Blog

[SINGAPORE] UOB Kay Hian (UOBKH) predicts continued growth in demand for data centres over the next five years – despite a challenging environment resulting from tariff volatility.

It maintained its “overweight” call on data centre real estate investment trusts (Reits), citing projections of rising global demand for data centres fuelled by the onset of next-generation, reasoning artificial intelligence (AI) models, alongside a potential supply shortage of AI-ready data centres. 

Jonathan Koh, UOBKH analyst, said the growing demand for data centres is poised to stay intact and shielded from damaging tariff effects. 

“(Data centres are) sheltered from reciprocal tariffs. Demand for data centres is correlated to new product cycles, rather than reciprocal tariffs imposed by the Trump administration,” said Koh.

US President Donald Trump in early April announced reciprocal tariffs customised to countries running trade deficits with the US, setting off a global, deep market sell-off. He later backed down and announced a 90-day pause for tariffs on all countries except China. Since then, markets have been whip-sawing between losses and gains.

“The advent of new reasoning AI models, including OpenAI GPT-5, will drive the next wave of demand for data centres,” Koh said. 

Top picks for the sector included Digital Core Reit with a US$0.88 target price versus its Monday close of US$0.475, Keppel Data Centre Reit with a S$2.55 target price versus its S$1.93 close and Mapletree Industrial Trust with a S$2.70 target price versus its S$1.93 close. All were assigned a “buy” rating by UOBKH.

Next-generation reasoning AI models

Koh noted that global data centre capacity demand is set to expand by a compound annual growth rate (CAGR) of 22 per cent from 2023 to 2030 to reach an annual demand of 219 gigawatts, citing McKinsey data.

That growth is driven by demand for AI-ready data centres which provide high computational power – set to increase by a CAGR of 33 per cent. 

“Future demand (for) data centre capacity depends on the pace of adoption for advanced AI use cases,” said the analyst. 

He cited Nvidia chief executive Jensen Huang on data that showed reasoning AI models consume more than 100 times more resources compared to conventional non-reasoning AI models.

This comes as the predicted capital expenditure for computing infrastructure is set to reach US$1 trillion by the end of the decade with AI-ready data centres facing a potential supply deficit, UOBKH pointed out.

Singapore data centres stand to gain

Data centres in Singapore will benefit from increased demand generated by the rise of next-generation reasoning AI models, UOBKH said.

This comes as the Republic’s data centres are well-suited for “mission-critical and low-latency (time-sensitive) applications, such as financial services and AI inference applications”. 

Additionally, the city-state is a “preferred data centre hub” that is “well-positioned for AI inference”. 

Koh noted that Singapore is a connectivity hub linking South-east Asia to the global network through 26 international subsea cables and three landing sites. 

Moreover, the Republic is set to remain as the region’s pre-eminent hub as it is staying ahead of competition in terms of connectivity. 

It is targeting to invest minimally S$10 billion over the next decade to double its capacity for international subsea cables and landing sites to support “pervasive usage of new AI applications”, Koh said.

“The government is working with the private sector and research institutions to scale the usage of autonomous systems using new technology, such as low-earth orbit satellites,” he added.