Low-Carbon Concrete and Steel That Cut Embodied Emissions by 30–65% in AI Data Centers

Neil L. Rideout

9/30/20267 min read

Low-Carbon Concrete and Steel That Cut Embodied Emissions by 30–65% in AI Data Centers

The AI revolution is driving an unprecedented expansion of data center infrastructure. Hyperscale facilities, high-density computing environments, and specialized AI campuses are being planned and constructed at a pace that would have been difficult to imagine only a few years ago.

Much of the discussion around sustainable data centers focuses on operational energy: renewable electricity, power usage effectiveness, efficient cooling, and increasingly sophisticated power management. These are critical considerations, but they address only part of the environmental footprint.

There is another carbon cost that is effectively locked in before the first server is switched on: embodied carbon.

Embodied emissions are generated through the extraction, processing, manufacturing, transportation, installation, maintenance and eventual disposal of construction materials. For data centers, two materials deserve particular attention: concrete and steel.

Recent industry experience demonstrates that strategically specifying lower-carbon concrete and steel can produce reductions in material-related embodied emissions in the range of 30% to 65% in appropriate applications, although the actual result depends heavily on material specifications, structural design, regional supply chains, electricity sources and the baseline against which the project is measured. AWS, for example, has reported reductions of more than 30% in the embodied carbon of certain concrete mixes, while low-carbon steel pathways can deliver substantially larger reductions compared with conventional blast-furnace production.

For AI data center developers, this represents an important opportunity.

Why embodied carbon matters in AI infrastructure

A conventional data center is already a substantial construction project. AI data centers can push that challenge further because they often require enormous electrical infrastructure, high-capacity cooling systems, reinforced structures, large foundations and significant quantities of structural steel.

The result is a facility containing thousands of tonnes of construction materials.

Research published in 2026 found that concrete and steel remain among the largest contributors to embodied carbon in buildings, with concrete accounting for approximately 38% and steel approximately 24% of embodied emissions in the study's projected 2050 North American building baseline.

Data center-specific analysis also demonstrates the importance of the issue. Uptime Institute has estimated that concrete and steel can represent up to approximately half of the building-shell product carbon footprint of a data center.

This creates an important distinction:

A data center can operate on clean electricity and still have a substantial carbon footprint associated with its construction.

That is why embodied carbon needs to be addressed during site selection, engineering and procurement, rather than treated as a sustainability issue after construction has begun.

Concrete: one of the biggest opportunities

Concrete is indispensable to most data center projects.

Foundations, equipment pads, slabs, structural elements, walls, loading areas and other components can require enormous volumes of concrete.

The principal source of concrete's embodied carbon is cement. Cement manufacturing requires high-temperature processing and produces significant process emissions. As a result, reducing the quantity or carbon intensity of cement within a concrete mixture can substantially reduce the material's footprint.

There are several approaches available today.

1. Supplementary cementitious materials

Supplementary cementitious materials, or SCMs, can partially replace conventional Portland cement.

Common examples include:

Slag cement
Fly ash
Silica fume
Calcined materials
Metakaolin
Other industrial or naturally occurring mineral materials

The National Ready Mixed Concrete Association identifies SCMs as one of the major opportunities available today for reducing concrete's carbon footprint. Its guidance notes that replacing a substantial portion of Portland cement with fly ash or slag can produce significant reductions while maintaining required concrete performance when properly engineered.

AWS provides a real-world example. For one data center project in Northern Virginia, its supplier was able to replace 40% of ordinary cement with slag while maintaining the project's requirements and schedule, reducing embodied carbon in the cement mix by more than 30%.

This is particularly important for AI data centers because sustainability does not eliminate the need for structural performance.

Concrete still needs to meet requirements for:

Compressive strength
Durability
Freeze-thaw performance
Shrinkage
Workability
Placement
Curing
Fire performance
Construction schedules

The objective is therefore not simply to specify "green concrete."

The objective is to specify performance-based low-carbon concrete.

