A Cautionary Story About the Real Costs of AI: Insights from Google and Amazon

It’s no secret that AI is a significant consumer of energy and water, outpacing earlier digital technologies in its demand. Recent studies highlight the environmental impact of Big Tech’s AI initiatives.

This week, both Google and Amazon released their sustainability reports, and the statistics are concerning. Despite their commitments to achieving net-zero carbon emissions in the years ahead, the rise of AI presents considerable hurdles in reaching these goals. Google’s overall carbon emissions have shot up by 25% compared to last year, while Amazon’s have increased by 16%.

A thorough review of these reports suggests that both Amazon and Google will need to make significant and potentially costly adjustments to their operations if they hope to attain their net-zero targets.

Although neither company explicitly links the rising emissions to AI, there is substantial indirect evidence indicating a connection.

AI at the Heart of the Matter

Both Amazon and Google acknowledge that their energy consumption has surged over the past year in tandem with the growing utilization of AI. They delve into carbon intensity, a measure of pollution produced per dollar of revenue, a concept previously employed by China during climate discussions, even amidst rising emissions. Moreover, both companies allocate several pages in their reports to discuss the environmental advantages of AI, which could be interpreted as a case of “protesting too much,” to quote Shakespeare.

The implications become clearer upon closer inspection of the data. Both companies are performing relatively well regarding carbon emissions linked to energy procurement. Years of investment in renewable energy have eased the situation; however, this may soon shift, as tech companies like Google increasingly invest in natural gas power plants to satisfy the high energy demands of AI.

Most of the rising carbon emissions for Amazon and Google are attributed to Scope 3 emissions—an overarching term for pollution not directly controlled by the company, such as procured goods and services or sold products. For these tech giants, Scope 3 emissions may encompass GPU purchases and the usage of their devices like smartphones and tablets.

Google combines two categories of Scope 3 emissions—capital goods and the use of sold products—though it remarks that the latter is minimal. (Most of Google’s hardware consists of small devices that consume little electricity.) This likely designates data centers as the main contributor. Last year, Google’s Scope 3 emissions increased by 2.1 million metric tons, effectively doubling since 2019, which serves as the baseline year for performance assessment.

Amazon’s rising Scope 3 emissions primarily stem from capital goods and energy consumption. The former can include data centers and warehouses, which clarifies why Amazon’s Scope 3 emissions surpass those of Google. A major factor indeed arises from data centers. “To meet substantial customer demand, in 2025 we expanded our data center capacity more than any other company, adding over 1.2 gigawatts (GW) in Q4 alone,” stated Amazon in its report.

Facing Hurdles

This expenditure explains why decarbonization has suddenly become more challenging. In the past, the energy consumption from offices and smaller data centers substantially contributed to their carbon footprints, which could be alleviated through investments in renewable energy.

However, AI has disrupted this paradigm. While tech companies could continue to leverage renewables along with battery systems for operating data centers, there is a tendency to revert to fossil fuels. This trend complicates their net-zero commitments, although not irreparably.

The emissions from constructing and outfitting data centers are particularly concerning. The steel and cement industries are notable polluters, and while startups are investigating low- to zero-carbon alternatives, they currently lack the scale to meet the demands of tech companies.

Additionally, the GPUs and memory chips driving the AI boom entail substantial energy consumption, with many advanced chip factories located in Asia, where fossil fuels dominate the energy grid. Compounding the issue, numerous chemicals used in semiconductor manufacturing are potent greenhouse gases, capable of warming the atmosphere far more than a similar volume of CO2. The spike in chip demand has likely exacerbated the carbon footprints of both Amazon and Google.

Nonetheless, these challenges are not insurmountable; however, Amazon, Google, and their peers face substantial obstacles. To meet their net-zero commitments, they will need to significantly boost their renewable energy purchases, invest heavily in advanced steel and cement production, and procure millions of tons of carbon removal credits. While this remains achievable, their reliance on AI has certainly complicated the process.