Investment in data centres, particularly artificial intelligence (AI) data centres, remains economically viable and attractive, but it must be evaluated in the context of three fundamental constraints, says Gartner Inc, a business and technology insights company.
The first constraint is the availability and cost of key components, especially AI-optimised servers such as graphics processing units, said Linglan Wang, director analyst at Gartner.
These servers are the core engines of AI data centres and represent the largest share of capital expenditure. Their supply is tight, which can significantly impact both timelines and costs, she said.
The second constraint is the cost and complexity of building data centre infrastructure, which includes land acquisition, construction, power connectivity and supporting systems. As data centres scale up to support AI workloads, they become more capital-intensive and technically complex to deploy.
The third and most critical long-term constraint is the cost of operating these facilities, particularly electricity. Power availability and pricing are often the determining factors in whether a data centre investment is financially sustainable, making reliable and scalable power infrastructure essential, said Ms Wang.
Water usage and cooling costs, while important from an operational and environmental perspective, represent a relatively small portion of total costs, she noted, with compute-related costs much greater. While concerns about water and cooling should be managed responsibly, they are not the primary economic barrier to data centre investment, according to Gartner.
"Promoting data centre investment in Thailand, or in any other country, can still generate strong economic returns if the country can provide competitive advantages in power, infrastructure and cost efficiency," said Ms Wang.
To ensure data centre investments deliver maximum economic benefit, a strategic and disciplined approach is required, with success dependent on securing reliable, affordable and scalable power, she noted.
Governments and investors should prioritise strengthening electricity infrastructure, expanding clean energy sources and ensuring long-term price competitiveness, said Ms Wang.
"Without this foundation, large-scale data centre development will be difficult to sustain," she said.
In addition, investment decisions must be grounded in real and sustainable demand rather than short-term hype, noted Ms Wang. While AI is driving significant growth, not all announced projects will materialise. Aligning investments with actual usage and economic needs will help avoid overcapacity and ensure healthy returns, she said.
Countries should focus on capturing value beyond the physical data centre, fostering a structure around high-value compute, said Ms Wang. The greatest economic returns come from AI workloads, cloud services and digital innovation built on top of the infrastructure, rather than from construction alone, she noted.
Finally, a data-driven and location-specific strategy is essential. Data centres should be developed in regions with clear advantages, such as strong energy supply, favourable regulations and good connectivity.
Careful planning and use of granular data will ensure that investments are both efficient and sustainable over the long term, Ms Wang said.