2. Portland-limestone cement

Another established approach is Portland-Limestone Cement, commonly known as PLC or Type IL in North America.

PLC can reduce the clinker content of cement while maintaining the performance characteristics required for many applications.

AWS has reported combining Portland-limestone cement with supplementary cementitious materials to achieve reductions of up to approximately 50% in certain concrete applications.

For large-scale data center construction, even a relatively modest percentage reduction becomes significant when multiplied across thousands of cubic metres of concrete.

3. Optimize the amount of concrete

Perhaps the simplest low-carbon concrete is the concrete that does not need to be poured.

Design optimization can reduce unnecessary material without compromising structural requirements.

This can involve:

Optimizing slab thickness
Reducing unnecessary concrete topping
Redesigning equipment pads
Eliminating unnecessary duct-bank concrete
Optimizing foundation geometry
Using appropriate structural spans
Reducing overdesign
Substituting other materials where technically appropriate

Meta has reported eliminating concrete requirements from certain electrical and telecommunications duct banks in newer data center designs, with some design changes reducing the carbon footprint associated with concrete by more than 30% compared with previous designs.

This illustrates a fundamental principle:

Embodied-carbon reduction should begin with design, not procurement.

Buying a lower-carbon version of the same building is useful. Designing a building that requires less material in the first place can be even better.

Steel: an equally important opportunity

Steel presents a different decarbonization challenge.

Traditional steelmaking frequently relies on blast furnaces and basic oxygen furnaces, with coal playing an important role in the process.

Alternative production pathways include electric arc furnaces (EAFs), which can use substantial quantities of recycled steel scrap.

The carbon intensity of EAF steel varies considerably depending on scrap content, electricity sources, production technology and other factors. That means simply specifying "EAF steel" is not enough.

The better approach is to require measurable environmental performance.

Environmental Product Declarations, or EPDs, can provide project teams with standardized information about the environmental impacts associated with particular products.

The Carbon Leadership Forum's current material-baseline work includes data for structural steel and other construction products, while its EC3 platform allows designers and procurement teams to compare material embodied-carbon information using third-party verified EPD data.

The potential of recycled and electric-arc-furnace steel

AWS reports that EAF steel used in its data center construction can commonly have approximately half the embodied carbon of conventional BOF steel, with some products achieving considerably lower levels.

Other industry examples demonstrate how dramatic the difference can become.

world steel reports that an HBIS direct-reduced-iron/electric-arc-furnace process reduced crude-steel carbon intensity by approximately 65% compared with the company's original blast-furnace/basic-oxygen-furnace process.

These numbers should not be interpreted as a universal 65% reduction for every EAF steel product. Electricity generation, raw materials and production methods matter enormously.

For a data center developer, that means the procurement specification should focus on verified carbon intensity rather than marketing terminology.

Designing with less steel

As with concrete, the lowest-carbon steel is often the steel that does not need to be manufactured.

Structural optimization can reduce steel requirements through:

Higher-strength structural steel
More efficient structural grids
Optimized column spacing
Improved beam design
Reduced unnecessary structural redundancy
Computational structural optimization
Modular construction
Prefabrication
Efficient equipment-support structures

AWS has reported using higher-strength structural steel to achieve a higher strength-to-weight ratio, potentially allowing less steel to provide the same structural function.

This is particularly relevant to AI data centers because their structural requirements can be substantial.

The goal should not be to minimize structural material at the expense of safety.

The goal is to maximize structural efficiency.

Where the 30–65% reduction comes from

The headline range of 30–65% should not be interpreted as saying that every AI data center can simply reduce its total embodied carbon by 65%.

Instead, it represents the range of reductions that can become possible across particular materials, systems or project strategies.

For example:

Low-carbon concrete:
Replacing a substantial proportion of conventional cement with lower-carbon binders can produce reductions of 30% or more in the affected concrete mix. AWS has documented a greater-than-30% reduction in the cement mix of one data center project.

