AI, data centres, and the coming energy supercycle in Hong Kong

The question is not how much electricity is available, but where it comes from and how reliably it can be delivered. 

Artificial intelligence (AI) may be experienced through the cloud, but the infrastructure behind it is decidedly physical. Servers need electricity. Higher computing densities need more sophisticated cooling. Facilities need dependable grid connections, redundancy, and credible access to lower-carbon power.

For Hong Kong, this makes the AI boom not only a technology story, but an energy and infrastructure question.

The scale of the shift is becoming clear. In its energy and AI base case, the International Energy Agency (IEA) projects global data-centre electricity consumption to more than double to 945TWh by 2030, with AI the most important driver of growth.

Hong Kong is already seeing pressure on the demand side. CBRE estimates active data-centre capacity at 687MW in the first quarter of 2026, with a further 653MW of upcoming capacity. Many existing facilities were designed for rack densities of 5 to 15kW, whilst AI workloads can require 40kW or more.

That difference changes electrical design, cooling requirements, connection needs, and asset economics.

Power is becoming part of the investment case
Hong Kong has strong foundations as a digital hub: international connectivity, deep financial markets, proximity to Mainland China and a growing policy focus on AI. Its overall computing capacity has reached 5,000 floating-point operations per second (PFLOPS), including 3,000PFLOPS at Cyberport’s AI Supercomputing Centre.

The strategic question is whether the energy system can develop at the same pace. Data centres can be developed faster than major generation, transmission, and grid infrastructure.

For investors and developers, power availability can therefore no longer be treated as a utility connection to be resolved late in development. It should be tested early, alongside land, fibre, permitting, and cooling. Grid capacity, connection schedule, redundancy, future load growth, energy procurement increasingly influence bankability, and long-term competitiveness.

Hong Kong needs a system solution
Hong Kong cannot respond to rising digital demand in the same way as markets with abundant land for large domestic renewable projects. Its geography and density make the energy challenge different.

In 2025, nuclear energy and renewable energy together accounted for 28% of Hong Kong’s sent-out electricity mix, largely reflecting imported nuclear power. The government aims to raise zero-carbon energy to 60 to 70% of the generation mix by 2035, including 7.5 to 10% from local renewable energy. The transition will therefore depend on both local renewable development and stronger access to regional zero-carbon electricity.

Land constraints make space-efficient renewable solutions relevant. Reservoir-based floating solar can add local generation without competing for the same land as housing, industry or other infrastructure. Hong Kong has piloted floating photovoltaics (PV) on several reservoirs and examined larger-scale deployment. But floating solar should be viewed as one contributor to a broader system, not as a substitute for the generation, grid capacity, and regional interconnection required by round-the-clock digital loads.

Battery energy storage can complement renewable generation by shifting output, managing peaks and supporting grid flexibility. But storage does not create primary energy and cannot replace adequate generation or network capacity. The opportunity is therefore not a single technology. It is coordinated development across renewable generation, storage, substations, networks, and regional power infrastructure.

Regional integration will become increasingly important
Hong Kong’s future power strategy is becoming more closely connected with the Greater Bay Area. Sandy Ridge sits within the Northern Metropolis and is intended to support a wider cross-boundary digital ecosystem. At the same time, Hong Kong is strengthening its ability to receive zero-carbon electricity from Mainland China.

AI infrastructure may be located in Hong Kong, but part of its lower-carbon energy solution will necessarily be regional. The question is not only how much electricity is available, but where it comes from, how reliably it can be delivered, and whether the surrounding network can support concentrated new demand.

Engineering discipline will shape long-term competitiveness
From September 2026, Hong Kong’s amended Buildings Energy Efficiency Ordinance brings data centres more fully within building-services efficiency requirements, alongside tighter energy-audit, and disclosure provisions. This reinforces a broader shift: energy performance is becoming an operating and asset-quality issue, not simply an ESG metric.

The stronger projects will establish the power case before facility design is frozen. That means testing renewable-resource and energy-yield assumptions, grid connection, power-system performance, storage requirements, redundancy, interconnection, and future expansion under realistic scenarios. Independent technical review matters because assumptions around supply, connection, and performance ultimately become investment assumptions.

This is where renewable-energy and power-system engineering becomes commercially relevant to the AI build-out. The task is not to design IT infrastructure, but to independently assess whether the generation, storage, grid, and interconnection strategy around it is technically credible, resilient, and bankable.

Hong Kong has the ingredients to remain an important Asian AI and data-centre hub. But computing capacity alone will not determine the outcome. The coming energy supercycle will be shaped by whether digital ambition can be matched with reliable power, lower-carbon supply and infrastructure engineered for long-term performance.