AI surge reshapes global data centre growth for 2026

AI computing demand, power constraints and regional policy shifts are set to reshape global data centre development in 2026, according to partners at engineering consultancy Cundall.

The firm sees diverging regional trends across Europe, the Nordics, the Middle East and Asia-Pacific. It also expects major changes in power infrastructure design and a sharper focus on skills in the UK market.

Barbara Sacha, Partner and Data Centre sector lead at Cundall, said power access, land availability and AI and cloud growth will determine where new projects move. She pointed to a cluster of European markets that are holding or gaining ground.

APAC growth corridors

Alex Saez, Partner, Data Centres at Cundall, views Asia-Pacific as the fastest-growing region for data centres in the period ahead. He said policy decisions, land supply and domestic digital growth are shifting the map of investment.

"The Asia-Pacific region is projected to be the fastest-growing market. Singapore remains a critical hub despite land constraints, with Malaysia emerging as a spillover destination offering lower costs and strong connectivity. India's rapid digitalisation and cloud adoption are driving hyperscale investments, while Australia continues to attract projects thanks to its stable power grid and proximity to Asia's major economies. Southeast Asian markets like Vietnam are beginning to gain traction as developers seek new growth corridors to meet rising local demand," said Saez.

Several governments across Southeast Asia have set targets for cloud infrastructure and digital trade. Developers are now assembling regional portfolios rather than focusing on single markets.

UK skills pressure

In the UK, Cundall expects the tight labour market in engineering and construction to remain a major constraint. It links this challenge to a rising domestic pipeline of facilities that support AI and cloud projects.

Andrew Livesey, Partner, Data Centres at Cundall, said skills remain a central concern for data centre build-out.

"In 2026, the skills shortage will be an ever-growing pressure for the UK data centre industry. As the UK seeks to become an AI superpower and investment in continues through AI growth zones and government-backed developments, the industry can't keep pace without expanding the talent pool. The most effective way to do this is to widen participation and open doors to those who may not traditionally see engineering as a possible career; we must look to increase the diversity of the industry. After years of UK engineers delivering major projects across the globe, 2026 will show that their expertise is needed at home more than ever to account for the growing pipeline of data centre projects here," said Livesey.

He expects more training routes in response.

"Apprenticeship opportunities will and should increase to meet this demand and to benefit the sector through a diverse work force. Whilst they demand more training time, they bring in talent that might otherwise not reach the industry. Organisations like The Land Collective are already helping to bridge this gap by connecting people with engineering opportunities, and I'm proud to be a champion sponsor of this at Cundall," said Livesey.

Major operators and contractors have already expanded apprenticeship intake in critical infrastructure and related trades. Industry bodies are also pressing for broader outreach to under-represented groups.

Power and AI loads

Cundall expects the next phase of AI build-out to change the design of electrical systems in large campuses. The firm cites rapidly rising rack densities and volatile load profiles from large GPU clusters.

Jamie Cameron, Partner, Data Centres at Cundall, said attention is moving from cooling to power engineering at rack and campus level.

"In 2026, the industry will shift from asking how we cool high-density AI racks to a tougher question: how we power them. Liquid cooling is now largely understood in design; however, the main disruptions will be on the electrical infrastructure. Large GPU clusters are creating spikes in power volatility when transitioning from periods of intense processing to data exchanging, this results in clusters jumping from 20-40% idle load to 140-160% in a few cycles before settling around 100% utilisation. Additionally, rack densities are pushing beyond 300kW with a roadmap towards over 2MW per rack," said Cameron.

He expects this shift to influence voltage levels in data halls and the use of emerging storage technologies in campuses.

"This will drive development to keep pace with power demand. There will be a move towards 800V DC distribution to power the high-density racks more efficiently. Next generation GPUs will be provided with supercapacitors to smooth out instantaneous peaks at rack level, while at the campus level medium voltage Battery Energy Storage Systems (BESS) will be used to soften the impact of multi-hundred-megawatt step loads on the grid. We won't see full-scale deployment in 2026, but we are likely to see proofs of concept and the standardisation of designs that will provide the foundation for the next generation of data centres," said Cameron.

Developers are now working with utilities and equipment vendors on pilot projects that use these approaches. Design standards for these schemes are likely to form the basis of later roll-outs across hyperscale campuses.