AI Is Making Google and Amazon’s Carbon Emissions Goals Harder to Hit

For years, Google and Amazon presented themselves as leaders in corporate sustainability, companies that had taken the energy-hungry tech industry and bent it toward renewables, carbon offsets, and net-zero commitments. Those claims just got a lot harder to make. Both companies released their 2026 sustainability reports this week, and the numbers are stark. Google’s total carbon emissions are up 25% compared to last year. Amazon’s are up 16%. Neither company points the finger directly at AI. But anyone reading the data closely will reach the same conclusion: the AI boom is the primary driver behind both companies’s surging environmental footprint, and AI carbon emissions are now one of the most serious reputational and regulatory risks facing the tech industry.

What the Numbers Actually Show

The biggest contributor to both companies’ growing emissions is not energy for offices or ordinary cloud computing workloads. It is what accountants call Scope 3 emissions — a broad category that covers pollution a company does not directly control, including the goods and services it purchases and the products it sells.

For Google and Amazon, Scope 3 includes things like GPU purchases from chip manufacturers, the construction of new data centers, and the energy used by customers running their products. These are all categories that have exploded alongside the AI buildout.

Google’s Scope 3 emissions increased by 2.1 million metric tons in the past year alone. They are now double what they were in 2019 — the year Google uses as its baseline for measuring environmental progress. That trajectory is moving in exactly the wrong direction for a company that has pledged to achieve net-zero emissions.

Amazon’s rising Scope 3 footprint mostly comes from capital goods and fuel and energy — categories that include data centers, warehouses, and the power that runs them. The company disclosed that in 2025, it added more data center capacity globally than any other company, including more than 1.2 gigawatts of new capacity in Q4 alone. That kind of infrastructure buildout has an enormous carbon cost before a single AI query is ever processed.

The Renewable Energy Problem

For most of the last decade, Google and Amazon managed to keep their emissions relatively flat by purchasing large quantities of renewable energy. Wind and solar power purchases could cancel out the emissions from running data centers, keeping the companies’ balance sheets looking green even as their infrastructure grew.

AI has broken that approach. The power demands of training and running large AI models are simply too large and too concentrated to be offset cleanly by renewable purchases. Both companies are increasingly investing in natural gas power plants — fossil fuel infrastructure — to meet the electricity demands of their AI data centers. That shift means their direct energy emissions are likely to rise even before accounting for the construction and chip supply chain effects.

Both companies still talk up renewables in their reports, and both are investing in solar and wind. But the International Energy Agency and independent analysts have consistently flagged that the pace of data center power growth is outrunning the pace at which renewable capacity can be brought online. The numbers in these two sustainability reports are starting to confirm that in black and white.

GPUs, Steel, and Cement: The Hidden Carbon Costs

Beyond electricity, there are two other major carbon cost centers that the AI boom has amplified: semiconductor manufacturing and physical data center construction.

Building a modern AI data center requires enormous quantities of steel and cement — two of the most carbon-intensive industries in the global economy. While startups are working on lower-carbon approaches to both materials, those solutions are not yet available at the scale that Google and Amazon need them. Every new data center is therefore a significant Scope 3 emissions event.

Then there is the GPU supply chain. Semiconductor manufacturing consumes vast amounts of energy. Many of the world’s leading chip factories — including those producing the Nvidia GPUs that power most AI training — are located in Taiwan, South Korea, and Japan, where electrical grids still rely heavily on fossil fuels. The chemicals used in chip production are also a serious concern: some of the gases used in semiconductor fabrication are greenhouse gases that can warm the atmosphere thousands of times more than an equivalent amount of CO2. A chip-buying spree at the scale Google and Amazon are running has a compounding environmental impact that their reports are only beginning to capture.

Why Both Companies Are in a Difficult Position

Neither Google nor Amazon uses the word “AI” when explaining their rising emissions. Their reports instead lean on language about “energy-intensive workloads,” “customer demand,” and “expanding infrastructure.” This is technically accurate but strategically careful — both companies have made public commitments to ambitious climate targets, and acknowledging that their flagship growth product is the main thing blowing those targets up is an uncomfortable message.

The approach also carries risk. Regulators in the European Union, the United Kingdom, and increasingly the United States are scrutinizing corporate sustainability reporting with much greater rigor than they have in the past. A gap between a company’s stated climate ambitions and the trajectory shown in its own data is exactly the kind of thing that draws investigations and enforcement actions.

Is There a Way Out?

The problems driving rising AI carbon emissions at Google and Amazon are real, but they are not permanent. The companies have three main levers available to them.

First, they can accelerate renewable energy procurement and invest in long-duration energy storage that can smooth out the intermittency of wind and solar. Second, they can invest in or contract with companies building low-carbon steel, cement, and semiconductor manufacturing — markets that are developing, if slowly. Third, they can purchase large quantities of high-quality carbon removal credits — not cheap offsets, but verifiable, durable removals — to cover emissions they cannot yet eliminate directly.

None of these is easy or inexpensive at the scale required. But the alternative — continuing to let the AI buildout drive emissions higher while maintaining the posture of climate leadership — is not sustainable either, in either the environmental or the reputational sense.

The 2026 sustainability reports from Google and Amazon are a milestone document: the moment when the AI industry’s environmental cost stopped being a future concern and became a measurable, year-over-year reality.

Read more tech related articles here.