AI demand reshaping data centres with new challenges ahead

Niklas Lindqvist, Nordic General Manager at Onnec, has shared insights into the challenges and opportunities data centre operators may encounter as 2025 approaches.

Operators are currently struggling to secure AI-ready infrastructure components, leading them to use outdated equipment. "Long wait times for essential equipment like batteries, generators and GPUs are forcing operators to settle for what they can get their hands on, not what they really need," Lindqvist stated.

The surge in AI demand is driving competition for top-tier technology among data centre projects and major infrastructure projects. "These delays are pushing lead times for critical parts to over two years, making it nearly impossible for data centres to stay on track. To avoid delays, many are forced to settle for 'second best' or older hardware just to get things up and running. The problem? This short-term fix is setting them up for major headaches down the road," Lindqvist explained.

He added, "AI-driven workloads require top-tier performance and using yesterday's technology means today's data centres will start to struggle to keep up with increasing processing demands. To make matters worse, with technology evolving so fast, parts ordered years in advance – like GPUs or batteries – are often outdated by the time they're installed. This means operators will face costly hardware swaps soon after data centres go live."

Lindqvist anticipates that by the end of 2025, there will be visible issues in some newly built facilities, stating, "Retrofitting hardware that's barely 18 months old will become the norm. That means operators need to shift their focus toward optimising and squeezing every bit of value from their equipment. This includes investing in quality cabling from the start to avoid costly replacements down the line. Otherwise, they'll be stuck playing catch-up, facing rising costs and disruptive impacts to their operations."

The demand for data centres is also accelerating construction timelines, which may pose health and safety risks. "As AI investment drives accelerated data centre construction, data centre operators will be pushed for faster timelines when delivering new projects. We will see an increased push for 24/7 construction to get services up and running quicker, with less acceptance of delays – meaning construction teams will need to deliver bigger projects in shorter timespans," Lindqvist commented.

He warned of potential safety incidents, highlighting a shortage of specialised workers, "To negate this risk, operators and construction teams must ensure they follow best practice and don't take unnecessary risks. Moreover, the industry must address the labour skills gap with enhanced training programmes, apprenticeships and collaboration between educational institutions, industry and the government. These efforts are long-term solutions, but the industry faces rising project risks now. In the coming year, operators must tackle these risks head-on, set realistic construction lead times – and resist the pressure to cut corners on safety for speed."

Lindqvist predicted significant growth in emerging markets in response to constraints in established FLAP-D markets. "Continued AI adoption will drive further demand for data centre infrastructure. But new regions beyond the traditional FLAP-D markets (Frankfurt, London, Amsterdam, Paris and Dublin) will experience the biggest growth as some FLAP-D regions may run up against power and land constraints," he said.

He continued, "Key markets are facing power shortages and stricter regulations, causing the Netherlands, Ireland and Germany to pause new projects. With AI prompting huge demand for data centre infrastructure and traditional markets starting to stall, it provides new regions of data centre power an opportunity to step up."

These emerging regions are capitalising on the opportunity, with Lindqvist noting, "These emerging regions are creating more favourable regulatory environments to attract investment and foster AI and data centre growth. Countries in the Nordic regions have developed AI strategies to bolster growth, Spain has been making major inroads to attract investment, and Italy's government has declared data centres as 'critical infrastructure'. These regions understand data centres are now strategic assets with significant economic value. Data centres will become the new oil for emerging regions poised to pounce and ride the AI wave to supercharge their economies."

On the energy front, AI growth is expected to boost demand for renewable energy sources over nuclear power solutions. Lindqvist explained, "AI's rapid growth is driving data centres' energy demands beyond current grid capacities. To sustainably meet these demands, the energy grid must diversify. While hyperscalers are eyeing up nuclear as a carbon-free solution, it's not the 'quick fix' it appears to be – whereas renewables will be."

Although companies like Google and Microsoft have plans to purchase nuclear power, Lindqvist noted the challenges involved in scaling nuclear energy. "Despite Google and Microsoft's plans to purchase nuclear power, with Google targeting 2030 and Microsoft 2028, nuclear power remains years away from impactful deployment. The challenges are significant: limited sites, long construction timelines and complex regulatory barriers mean nuclear will struggle to scale in time. And in the meantime, AI demand is expected to surge."

He stated, "Instead, we'll see a stronger shift toward renewable energy sources like wind, solar and hydropower, which are more accessible and adaptable to rapid industry growth. These sources not only support long-term sustainability goals but also reduce carbon footprints. Data centres too, are stepping up in this transformation. They're becoming 'smart prosumers' – using electricity when it's available and feeding power back into the grid when it's needed elsewhere."

Data centres are already contributing to this energy transformation, as Lindqvist illustrated. "We're already seeing data centres repurpose waste heat, like in the Paris 2024 Olympic pools and the UK's heat network zones. Next year, they'll play an even bigger role in stabilising energy grids that are evolving to meet surging demand. With small modular energy sources and new technology, data centres will help turn traditional grids into a dynamic, interconnected tapestry of energy solutions."

Lindqvist also noted that operators need to be cautious about partnerships. "One of the big challenges facing DC operators in the year ahead will be managing risk around new customers. Traditionally, operators have worked with established companies who are more stable and predictable. Now, the boom in the AI market has led to floods of VC investment for AI companies, who are investing heavily in DCs to power their thirsty products. This is great news for the DC industry, but means the market is now a lot more volatile. Investment funding can run out or go on pause, delaying projects from moving on. As such, DC operators will need to think carefully about their due diligence to ensure they don't get stuck with the bill if customers run out of cash."