tuzzolino stated that Orange has a strong commitment to be net zero in 2040 and applies the ITU methodology L.4070 with SBTi validation . The original target was to reduce Scope 1 and 2 emissions by 30% by 2025, but Orange achieved a 49% reduction in Scope 1 and 2 and a 61% reduction in Scope 3, surpassing the objective .

Philippe Tuzzolino argued that for a company to credibly claim that its ICT solutions help other sectors reduce their emissions, it must first demonstrate that its own operations are on a net zero trajectory. Without this, claims about avoided emissions in other sectors are undermined by the company's own pollution. This principle of leading by example is presented as a foundational requirement for credible sustainability claims.

Philippe Tuzzolino stated that if a company wants to avoid emissions to customers or help other sectors reduce their own emissions, it must first be exemplary and commit to net zero itself . He argued that if a company is polluting, it cannot credibly say it can avoid emissions to other sectors .

Philippe Tuzzolino presented a detailed case study in which Orange applied L.1480 to assess the avoided emissions from teleworking. The study compared a 2020 baseline (35% teleworkers, 1.5 days per week) with a 2022 scenario (80% teleworkers, 2 days per week). The assessment accounted for both the positive effects of reduced commuting and the negative effects of increased home energy use and new rebound travel behaviour.

Philippe Tuzzolino described the teleworking case study with a 2020 baseline of 35% teleworkers at 1.5 days per week , increasing to 80% teleworkers at 2 days per week in 2022 . The assessment using L.1480 took into account life cycle analysis, new mobility implemented by teleworking, use of ICT solutions at home, and commuting from home to work . Negative aspects included video conferencing ICT use, maintained commuting, new rebound travel such as trips to supermarkets and libraries, and increased home heating and cooling .

The teleworking case study ultimately showed a net positive result in terms of avoided emissions, but Philippe Tuzzolino emphasised that this outcome is highly dependent on the specific conditions of implementation. For example, avoiding long-distance travel or international meetings produces a stronger positive effect than avoiding short local commutes. This sensitivity to conditions underscores the importance of rigorous, context-specific assessment.

Philippe Tuzzolino stated that the result of avoided emissions is nevertheless positive globally for this example , but noted that the positive effect is stronger when teleworking avoids long-distance travel or meetings with different countries compared to regular commuting . He emphasised that conditions are very important and that the result could change if conditions change .

Philippe Tuzzolino explained that Orange achieved its significant emission reductions through a range of concrete operational changes, including switching to renewable energy, electrifying its vehicle fleet, virtualising servers, and implementing circular economy practices. Crucially, these changes did not disrupt Orange's core business operations, demonstrating that sustainability and business continuity are compatible. The transition involved changing the business model rather than reducing business activity.

Philippe Tuzzolino described action plans including electric vehicles in all fleets, reduced energy consumption of data centres with free cooling, virtualisation of servers, and implementing a large share of renewable energy across all operating countries . He also mentioned a strong circular economy programme, with more than 90% of boxes reused for customers , and switching from copper to optic fibre networks to reduce energy consumption .

Philippe Tuzzolino argued that the transition to renewable energy and circular economy practices does not fundamentally harm business performance. When a company uses renewable energy, its CO2 emissions decrease significantly while business activities continue as normal. The key is changing the business model — for example, through circularity in product procurement and network upgrades — rather than reducing the scale of business.

Philippe Tuzzolino stated that using renewable energy decreases CO2 emissions drastically because the energy is not carbon-based, but the company continues to have action in the business . He noted that changing the network from copper to optic fibre reduces energy consumption drastically, and using electric vehicles for technicians reduces CO2 emissions from customer interventions, all while continuing to make business .

Philippe Tuzzolino explained that Orange extends its sustainability efforts to its customers through circular economy programmes, particularly by offering refurbished and reused devices at attractive prices. This approach makes sustainability accessible to different customer segments, including students and price-sensitive consumers who might not otherwise afford new devices. The reuse programme covers millions of devices in service.

Philippe Tuzzolino described Orange's 'Reuse' programme offering special deals with attractive prices for students and different levels of population who prefer to buy reused mobile phones because new iPhones are very expensive . He also noted that more than 90% of boxes are reused for customers, with many millions of boxes in service .

An audience member asked whether the cost efficiencies achieved through sustainability measures — such as switching to renewable energy and implementing circular economy practices — are passed on to end customers in the form of lower prices. This question highlights a broader concern about whether telecom operators' sustainability performance translates into tangible consumer benefits. The questioner also noted that high KPIs are sometimes cited by operators as a reason for high prices.

An audience member asked whether Orange reduces prices to customers as a result of its sustainability-driven cost reductions, and noted that telecom operators sometimes front high KPIs as impacting their businesses .

An online audience member raised the concern that the rapid expansion of data centre infrastructure required for AI training and inference could negate the process efficiency gains that AI enables. This risk has already been observed in practice, with companies such as Google reporting that their data centre emissions are rising due to AI workloads. The question challenges the assumption that AI will necessarily deliver net environmental benefits.

The question from an online participant asked whether the data centre build-out for AI training and inference will effectively cancel out the savings gained from process efficiency, noting that this has already been reported by Google and observed for Tesla products as well .