Brownfield AI Data Centers: How Tailored Engineering Can Turn a Head Start Into Commercial Value

Speed to market is becoming critical in AI data center development. As demand for AI services grows, AI labs need more computing capacity and facilities ready to support their hardware. Yet, as SemiAnalysis’s article on SpaceX’s infrastructure plans highlights, finding that capacity at scale and within a short timeframe remains difficult. This creates an opportunity for developers who can bring suitable infrastructure into operation sooner.

Brownfield development can help meet that need. Reusing existing infrastructure can mean less to build and an earlier start to operations. But turning that head start into economic value requires an engineering and delivery strategy tailored to the site, so that conversion work does not outweigh the benefits of reuse.

Drawing on the brownfield discussions at DCD Connect London, this article considers those opportunities and trade-offs, using Sauda to illustrate how Terakraft has approached them in practice.

Turning existing assets into project value

A brownfield site can give a developer a head start, with advantages that vary by location. An existing power connection, for example, can be particularly valuable where securing new supply is constrained by approvals or grid capacity, as discussed in the IEA’s recent Electricity report.

But having those assets is only the starting point. As Terakraft CEO Giorgio Sbriglia explained at DCD Connect London, power, buildings and permits must be assessed together to understand what determines the delivery timeline. Engineering then defines how to convert the site into an AI data center, establishing what can be reused and what work remains.

This is where more time spent upfront can save time later. Drawing on Terakraft’s experience, engineering may take 6–12 months for a brownfield project, compared with 3 months or more for greenfield development. Yet that additional design effort can remove substantial work from construction. At Sauda, it has helped support a construction timeline of around 12 months, compared with an indicative 24 months or more for greenfield development.

Permitting can provide another head start where existing industrial land use simplifies the approvals required. At Sauda, the process took around 12 months, against an indicative 18–24 months or more for greenfield development, although timelines vary with the proposed use and local requirements. Engineering, permitting and construction can overlap: the real advantage lies in how they come together to bring operations forward.

For investors, this is where tailored engineering can translate into concrete commercial value, reducing both the capital required and the time to revenue. At Sauda, retaining suitable infrastructure has helped bring the CAPEX budget to around €5 million per MW, excluding servers, compared with a greenfield reference of approximately €10 million per MW.

Alongside this lower capital requirement, a shorter delivery programme can reduce the period during which funds are committed before the facility begins operating and generating revenue.

Beyond the financial benefits, a brownfield approach can also create value for the local community. Bringing unused facilities back into productive use can limit development on additional land, while restoring neglected industrial buildings can improve the surroundings. These potential benefits provide a basis for dialogue about the project and its impacts, which can also help the redevelopment progress.

Where brownfield AI data centers can lose their head start

The same assets that give brownfield AI data centers a head start can also introduce constraints. As discussed in the DCD roundtable on retrofit and brownfield investment, existing space can limit equipment choices, while adapting a facility for higher capacity can require substantial modifications. A preferred cooling unit may not fit the building, turning an expected saving into additional design and construction work.

Equipment availability can further change the delivery plan. Even with a building and power connection in place, new electrical or cooling systems may introduce procurement delays. The developer must weigh these costs and timing implications against the benefits of reuse before committing to the conversion.

At Sauda, this balance shaped our decisions about both the design and the organization of construction.

Putting the brownfield approach into practice: Sauda AI data center

Sauda I, Norway: the original hydropower station (left) and the same building converted into Terakraft’s AI data center (right)

At Sauda in Norway, we set out to convert a former hydropower facility into a high-density AI data center designed for GPU hosting and liquid cooling colocation. The building and existing power infrastructure provided the starting point, while access to lake water informed the cooling design.

As Giorgio explained at DCD Connect London, the modular solution initially considered did not fit the facility. We moved to site-specific engineering and construction, developing the design around the building and its intended workloads.

The same flexibility carried into execution. When the original turnkey route for the mezzanine faced delays, we divided the work into separate packages and engaged contractors with immediate availability, as documented in our December 2025 construction update.

Smaller packages also opened the work to local companies. Their proximity and established working relationships reduced mobilization needs. Restructuring the contracts addressed a schedule constraint while enabling more businesses in the area to participate.

Sauda shows how Terakraft is putting a brownfield approach into practice, adapting the design to existing resources and reorganizing construction when constraints emerge. These decisions support progress toward operations and illustrate what is needed to turn a site’s initial head start into economic value: making use of suitable assets while managing the cost and time required for conversion.