The villain of the moment

Every era picks an industrial villain. In the 1970s it was the smokestack. In the 2000s it was the SUV. Today it is a windowless warehouse full of humming GPUs.

The case against data centres writes itself, and much of it is fair. A planned Google campus in rural Virginia could draw up to 2 million gallons of water a day . Gallup found that 71% of Americans oppose an AI data centre being built near them. In Canada, a Leger poll found 81% of respondents worry these facilities will spike their power bills. Local opposition has already blocked or delayed tens of billions of dollars in U.S. projects.

But paradoxically, what if the data centre boom ends up being one of the best things to happen to climate technology this decade?

According to the International Energy Agency, data centres account for roughly 180 million tonnes of CO2 a year , about 0.5% of global combustion emissions. In the IEA’s base case, that figure peaks around 320 million tonnes by 2030 and stays under 1% of the global total.

Now compare that with the sectors that actually make the modern world. Industry emits about 9 billion tonnes of CO2 a year , roughly 50 times the entire data centre fleet, and steel, cement and chemicals account for around 70% of it. Add oil and gas operations and fertilizer, and the centre of gravity of the climate problem is obvious. It is not in data centres. It is in kilns, blast furnaces and refineries.

A data centre’s footprint comes from two places. Operational emissions come from the electricity it burns and embodied emissions come from the steel, concrete and chips used to build it. The second is easy to overlook and enormous. Microsoft’s 2026 Environmental Sustainability Report shows its total emissions rose 25% in a single year, driven primarily by data centre expansion. Supply-chain (Scope 3) emissions still make up about 86% of its footprint, and emissions from capital goods, the servers and buildings, have more than doubled since 2020 .

The most important first customer in climate tech

Hyperscalers know all of this. Their licence to operate now depends on showing communities and regulators that they can build sustainably. So they are doing two things at once: Bring Your Own Clean Energy (“BYONCE”), and buying the low-carbon materials that don’t yet exist at scale.

Google signed the first corporate offtake for a gas plant with carbon capture , the 400 MW Broadwing Energy project in Decatur, Illinois, which will store about 90% of its CO2 underground. Earlier this month it announced a $17 billion data centre in Laramie County, Wyoming , powered by a 702 MW fuel cell project with carbon capture. Google says that brings its contracted carbon-capture power to nearly 1 GW.

Meta just signed long-term purchase agreements for low-carbon cement from Sublime Systems and Fortera, near-zero-emission steel from Stegra, electric trucking with Nevoya and recycled plastics from MacroCycle. The Fortera deal alone backs at least 200,000 tonnes of cement and helps finance the company’s planned commercial plant in Missouri.

Microsoft signed a 10.5 GW framework with Brookfield , nearly eight times larger than any prior corporate PPA. It agreed to buy near-zero-emission steel from Stegra for its data centres. And it contracted 70 million carbon removal credits to date , the large majority of the entire market.

Crossing the valley of death

To see why this matters, you have to understand what a new cement or steel company is up against. A conventional kiln or blast furnace was paid off decades ago. Its owner is selling product at marginal cost. The newcomer must finance a first-of-a-kind plant, carry technology risk, and still match that price. Banks won’t lend without a buyer, and buyers won’t sign without a plant. That is the valley of death, and most climate technologies die in it.

A 15-year offtake from a creditworthy hyperscaler breaks that loop. It is the document a lender needs to see to finance a project.

The ripple effects reach beyond any single deal. U.S. utilities now plan a record $1.4 trillion in capital spending over five years, and the Department of Energy’s draft transmission study names data centres as the main driver of new grid investment . A modernized grid serves everyone long after today’s models are obsolete.

None of this is guaranteed. AI models will get more efficient, and the IEA’s own scenarios show demand growth slowing if they do. Corporate appetite can also cool fast: Microsoft cut its carbon removal purchases by roughly 80% this year as AI spending climbed. A market resting on a handful of buyers is fragile.

The criticism of data centres is fair, and communities are right to demand better. But the climate will not be won or lost inside a server hall. It will be won or lost in the kilns, furnaces and power plants that have run the same way for a century, and that almost no one has been willing to pay to change. For the first time, someone is.

So maybe the headline asks the wrong question. Data centres will never be good for the climate on their own. But they may end up paying for the things that are.

Disclosure: Phil De Luna is co-founder and CTO of CURA, which is developing low-carbon cement and lime technology.