‘Electricity can be a bottleneck for Vietnam’s ambition to develop AI’

At the Vietnam – Asia Digital Transformation Forum (DX Summit 2026) on May 27, experts said that power supply could become a major bottleneck in Vietnam’s ambition to develop AI and data centers (data centers – DC).

According to Mr. Phung Viet Thang, Country Director of Intel Vietnam, when it comes to AI, many people are often interested in computing capacity or supercomputer systems, but do not pay enough attention to the energy problem behind it. “It is not possible to build power plants as fast as building AI factories,” he said.

This reality is not only in Vietnam, but also on the world level. He said that with large computing centers, many businesses have to urgently build solar power systems to supplement energy sources for AI Data Centers, because the speed of expanding AI infrastructure is currently significantly faster than developing new power sources. Therefore, energy becomes an important issue that needs long-term attention.

This observation was agreed by many experts at the forum, in the context that Vietnam is promoting the development of AI infrastructure and data centers.

According to a May 2026 report from the Ministry of Science and Technology, the whole country currently has 42 data centers with a total design capacity of about 372.75 MW, an increase of more than 4 times compared to 2021.

Documents of the Fifth Session of the Government Steering Committee on the Development of Science, Technology, Innovation, and Digital Transformation also point out that Vietnam has formed a system of data centers and AI supercomputer clusters, but there are still many limitations such as small scale, small number of high-performance GPUs, unevenly distributed infrastructure, and lack of stable high-capacity power sources.

According to Mr. Nguyen Tuan Phong, VDI Deputy General Director of Technology, the world and Vietnam are witnessing an “explosive level of investment” in AI. However, AI Data Center is completely different from traditional data centers, especially in power consumption. He compared a regular server to “a regular car”, and an AI supercomputer to a “high-speed racing car”.

Mr. Phong said that the electricity consumption of high-performance AI systems can be 10-15 times greater than that of conventional servers. Therefore, when building an AI Data Center, right from the design stage, we must carefully consider the ability to meet the power source and cooling system.

Solve the energy problem

From an energy infrastructure perspective, Dr. Nguyen Manh Cuong, representative of the Power System Development Department, Institute of Energy, Ministry of Industry and Trade, said that electricity demand for AI and data centers in Vietnam will increase sharply in the near future. “The proportion of power consumed by data centers accounts for at least about 0.5% of total national electricity consumption,” he said. Citing an example in the US, he said this ratio is currently about 2-6% of the total national electricity. This shows that the room for data center development in Vietnam is still large, which means pressure on power infrastructure will also increase.

According to Dr. Cuong, the total capacity of Vietnam’s power system in 2025 will reach about 92 GW, but the base power source is only about 68 GW, equivalent to 74%. The rest is renewable energy such as wind power and solar power, with about 24 GW, accounting for 26%. He warned that if investment in power sources, especially base electricity, is not improved, the risk of shortage may occur in 2027.

This expert also commented that the risk of electricity shortage is more evident in the Northern region. According to him, Vietnam’s data centers in the future should prioritize being located in the Central or Southern regions, where the potential for renewable energy is higher.

The electricity problem for data centers lies not only in the generation source, but also in the transmission capacity. According to Dr. Cuong, currently Vietnam’s power lines and transformer stations carry a high load, about 70-75%. Meanwhile, in developed countries, for example in Germany, the line load carrying ratio is just under 50%.

According to Dr. Cuong, businesses investing in data centers in Vietnam have three options for power sources. The first is to buy electricity directly from the electricity company with large capacity requirements and multiple power sources to ensure safety. However, peak hour electricity prices are “not cheap”, sometimes up to about 5,000 VND per kWh.

The second option is to build your own power plant such as solar power, wind power or hydropower to supply the data center, then buy more electricity from EVN if there is a shortage.

The third option is to buy electricity from energy projects elsewhere and use the national grid to transmit it to the data center. With this option, businesses will pay additional electricity system service costs to the state. “In terms of institutions, there is no shortage, mainly the economic problem of businesses,” he said.

According to experts, as AI Data Centers consume more and more electricity, Vietnam’s problem is no longer just investing in GPUs or supercomputers, but also the ability to ensure a stable power source for AI infrastructure in the long term.