Summary
- AI demand, grid delays and space constraints are increasing the pressure to make better use of existing data centre infrastructure.
- Component-level cooling retrofits may reduce electricity demand and extend equipment life without complete system replacement.
- Procurement decisions based on upfront cost alone can miss maintenance, efficiency and lifecycle benefits.
Right now, we’re in a time of unprecedented momentum in the data centre industry. With the boom in artificial intelligence, AI workloads are projected to account for 70% of new capacity requirements and demand is set to triple by 2030, according to UK government data.
Building new facilities to accommodate this increase is slow, costly and restricted by space constraints, grid queue delays and lengthy substation upgrade wait times.
We’ve got investors in our industry with substantial funds but operators have to tell them to return in three years due to limits on power and space. Keeping growth going in Europe requires us to optimise our existing network through strategic upgrades.
If we’re realistic about future-proofing our data centres, retrofitting is the fastest route to scale. But what value does retrofitting add and why do we need it? What does it add to the equation in this new AI era?
The cost of throwaway culture
For years, the data centre industry has fallen into a bad habit: replacing entire data centre systems whenever a single internal part fails or becomes outdated.
We don’t expect our mission-critical infrastructure to act like disposable consumer electronics. Major cooling equipment, for instance, is built to last 20 years and contains up to 90% reusable materials such as copper, aluminium and casing. The units do not need to be fully replaced when one component malfunctions; for instance, a chilled water unit lasts two decades but an internal fan carries a limited service life of 100,000 hours.
True sustainability means moving away from this throwaway mindset, but progress is often blocked by suppliers who shift to non-repairable designs that force complete unit replacements over simple part repairs. It’s also true that a unit is often 10–15% more expensive to manufacture using sustainable materials and I’ve seen first-hand that customers are often unwilling to pay the difference.
Some operators are trying to get around this by simply turning up the thermostat in their data centres, but this leads to their workers having to complete tasks in dangerous 38.5°C conditions.
Achieving full circularity requires us to push suppliers for component-level maintainability. Without this shift, the hidden costs of ignoring retrofits quickly mount — leading to ballooning budgets, unnecessary downtime and significant carbon emissions from scrapping hardware prematurely.
The fastest route to scale
Targeted upgrades of existing equipment through retrofitting are the best route to scale. They prevent having to build fresh data centres or tear out infrastructure. Replacing legacy internal fans with high-efficiency alternatives will drastically cut power draw while extending existing facility lifespans by up to 10 years.
Even if this is a clear solution to the space constraints we’re facing, purchasing managers consistently prioritise lower initial purchase costs over long-term energy savings. There’s also slight tunnel vision across teams in the industry as cost can overshadow operational value.
Once, I had a project where I suggested we use three fans instead of two. We explained that energy savings would lead to a break-even point in 2.9 years. In the end the purchasing manager went for the two-fan unit which leads to lower upfront costs but it’s not the most sustainable option.
There’s a lack of communication between purchasing departments, facility managers and end-users which serves as an industry barrier. It’s understandable why people opt for cheaper alternatives but in an industry where something can change overnight it’s important to forward plan and ensure teams have technology that can withstand these changes.
It reminds me of an orange analogy I’ve shared over my thirty years in this industry. Three buyers end up bidding for a single orange because nobody communicated that one wanted the peel, one the juice, and one the pulp. When teams operate in silos, procurement teams buy for the initial price tag rather than the combined operational outcomes everyone actually needs.
The future
Right now, we’re seeing the perfect storm in our industry: growing AI demand, strict new regulations and long power delays are making building new data centres simply impractical; problems which are leading to some poor decision making.
Yet those quick fixes — like raising room temperatures to dangerous levels — aren’t the answer and pose some serious health and safety risks.
Retrofitting existing equipment is the fastest, safest and most sustainable way to unlock long-term capacity. Before committing to costly new builds or total replacements, operators must audit their current assets for targeted upgrades. We need to remember that we don’t need to rebuild the future of the industry from scratch. We just need to stop throwing away the one we’ve already built.
About the author
Ad Vink is the enterprise business leader – EMEA at Apx Data Centre Solutions. He is an expert in close control cooling systems and providing cooling solutions for data centres, computer rooms, telecom and industrial applications.
