Four of the world’s biggest AI and cloud providers watched their emissions climb by as much as 239 percent between 2020 and 2024. Over those same four years, fourteen major telecoms operators cut theirs by 11 percent. That gap, sitting in a report the UN’s telecoms agency published this week, is the clearest picture yet of what the AI build-out is doing to the technology industry’s climate arithmetic.
A different kind of energy problem
The Greening Digital Companies 2026 report was launched on 2 September at the World Telecommunication Policy Forum in Geneva, produced jointly by the International Telecommunication Union and the World Benchmarking Alliance. Now in its fifth edition, it tracks greenhouse gas emissions across Scopes 1, 2 and 3, energy consumption, climate targets and transition plans at 200 digital companies. The figures come from 2024, the most recent full year of corporate disclosure.
Together those 200 companies reported 301 million tonnes of operational emissions in carbon dioxide equivalent. That amounts to roughly 0.8 percent of global energy-related emissions, which sounds modest until you look at the electricity column. The same group consumed close to 500 terawatt-hours, about 1.7 percent of all the electricity used on the planet, and the report expects that demand to keep climbing as AI, cloud computing and digital infrastructure expand.
Concentration is the striking part. The ten largest consumers alone accounted for 269 TWh, more than the entire country of Australia uses in a year. China Mobile topped the list at 63 TWh, nearly double Alphabet and Samsung, which each reported 32 TWh, with Microsoft close behind at 30 TWh. Amazon did not report electricity data for 2024 at all, which leaves a conspicuous hole in any attempt to size the sector properly.
“While digital technologies offer immense potential for climate action, their rising energy demands and emissions cannot be overlooked,” said Doreen Bogdan-Martin, who heads the ITU. “Environmental sustainability must be built into how we design, power and scale the technologies shaping our shared digital future.”
Targets are easy, progress is not
Commitments have spread fast. Of the 200 companies assessed, 151 now hold near-term emission reduction targets, which the report credits to a mixture of voluntary leadership, investor expectations and regulation that has started to bite. Only 85 were judged to be genuinely on track based on progress to date. Just 81 had comprehensive plans setting out how they intend to hit their goals.
Renewable sourcing follows the same pattern. Technology firms sit among the largest corporate buyers of clean electricity anywhere in the world, and yet only 25 of the 200 companies in the report said they run entirely on renewable power. Buying certificates at scale is not the same thing as running a business on carbon-free electricity, and the gap between the two is where most of the sector currently lives.
Gerbrand Haverkamp, executive director of the World Benchmarking Alliance, pointed at the part of the problem that rarely makes headlines. “Digital companies need to engage suppliers and address emissions across the products and services they rely on,” he said. Scope 3 emissions, the ones embedded in hardware, construction and supply chains, are harder to measure and far harder to cut than the power bill for a data centre.
Who scored well, who did not
Each company received a climate assessment score built from its targets, its data quality and its actual performance. Swisscom was the only firm to earn a perfect score, followed by Accenture, Deutsche Telekom, Vodafone, Capgemini and Telefonica. European telecoms operators dominating a sustainability ranking is not an accident; they have had regulators and disclosure rules leaning on them for years.
At the other end, eighteen companies received no score whatsoever, including Elon Musk’s X and SpaceX. The bottom thirty among firms that were scored included Huawei, Spotify, Nintendo, Weibo, Xiaomi, Zoom and Toshiba TEC. Some of those are small energy consumers with weak reporting rather than heavy emitters, which is exactly the ambiguity the ranking is designed to expose.
The efficiency defence
Artificial intelligence “is both a driver of efficiency and a growing challenge”, the ITU said in its statement accompanying the report. The industry has a genuine case to make here. AI is already being used for grid-level energy optimisation, renewable output forecasting and efficiency work inside data centres themselves. Those gains are real and they compound.
They also arrive attached to a cost the report is blunt about. Every efficiency win has to be measured against the infrastructure being poured into the ground to deliver it, and on 2024 numbers the infrastructure is winning. Next year’s edition, working from 2025 data, will be the first to capture a full year of the current generation of frontier model training runs. If the curve has not started to bend by then, the pledges start to look like accounting rather than strategy.
For more coverage of AI and climate, visit Mylistingo.
The Netherlands, for internationals
One email a week: the week's news for internationals in the Netherlands, new practical guides and what changed in the cities. No spam, unsubscribe any time.
