Retrofitting, not rebuilding: A sustainable approach to data centre expansion
Can refurbishing preloved facilities offer a greener, more economically conscious alternative to constant new builds?
As AI reshapes the world as we know it, the infrastructure supporting this transformation is being pushed to its limits. Globally, data centre growth is booming - with McKinsey estimating as much as $7 trillion in new data centre investment by 2030, over $1.2 trillion per year for the next five years. But with this scale comes an even greater sustainability challenge.
The conventional response to rising demand has been to build more: more square footage, more servers, more power. However, rushing to build is both unsustainable and unnecessary. Much of today’s buildout is focused on training large-scale, frontier AI models and research clusters. Yet there is equally growing demand for inference infrastructure and other applications that benefit from being closer to data sources and typically require less power. Many existing data centres are misused, underused, or inefficient, yet the technology is already available to modernise them into high-performance, AI-ready facilities without the need for new construction.
According to Synergy Research Group, just 20 metro areas account for 60% of global colocation capacity, placing huge strain on regional grids and resources. Meanwhile, in China, up to 80% of newly built data centres are reportedly sitting idle (according to Jiazi Guangnian and 36Kr) - an illustration of what happens when expansion outpaces strategic planning.
The solution: retrofitting legacy infrastructure
Retrofitting existing data centres offers a greener, more economically conscious alternative to constant new builds. By modernising what we already have, we can avoid overbuilding, reduce environmental impact, and increase operational resilience, especially in saturated hubs like Northern Virginia, Dublin, and Singapore, where land and energy are in short supply.
Other drivers make this approach compelling. Data residency and regulatory boundaries mean AI models certified for sectors like healthcare must often run close to the data in regulated environments. Cost is another driver, as refurbishing existing sites is significantly more cost-effective for some organisations than investing in new facilities. Another critical consideration is the carbon impact, as leveraging existing data centre infrastructure allows operators to reduce both operational emissions and the embedded carbon from construction and new equipment, including Scope 3 emissions.
If we continue clustering new builds in the same regions, the pace of AI progress will become unsustainable. Energy supply won’t keep up, and neighbouring communities and businesses will bear the burden. Long-term growth requires balance, optimising what is already built and where new facilities are needed, expanding into geographies with the infrastructure to support them.
Bridging the regulatory and sovereign divide
Beyond this, Sovereign AI is becoming a significant driver of capacity demand. Nations around the world are waking up to the strategic importance of AI - not just for innovation, but for national security, digital independence and data protection.
As a result, governments are accelerating investment in domestically controlled capacity and reassessing how existing infrastructure can be repurposed to meet these goals.
READ MORE: Local language models: Building a personal AI assistant with LM Studio
This shift highlights how national interest is now directly influencing data centre expansion strategies, and it’s a trend that deserves closer exploration in future analysis.
The industry must push beyond incentives and confront the environmental impact of rapid growth. Companies must take accountability for their carbon footprint or risk increased regulatory and public scrutiny.
Accountability and optimisation with Data Centre Infrastructure Management
Ensuring accountability and smarter decision-making in data centre operations is key, and this is where modern Data Centre Infrastructure Management (DCIM) tools play a transformative role.
Beyond real-time monitoring of power, cooling and asset performance, modern DCIM platforms now help operators analyse historical power contracts versus actual usage, revealing stranded capacity that can be repurposed for new workloads.
This is particularly relevant as electrical contract capacity utilisation is increasingly shaping upstream decisions in the energy and infrastructure supply chain. Just recently, Google signed a $3 billion hydropower deal in the US, the largest clean energy agreement of its kind, illustrating how major players are moving to secure power for future AI demand.
READ MORE: Sharepoint under attack: Microsoft rocked by "high-severity, high-urgency" Toolshell zero-day
Coupled with predictive analytics for demand forecasting and energy efficiency, DCIM delivers the detailed, auditable insights needed to modernise older facilities and meet growing demand responsibly.
For instance, in the UK, legislation such as the Climate Change Act and the Streamlined Energy and Carbon Reporting (SECR) mandate comprehensive reporting on energy use, carbon emissions and sustainability indicators.
Features like Carbon Footprint Reporting make it easier to track emissions live and take immediate, data-driven decisions like quickly addressing inefficiencies or overheating issues, to stay compliant.
This constant access to in-depth information empowers operators to proactively manage resources, replace underutilised legacy hardware with modern systems, and transform older sites into efficient, AI-ready hubs.
Supporting AI without compromising sustainability
AI workloads demand far more power than traditional computing. Without efficiency improvements, data centres’ share of global electricity usage could rise from 1.5% today to as much as 8% by 2030 - a direct threat to global climate goals.
By retrofitting instead of rebuilding, and leveraging modern DCIM systems, operators can responsibly support AI growth. This includes modernising underused facilities, balancing workloads across geographic regions, and making smarter use of stranded power contracts, all while reducing emissions and operational costs.
A global call for smarter infrastructure
The data centre industry stands at a critical moment. The decisions we make today will shape both the pace of AI innovation and the sustainability of the infrastructure powering it. It’s time to move past the default thinking of building when you see growth, and instead start building smarter, reusing what is already available.
Retrofitting legacy facilities with the right tools isn’t just an operational win; it’s a climate imperative. If we are serious about sustainable digital transformation, optimising what we already have must come first.
Rami Jebara is Chief Technology Officer of Hyperview
Have you got a story or insights to share? Get in touch and let us know.
