The climate case for AI has not been made

Countries are meeting in Antalya, Turkey this week for the United Nations climate conference, where artificial intelligence is expected to loom large. The host has already called on nations to commit to AI being designed, powered, and deployed in ways that support climate action. But the article's authors argue that industries worldwide are doing far less than the technology's champions suggest, despite highly optimistic claims from AI companies.

Those claims are substantial. In 2024, OpenAI chief Sam Altman said AI would deliver "astounding triumphs," among them "fixing the climate." A 2023 op-ed by Google's head of sustainability and a coauthor put numbers on it, arguing AI could trim global emissions 5% to 10% by 2030. A peer-reviewed study published last year, led by researchers at the London School of Economics' Grantham Research Institute and the firm Systemiq, estimated AI could cut annual global emissions by the equivalent of 3.2 to 5.4 billion tons of carbon dioxide by 2035 — at the high end, roughly comparable to US annual energy-related emissions, even after subtracting AI's own energy footprint. The gains were projected to concentrate in three areas: power, meat and dairy, and vehicles.

To test whether that 2035 figure reflects reality, the second author of the piece conducted and reviewed an AI-assisted analysis of public information from significant market actors in each sector. The result: few companies have announced they are pursuing the specific emissions-reducing routes underpinning the study's projections, and fewer still say they have achieved meaningful climate progress through them.

The power sector offers the most concrete math. AI-driven forecasting of supply and demand could push up to 20% more renewable electricity from already-built solar and wind facilities onto the grid, preventing an estimated 1.8 billion tons of emissions annually. Yet of the grid operators reviewed across China, the United States, and Europe, only one was found to be running a meaningful, working AI program of this kind. A partnership between the UK's National Grid ESO and Open Climate Fix halved the operator's solar-forecast error — the gap between predicted sunlight and what actually reaches panels — avoiding hundreds of thousands of tons of carbon dioxide a year. Elsewhere in the EU, the US, and China, the public record showed little of the AI integration the study envisioned.

Meat and dairy show the widest gap between technological readiness and market uptake. Hitting the projected reductions would require AI to close the taste and cost gap between traditional proteins and alternatives such as lab-grown or plant-based products. Among the clearest efforts, food-tech firm NotCo formed a joint venture with Kraft Heinz around its Giuseppe platform, which parses plant molecular data to find combinations that replicate the taste, texture, and smell of animal products. The review suggests few other alternative-protein companies have AI so deeply integrated into product development.

What remains open is whether the headline projections stay aspirational: whether other grid operators replicate the UK's forecasting win, whether more companies across the three sectors disclose AI deployments that match the study's pathways, and whether commitments emerging from Antalya translate into visible, measurable deployments.