Data centre growth driven by demand for artificial intelligence (AI) is colliding with environmental, social, and governance (ESG) and infrastructure constraints, according to GlobalData.
Annual data centre power consumption is expected to rise from 590 terawatt-hours (TWh) in 2025 to 1 520TWh in 2030, accounting for 4,5% of global electricity consumption – up from 2,1% in 2025. However, rising water use in already water-stressed areas, growing public opposition, grid connection delays, and shortages of materials and skilled labour are delaying data centre projects and could slow AI progress.
GlobalData’s latest Strategic Intelligence report – The Constraints Reshaping Data Centre Development – reveals that most announced data centre projects remain stuck in the planning stage. Projects that broke ground between Q3 2025 and Q3 2026 waited an average of just over two years from announcement to the start of construction – up from about 18 months for those that started between Q1 2020 and Q2 2025.
“Data centres are increasingly located in water-stressed areas and their construction, along with the energy-intensive training and running of large language models (LLMs), are adding new sources of emissions,” says Martina Raveni, senior analyst, Strategic Intelligence team at GlobalData. “Together with concerns over noise, land use, and rising electricity bills, these impacts are hardening public opposition. According to Data Centre Watch, in Q1 2026 alone, at least 75 US data centre projects worth roughly $130-billion were blocked or delayed – the largest single-quarter total on record. Similar backlash is emerging in countries including the UK, Chile, South Korea, and Japan.”
Grid access has become a major bottleneck. According to the IEA’s 2026 Electricity Outlook, more than 2 500GW of renewable, large-load, and storage projects are waiting in connection queues worldwide – and developers in Europe’s main data centre hubs can face waits of seven to 10 years. Building new grid infrastructure can take five to 15 years, compared with the one to three years required to build a data centre. Grid congestion is pushing operators toward on-site gas power, which risks increasing emissions even as grids decarbonise.
“Data centre buildout also depends on fragile supply chains,” says Raveni. “A single hyperscale facility can require up to 50 000 tons of copper, while AI data centres need 10 to 20 times more fibre than traditional builds, at a time when fibre manufacturers are already at capacity. On top of this, a shortage of electricians and other skilled trades is delaying projects.”
The industry is responding on several fronts, although progress remains uneven. Operators are adopting closed-loop cooling, expanding water stewardship commitments, diversifying their power supplies through renewables, battery storage, and fuel cells, and extending equipment lifespans through circular economy practices. Community engagement and workforce development are increasingly treated as core requirements for project delivery.
“These constraints are already translating into real delays for AI infrastructure,” says Raveni. “As capacity tightens, compute becomes more expensive and harder to access, pushing AI developers toward smaller models and more limited use cases. Unless grid access, equipment supplies, and construction timelines become more predictable, AI progress will be shaped as much by power and hardware logistics as by algorithmic innovation.”