Google and Amazon Reveal the Rising Cost of the AI Boom

Artificial intelligence has long been recognised as one of the most resource-intensive technologies, consuming far more electricity and water than previous generations of digital computing. Now, newly released sustainability reports from Google and Amazon provide a clearer picture of the environmental impact associated with the rapid expansion of AI.

Both companies published their latest sustainability reports this week, revealing a noticeable rise in carbon emissions despite longstanding commitments to achieve net-zero emissions in the years ahead. Google reported that its overall carbon emissions have increased by 25% compared with last year, while Amazon recorded a 16% rise over the same period.

A closer examination of the reports suggests that both companies may need to make significant—and potentially expensive—changes to their operations if they hope to meet their long-term climate commitments.

Neither Google nor Amazon directly attributes the increase in emissions to artificial intelligence. However, the reports contain multiple indicators suggesting that the rapid growth of AI infrastructure is playing a major role.

AI appears to be driving the increase.

Both companies acknowledge that their energy consumption has risen substantially during the past year as demand for AI services has accelerated. They also highlight carbon intensity—the amount of carbon pollution generated per dollar of revenue earned—a metric that has previously featured in international climate discussions, including negotiations involving China, even as the country’s emissions continued to rise. In addition, both reports dedicate significant space to explaining how artificial intelligence could help improve environmental sustainability, underscoring the growing importance of AI within their broader climate strategies.

The data becomes more revealing when examined in greater detail. Google and Amazon have generally managed to limit emissions associated with purchased electricity through years of investment in renewable energy. However, that progress may become more difficult to maintain as technology companies increasingly invest in natural-gas power generation to meet the rising electricity demand driven by AI data centres.

Instead, much of the increase comes from what are known as Scope 3 emissions. These emissions include pollution generated through activities outside a company’s direct control, such as purchased goods and services or products used by customers. For companies like Google and Amazon, this category includes items such as graphics processing units (GPUs), data centre equipment, and consumer hardware.

Google combines capital goods and the use of sold products in its Scope 3 reporting, noting that emissions from customer use of its hardware remain relatively insignificant because most of its devices consume relatively little electricity. That leaves investment in data centres as the most likely contributor. During the past year, Google’s Scope 3 emissions increased by approximately 2.1 million metric tonnes, bringing them to roughly double the level recorded in 2019, the baseline year the company uses for measuring climate progress.

Amazon’s increase in Scope 3 emissions is largely linked to capital goods, fuel, and energy. Capital goods include major infrastructure such as warehouses and data centres, helping explain why Amazon experienced an even larger increase than Google. The company noted in its report, “To meet strong customer demand, in 2025 we added more data centre capacity globally than any other company, including more than 1.2 gigawatts (GW) in Q4 alone.”

The growing challenge of reaching net zero

The scale of this infrastructure expansion illustrates why reducing emissions has become increasingly difficult. In previous years, the largest contributors to carbon emissions were office buildings and comparatively smaller data centres, making it easier to offset emissions through renewable electricity purchases.

The rapid expansion of AI has fundamentally changed that equation. Although renewable energy, combined with battery storage, can still power many facilities, several technology companies are increasingly turning to fossil-fuel generation to meet soaring electricity demand. That shift could make their net-zero commitments more difficult to achieve, although it does not necessarily make those goals impossible.

An even greater challenge comes from constructing and equipping new data centres. Industries that produce steel and cement remain among the world’s largest sources of carbon emissions. While several emerging companies are developing low-carbon alternatives, those technologies have yet to scale to the levels required by major technology firms.

The AI boom has also sharply increased demand for GPUs and advanced memory chips. Semiconductor manufacturing is highly energy intensive, and many of the world’s most advanced fabrication facilities operate in regions where electricity grids continue to rely heavily on fossil fuels. In addition, several chemicals used during chip manufacturing are themselves extremely potent greenhouse gases, capable of trapping thousands of times more heat than an equivalent amount of carbon dioxide. The rapid expansion of chip production has likely contributed significantly to the growing carbon footprints reported by both Amazon and Google.

Despite these challenges, the situation is not viewed as impossible to address. To honour their net-zero commitments, Amazon, Google, and other major technology companies will likely need to accelerate renewable energy investments, support cleaner methods of producing steel and cement, and purchase millions of tonnes of carbon removal credits. Achieving those objectives remains possible, but the industry’s rapid embrace of artificial intelligence has made the path considerably more demanding.