For years, tech giants have aggressively promoted the idea that artificial intelligence (AI) will be a key solution to climate change. Google, for example, has claimed that AI could cut global greenhouse gas emissions by 5–10% by 2030 – a significant number equivalent to the annual emissions of the European Union. However, a closer look reveals that these claims are based on shaky evidence and often serve to justify an explosive, energy-intensive build-out of AI infrastructure.
The reality is that much of the hype surrounding AI’s climate benefits lacks scientific backing. Energy researcher Ketan Joshi investigated Google’s claims and found that the 5–10% reduction figure originated from a 2021 BCG analysis that relied on “experience with clients” – a vague and unsubstantiated source. This estimate conveniently emerged before the current AI boom, driven by energy-hungry generative models like ChatGPT.
The Energy Cost of AI Growth
Tech companies are racing to develop AI, but this comes at a steep environmental price. In the US, the expansion of data centers to power these AI systems is so large it’s keeping coal plants operational and adding hundreds of gigawatts of new gas power to the grid. Despite these costs, tech executives insist that the benefits of AI outweigh the energy demands. Jeff Bezos’s Earth Fund has hosted events promoting AI as an “environmental force for good,” while former Google CEO Eric Schmidt argues that focusing on AI is more effective than trying to meet existing climate goals. OpenAI CEO Sam Altman has even promised that AI will “fix” the climate.
However, a new report by Joshi, supported by environmental organizations, reveals that only a quarter of over 150 claims about AI’s climate benefits are backed by academic research. More than a third of these claims lack any publicly cited evidence at all.
The Generative AI Disconnect
The problem isn’t just the lack of proof; it’s what kind of AI companies are touting. Many older, less energy-intensive machine learning applications have long been used in scientific fields to reduce emissions. But it’s generative AI – ChatGPT, Gemini, and similar models – that is driving the current data center build-out. Companies often conflate these two, falsely suggesting that all AI is equally beneficial.
Transparency and Accountability
Experts argue that tech companies need to be transparent about the energy costs of their AI development. Joshi advocates for full disclosure of energy consumption, stating that if companies fear exaggeration, they should reveal exact figures: “If [tech companies] are worried that people are overstating or exaggerating the climate impacts of generative AI, then there should be nothing stopping them from saying, ‘Well, OK, our energy growth this year was six terawatt-hours, and two of them were for generative AI.’”
The narrative that we need massive AI models – and the quasi-infinite energy to power them – serves to convince us that this is the only possible future. Smaller, more efficient models can often achieve similar results at a fraction of the environmental cost, but they are ignored in favor of the bigger-is-better approach favored by tech giants.
Conclusion
The claims that AI will save the planet are largely unsubstantiated hype. The rapid expansion of generative AI infrastructure is driving up energy demand without clear evidence of equivalent climate benefits. Until tech companies prioritize transparency and accountability, their promises remain empty gestures in the face of a worsening climate crisis.
