← Back to feed

Microsoft, Amazon, and Google released their annual sustainability reports over the past several weeks. Taken together, the numbers tell a consistent story: AI infrastructure is erasing years of progress on corporate climate commitments.

The Numbers

In the financial year ending March 2026, the three companies collectively emitted 119 million metric tonnes of carbon dioxide equivalent (mTCO₂e). The prior year’s figure was approximately 101 million mTCO₂e — equivalent to Czechia’s total 2024 emissions. The 2026 total puts the three companies at roughly a third of France’s national emissions.

Microsoft: Carbon emissions increased 25% to 20 million mTCO₂e, “driven primarily by the expansion of our datacentre infrastructure,” per the company’s own report.

Google: Emissions up 18%, attributed to “increases in supply chain activities that supported the rapid expansion of our business.” Google separately claims its AI systems helped reduce emissions elsewhere by 41 million tonnes — a number that, if accurate, would exceed its own increase.

Amazon: Overall emissions up 16%. Supply chain emissions — which include data center construction — up 20%. The report describes this as “making progress” toward a 2040 net-zero target.

What Changed

Microsoft’s emissions had flatlined at approximately 16 million mTCO₂e in both 2023 and 2024. The jump to 20 million in FY2026 represents a step-change, not a gradual drift. The company’s sustainability report also notes there are now fewer carbon credits available on global markets, reducing the offset toolkit that companies had been relying on.

The broader backdrop: the global tech sector is on track to spend $765 billion this year on AI infrastructure, the majority on data center construction. JLL estimates 1,200 data centers will be built globally between now and 2030.

The Uptime Institute — which rates and inspects data centers — estimates that large projects announced last year alone would, if fully built, consume 1.3% of the world’s current total electricity usage. The majority of that new demand comes from US projects.

The Contradiction

All three companies maintain net-zero commitments: Google and Microsoft by 2030, Amazon by 2040. None of those timelines has changed. The math has.

Cecilia Rikap, an economics professor at University College London, noted a secondary effect that rarely surfaces in these reports: when companies migrate workloads to cloud infrastructure, they shift their AI-related emissions to hyperscaler balance sheets. The hyperscalers’ reported figures therefore represent a floor, not just their own operations.

Shaolei Ren, an electrical engineering professor at UC Riverside, flagged a compounding problem: carbon credit markets are thinning at the same time demand for offsets is accelerating. “Everyone is talking about the lack of physical goods and infrastructure like power,” he said, “but there may also be a lack of virtual goods — carbon credits.”

Stakes

The scale of the AI buildout makes this trajectory difficult to reverse quickly. The companies have committed to multi-decade infrastructure contracts — including multi-gigawatt data center leases and chip supply deals stretching into the early 2030s. The emissions curve is a downstream consequence of capital allocation decisions that are now largely locked.

Whether net-zero commitments at 2030 or 2040 are achievable from this baseline, against this spending trajectory, is a question the annual sustainability reports do not answer.