Both companies’ latest sustainability reports suggest a difficult reality: rapid AI expansion is making ambitious net-zero targets harder to achieve.
Two of the biggest names in cloud computing just admitted their climate math isn’t adding up. Google and Amazon released their annual sustainability reports this week.
Buried in the spreadsheets is a warning every AI company should be reading closely.
The Numbers Nobody Wanted to Publish
According to their latest sustainability disclosures, Google’s emissions rose 25% year over year, while Amazon reported a 16% increase. Both companies have public net-zero pledges. Both are now further from hitting them than they were twelve months ago.
Neither report names AI as the culprit outright. But the timing, and the categories driving the increase, tell their own story.
Here’s the tell: emissions from the electricity these companies buy are actually holding steady. Years of renewable contracts get credit for that.
The real spike is in Scope 3 emissions — pollution tied to a company’s supply chain, not its direct operations. Google’s Scope 3 footprint jumped 2.1 million metric tons last year, now double its 2019 baseline. Amazon’s growth traces largely to infrastructure spending, including a reported 1.2 gigawatts of new data center capacity added in Q4 alone.
Translation: it’s not the servers running that’s the problem. It’s building them in the first place.
Why This Breaks the Old Climate Playbook
Most sustainability coverage treats these reports as a scorecard — did the number go up or down? That misses the real story: AI has changed where Big Tech’s emissions come from, and the new source is much harder to engineer away.
For years, tech companies offset their biggest climate liability — data center electricity — by signing more renewable power deals. Solvable problem, known playbook.
AI broke that playbook. Data center construction runs on steel and cement, two of the most carbon-intensive industries on Earth, and low-carbon versions of both aren’t ready to scale.
Chip manufacturing adds another layer. Fabrication is energy-hungry, much of it happens on fossil-fuel grids, and some of the chemicals involved trap heat thousands of times more effectively than CO2. None of it shows up on a company’s energy bill. All of it shows up in Scope 3.
There’s a quieter tell too. Both companies lean on “carbon intensity” — emissions per dollar of revenue — as their framing metric. Carbon intensity matters because it measures economic efficiency, but critics note that improving intensity can still coincide with rising total emissions. When a report emphasizes intensity over totals, the totals usually aren’t the flattering number.
Two Things to Watch Next
Expect more tech companies to follow Google into natural gas as a data center stopgap, even while they keep promoting renewables. Many companies increasingly view natural gas as a near-term bridge because renewable deployment, storage, and transmission upgrades aren’t expanding quickly enough to match AI infrastructure timelines.
Expect the pressure to shift downstream, too. Chipmakers, steel producers, and cement manufacturers are about to hear a lot more from their biggest customers about their emissions.
None of this puts net-zero out of reach. Google and Amazon can still get there through heavier renewable investment, funding cleaner steel and cement production, and buying carbon removal at scale.
But “still possible” and “on track” are different sentences. This week’s reports quietly moved both companies from the second column to the first.
The AI boom didn’t just raise Big Tech’s power bills. It rewrote which part of the business is actually driving the pollution — and that’s a much harder problem to buy your way out of.
Related: AI Is Quietly Draining the World’s Water — And No One Is Counting
