AT last week’s UN General Assembly (UNGA), the secretary-general focused on three existential threats to humanity: climate change, unchecked AI, and spiralling income inequality. It is clear that these threats are complex, intersecting and perceived with varying levels of concern by geopolitical heavyweights. But these threats are not intractable. The dizzying speed at which narratives shift in this social media era means we have moved quickly from catastrophising about the climate apocalypse to predicting myriad ways in which AI will ravage society and the planet. But such widespread despair never produced a good outcome.
It was helpful, then, to see more nuanced positions emerging from New York Climate Week, which unfolded at the same time as the UNGA. Here too, climate and AI were at the top of the agenda, but with less zero-sum messaging.
Undoubtedly, AI’s environmental impact is significant, and arguably unconscionable, given the scale of the climate crisis, and the failure to reduce greenhouse gas emissions. Data centres that power AI use land, already scarce water supplies, and immense amounts of energy, still largely derived from fossil fuels. By 2030, 40 per cent of electricity used by data centres will be for AI workloads — that equates to the residential electricity needs of all of sub-Saharan Africa (1.3 billion people), according to a 2025 study by Gartner, a technology insight company. Bloomberg New Energy Finance estimates that the AI load on data centres will increase global power sector emissions by 10pc over the coming decade. This is on top of emissions associated with other computing linked to cloud storage, video-streaming and the growing use of cryptocurrencies.
The AI debate is heating up just as scientists are confirming that this year marks the tipping point beyond which restricting global warming to 1.5 degrees Celsius higher than pre-industrial levels is unachievable. The more evident the scale of the climate crisis becomes, the more perverse it seems to double down on a technological advance that has an insatiable appetite for electricity (fossil fuels) and water.
But AI is here to stay, and plans to colonise Mars are still nascent. And so at New York Climate Week, more attention was paid to mounting claims that AI can help tackle the climate crisis. Examples of this include the use of AI for more intelligent load management for renewable energy; using AI to track pollution more effectively by pulling together disparate data sources from weather stations, satellites and private monitors; and better modelling of climate change impacts for adaptation and mitigation. There was also recognition that in a US that at the federal level is rolling back on climate commitments, American Big Tech companies are playing an important role funding renewable energy projects through multimillion-dollar contracts.
The climate aspect should be a core focus of AI regulation.
But this more balanced AI-fights-climate change story should not enable greenwashing by Big Tech. While there are hopes that data centres will increasingly be powered by renewable sources, the overall effect of bringing coal- or gas-fuelled power plants online now to sate immediate energy needs is hugely damaging. It stymies efforts that have been underway to retire fossil fuel-based power generation.
In a recent paper, Ketan Joshi, a clean energy and corporate accountability researcher, highlighted that many claims of AI’s climate impact link back to case studies being published by Big Tech and other corporates that are deeply conflicted, rather than independent academic research. Joshi’s paper also points out that the AI that can be used to benefit the climate is ‘traditional’ machine-learning which has a much lower environmental impact than consumer-generative AI tools, such as Copilot, Gemini or ChatGPT, that are more resource-
intensive.
With this awareness comes the pressure on AI consumers to make more ethical choices about when and how they use AI. Since generative AI has a greater environmental cost, users should be more informed about when they genuinely need the extra computational power (stick to Google for the recipes; images and video are more wasteful than text; don’t automatically opt in for the AI search option).
But it’s unfair to put the onus on end users. The climate aspect should be a core focus of current AI regulation and sovereignty debates. As with political discourse in recent years, it’s important that these issues not be hijacked by zero-sum thinking, knee-jerk polarisation and absolutism. The dire warnings punctuated by glimpses of hope in New York last week are a reminder that the only way forward — on the climate, AI and much else — is on the basis of empirical evidence, nuance, transparency and accountability.
The writer is a political and integrity risk analyst.
X: @humayusuf
Published in Dawn, September 28th, 2026
