Scope 3 emissions in AI data centers

Manfred Engelhard, Senior Director Technology Management at Exyte, smiles while working at a computer. He plays a key role in Exyte’s mission to reduce Scope 3 emissions in AI data centers through sustainable design, smarter materials, and innovative energy systems.

As AI reshapes the modern world, the challenge becomes clear: how can Exyte support it without compromising the environment that sustains us? Manfred Engelhard, Senior Director Technology Management, Data Centers explains more in this article.

Reducing scope 3 emissions in AI data centers

AI depends on high-density compute, advanced cooling systems, and enormous energy loads. Every breakthrough in generative intelligence is backed by data centers consuming resources at a scale that is unsustainable, unless the industry rethinks how to design and deliver them.

Companies cannot afford to stand by, and Exyte is planning, engineering, and building for what is next. This also means confronting a hard truth: If AI is to be a force for good, companies must address Scope 3 emissions with the same urgency and innovation driving the technology itself.

The unseen footprint of AI infrastructure

Most discussions around carbon neutrality focus on Scope 1 and Scope 2 emissions. These cover direct emissions from operations and indirect emissions from purchased energy. Scope 3 and the emissions linked to building and supply chain activities, are the biggest slice of the carbon pie.

This is especially true for AI-ready data centers, which are more complex and resource-intensive than traditional facilities. Steel, concrete, and long-haul logistics are difficult to avoid, but they aren’t impossible to manage. Exyte is not content with the status quo. The company’s mission is to confront this challenge through smarter design, better materials, supply chain planning and cleaner execution.

Sustainability begins at the concept stage

“Our approach at Exyte starts with planning for the full lifecycle of the data center, not just its operational phase,” says Engelhard. “From concept to commissioning, every decision is an opportunity to reduce environmental impact.”

That means choosing the right materials and evaluating alternatives. In some locations, Exyte proposes stone piling instead of concrete foundations to lower embedded carbon. There is also a priority to use materials that come with Environmental Product Declarations (EPDs), giving visibility into Scope 3 impacts early in the process.

Engelhard adds: “Additionally, Exyte uses digital tools like the Embodied Carbon in Construction Calculator (EC3), which helps us model and compare carbon values of building components. This allows us to guide decisions based on data, not assumptions.”

Building smarter with every detail

Sustainability isn’t just about grand gestures. It is about doing many small things right, consistently and at scale.

One such detail is the use of hydrotreated vegetable oil (HVO) as an alternative to traditional diesel. This cleaner-burning renewable fuel significantly reduces carbon emissions without requiring any equipment adaptations. Similarly, just-in-time delivery models are being adopted to minimize idle time on construction sites, cutting down on unnecessary fuel consumption, and improving overall efficiency.

Material choices are also evolving. Low Carbon Concrete, for instance, is gaining traction as a sustainable alternative to conventional concrete, offering a reduced carbon footprint without compromising structural integrity. Meanwhile, construction machinery is being upgraded with low-emission technologies, including the use of AdBlue to reduce nitrogen oxide (NOx) emissions.

Lean construction principles are another cornerstone of this smarter approach. By focusing on eliminating waste and streamlining processes, these methods not only enhance productivity but also contribute to environmental goals.

Even minor upgrades such as choosing cleaner fuel or reducing unnecessary transport contribute to a meaningful emissions reduction when implemented across large projects.

Exyte is actively involved with organizations like the Lean Construction Institute to help standardize these practices and bring efficiency to the forefront of data center construction.

Energy systems that support more than AI

Looking ahead, the goal is to create energy efficient facilities that do not only consume energy but contribute to the broader energy ecosystem. This is especially important as AI workloads surge and utility grids feel the pressure. The on-site generators can be planned with a bi-directional grid coupling and utilized as peak shaving grid support if they are not needed for the DC.

Exyte is already integrating provisions for heat reuse systems being fed from the Data Center facilities that allow waste heat from servers to support local district heating networks. In Germany, where new energy efficiency laws will soon require this, Exyte is designing ahead of regulations.

“In other regions, project teams are exploring innovative energy coupling strategies,” explains Engelhard. “For example, capturing cryogenic energy during re-gasification from LNG terminals and using it to drive data center cooling systems is a solution that turns waste into value.”

Another concept under development involves powering data centers with fuel cells and gas engines running on hydrogen / natural gas blends, with optional carbon capture systems in place. The captured CO2 could be used in industrial processes or even converted into synthetic fuels, supporting a broader shift toward a circular carbon economy.

Partnering for impact

Sustainability is a shared expectation between builders, clients, and the communities around them. To foster this, Exyte works closely with hyperscalers and cloud providers to meet their climate goals, while aligning with the company’s own ESG roadmap toward carbon neutrality by 2040.

AI has raised the bar, and Exyte is rising to meet it. But it will take transparency, collaboration, and a willingness to rethink typical build methods. Engelhard says: “Scope 3 is complex. We can make significant progress if we start early, act decisively, and stay committed to this. Because, in the end, the question is not just whether we can build data centers fast enough to support AI, it is if we can do it responsibly and build a future we are proud of.”