AI vs. Climate: Google and Amazon Break Zero-Emission Promises

The Hidden Cost of Artificial Intelligence

Generative AI has revolutionized technology, but its environmental impact is alarming. According to a TechCrunch report, Google and Amazon are struggling to meet their net-zero emission promises due to the explosive growth of AI. Training a single large language model, like GPT-3, can emit around 500 tons of CO₂, comparable to the lifetime emissions of five cars. However, inference (model usage) multiplies that cost: a ChatGPT query is estimated to consume about 10 times more energy than a Google search. This surge in energy demand has led AI data centers to already account for about 2-3% of global electricity consumption, a figure that could double by 2026 according to the International Energy Agency.

What Happened?

In 2024, Google reported that its greenhouse gas emissions had increased by 48% compared to 2019, reaching 14.3 million tons of CO₂ equivalent, instead of decreasing toward its net-zero goal for 2030. The company attributed this increase to the energy consumption of its data centers, where AI workloads have skyrocketed. Amazon, for its part, reported a 39% increase in emissions in 2023, reaching 71.3 million tons, driven by the expansion of its data centers for AI services like AWS Bedrock. Both companies had made ambitious climate promises: Google committed to operating on 24/7 carbon-free energy by 2030, while Amazon pledged net-zero emissions by 2040. However, the demand for AI computing has outpaced their investments in renewable energy. For example, Google reported that in 2023, 64% of its energy came from carbon-free sources, but AI growth has caused this percentage to decline. Amazon, although it has installed wind and solar farms, has had to resort to carbon credits to offset some of its emissions.

Why Does It Matter?

This failure is not just corporate: it sets a precedent for the entire tech industry. If leaders like Google and Amazon cannot meet their climate goals, what hope is there for startups and smaller companies that rely on the cloud? Moreover, AI is being integrated into all sectors, from healthcare to finance, multiplying its carbon footprint. A study from the University of Massachusetts Amherst estimated that training a single large language model can emit as much CO₂ as 300 transatlantic flights. Consumers and regulators are beginning to push for transparency and action. The European Union, for instance, has proposed the Ecodesign Directive for sustainable products, which would require tech companies to report the environmental impact of their AI services. In the United States, the FTC has started investigating carbon neutrality claims. Additionally, AI's energy cost could translate into higher prices for users, as companies pass on electricity and cooling costs to customers.

Consequences and the Way Forward

The consequences are multiple: increased regulatory scrutiny, potential loss of investor and customer trust, and a slowdown in AI adoption if the energy problem is not resolved. Google and Amazon are investing in nuclear and geothermal energy, but these solutions will take years to materialize. For example, Google signed an agreement with Kairos Power to build small modular nuclear reactors, but they are not expected to be operational until 2030. Amazon, meanwhile, bought a data center powered by nuclear energy in Pennsylvania. In the meantime, AI model efficiency and strategic placement of data centers near renewable sources are insufficient palliatives. Some companies are exploring specialized hardware, such as more efficient AI chips (e.g., Google's TPUs or Amazon's accelerators), which can reduce energy consumption by up to 50%. However, the exponential growth in AI demand could offset these gains. Additionally, locating data centers in regions with cold climates or abundant renewable energy, such as Nordic countries, can reduce the carbon footprint but does not eliminate the problem.

What Readers Should Know

It's not about abandoning AI, but demanding accountability. Companies must report emissions with greater granularity, invest in more efficient hardware, and prioritize green computing. As users, we can opt for AI services that are transparent about their environmental impact. For example, some startups offer AI models trained with renewable energy and offset their emissions. The climate battle is also fought in the servers. If urgent measures are not taken, AI could become a drag on global climate goals, rather than a tool to achieve them.