Asia faces 100GW energy surge as AI and data centres dominate power demand

Data centre electricity use will quadruple to 832 TWh by 2030, Morgan Stanley says.

Asia’s demand for data centre compute power is projected to exceed 100 GW by 2030, with AI-specific capacity estimated at 45 GW, according to the Morgan Stanley report “Powering AI: The 100GW Opportunity.”

Power consumption from data centres is expected to account for 3% of total regional power demand and 15% of new demand creation, with an annual growth rate of 23% through 2030.

Total regional data centre capacity is projected at approximately 98,757 MW. China is expected to host around 71,500 MW, followed by ASEAN (14,000–15,000 MW), India and South Korea (11,800 MW), Japan (5,900 MW), and Australia (4,000 MW).

Excluding China, Asia’s data centre capacity is expected to grow nearly fivefold by 2030.

Power demand for AI data centres in Asia is forecast at 395 TWh, whilst non-AI data centres are expected to consume approximately 429 TWh in 2030.

Total annual electricity consumption across all data centres in the region is projected at 832 TWh under the base-case scenario, compared with 193 TWh in 2023.

The projected power usage effectiveness (PUE) is expected to improve from 1.52 in 2023 to 1.31 by 2030, whilst total regional power draw is projected at approximately 104 GW.

Investment in the sector is concentrated across multiple countries. In ASEAN, AWS plans $6b in Malaysia by 2037, Microsoft $2.2b from 2024 to 2028, Google $2.4b by 2035, Nvidia-YTL Power $4.3b for AI cloud and supercomputing, GDS Holdings $2b, and ST Telemedia $1b.

In Thailand, Microsoft and Google have committed a combined $3.5 b, AWS $5b, and TikTok $8.8b for datacenter and cloud infrastructure.

In Singapore, AWS plans $8.6b by 2028, Google $5b for a fourth datacenter, and the government has committed $700m for AI development.

Greater China investments include Alibaba $52b for AI infrastructure, ByteDance $22b for cloud capex, and Tencent $13b for cloud projects. In India, TCS is committing $7b for 1 GW of datacenter capacity, AWS $4b, and Microsoft $3.5b. 

Total tracked investment in Asia is approximately $211.2b.

Financial modelling of a 100 MW AI hyperscaler shows total capital expenditure of $31.9m per MW, with GPU costs of $20.4m per MW and powered shell costs of $11.5m per MW.

Annualised operating projections indicate revenue of $1.067b, cost of power $107m, EBITDA of $952m, net profit of $378m, and a return on equity of 23.7%.

Sensitivity analysis indicates that a 17% increase in power costs reduces ROE by approximately 100 basis points.

Scenario analysis shows a base-case capacity of 100 GW in 2030, a bear case of 49 GW, and a bull case of 130 GW, reflecting variations in AI adoption, edge computing, and regulatory factors including data privacy and offshore training.

Key publicly listed companies positioned to benefit from Asia’s datacenter growth include Tenaga Nasional, Korea Electric Power, Hokkaido Electric, Sembcorp Industries, Keppel Ltd, GDS Holdings, NEXTDC, Delta Electronics, CATL, and HD Hyundai Electric.