Low-carbon steel:
EAF, recycled-content and emerging DRI/EAF pathways can produce very large reductions compared with conventional steelmaking, with specific documented processes reaching approximately 50–65% reductions or more depending on the baseline.

Whole-building strategies:
Material optimization, efficient structural design and low-carbon products can compound their benefits. RMI research summarized by the Carbon Leadership Forum found that upfront embodied carbon in studied building typologies could be reduced by up to 46%, with some strategies requiring less than a 1% cost premium.

The important point is that 30–65% is a potential reduction range, not a universal guarantee.

AI can help make the buildings themselves smarter

There is an interesting irony in the AI infrastructure boom.

The technology driving demand for these enormous buildings can also help optimize the materials used to construct them.

Meta has already demonstrated this concept.

The company worked with Amrize and the University of Illinois to develop an AI-based system for optimizing concrete mixtures. The resulting formulations were tested and deployed at a Meta data center construction site in Rosemount, Minnesota. The system considered strength, workability, curing speed and sustainability simultaneously.

This points toward a future in which AI is used not simply to operate data centers but to design the physical infrastructure supporting AI itself.

Structural engineering, concrete formulation, construction sequencing and procurement can increasingly become data-driven optimization problems.

What data center developers should require

For a new AI data center, low-carbon construction should be incorporated into the project requirements from the beginning.

A practical procurement strategy should include:

1. Establish a baseline

Determine the expected embodied-carbon intensity of conventional concrete, reinforcing steel and structural steel.

2. Require EPDs

Require suppliers to provide current, third-party verified Environmental Product Declarations where available.

3. Set carbon-performance targets

Rather than specifying a particular material solely because it is marketed as sustainable, establish measurable global-warming-potential limits.

4. Optimize material quantities

Challenge engineers to determine whether the same performance can be achieved using less concrete and steel.

5. Conduct whole-building LCA

Material decisions should ultimately be evaluated at the building level, rather than looking at isolated products.

6. Protect performance requirements

Carbon reduction should never override structural safety, reliability, durability, fire resistance or critical-facility requirements.

7. Lock in procurement early

Low-carbon materials can have regional availability constraints. Early engagement with concrete and steel suppliers can identify what is realistically available before construction begins.

The Carbon Leadership Forum's model specifications specifically demonstrate how EPD requirements and global-warming-potential limits can be incorporated into construction specifications for concrete, rebar and structural steel.

The next generation of AI data centers

The AI data center of the future will need to be efficient in more than one dimension.

It will need to use electricity efficiently.

It will need highly efficient cooling.

It will need resilient power infrastructure.

And increasingly, it will need to consider the carbon associated with the physical materials used to construct the facility.

The industry's largest developers are already moving in this direction. AWS reports construction of data centers using lower-carbon concrete and steel, while Meta is deploying low-carbon concrete approaches and investigating additional low-carbon structural systems.

The message for developers is straightforward:

Sustainability does not begin when the servers are turned on. It begins when the first foundation is designed.

Concrete and steel will remain fundamental to high-performance data center construction for the foreseeable future. The opportunity is to make those materials substantially more efficient, more transparent and less carbon-intensive.

For the rapidly expanding AI infrastructure sector, that could mean building facilities that deliver enormous computing capacity while significantly reducing the emissions embedded in the infrastructure itself.

The data center industry has spent years optimizing watts per computation.

The next frontier is optimizing carbon per building.

And with the right combination of low-carbon concrete, lower-carbon steel, material efficiency, verified EPD data and intelligent structural design, the industry has an opportunity to cut a substantial portion of construction-related emissions before the first AI model ever runs inside the building.

Contact

Head Office

Green Life Enterprises LLC
8 Biltmore Est
Suite 310,
Phoenix, AZ 85016

greenlifedatacenters@gmail.com

+1-813-220-0001

Canadian Office

Green Life Enterprises LLC
3142 Nicholson Ave
Suite 10
New Waterford, Nova Scotia B1H 1N8