Foreign media believe that the AI boom is often interpreted as a race in GPU procurement and capital spending by cloud providers, but the more difficult segment to catch up on is now shifting to electricity. While data center construction cycles are relatively short, transmission lines, substations, and new power generation capacity often progress much more slowly, making power supply a practical constraint on AI expansion.
Data center electricity consumption continues to rise
The International Energy Agency predicts that global data center electricity consumption will rise from approximately 485 TWh in 2025 to about 950 TWh in 2030, with AI-focused facilities growing even faster. The article highlights that the issue is not just how much electricity AI consumes, but that demand is emerging too rapidly for grids and supporting infrastructure to keep pace.
In addition to the GPU, there are supporting systems.
The article states that AI clusters require more than just processors. Large-scale accelerator deployments also depend on networking equipment, power conversion, cooling systems, grid connectivity, and stable power supply. As server density continues to increase, the infrastructure requirements to support these devices are rising accordingly.
- GPU and network interconnection
- Power Conversion and Thermal Management
- Grid connection and power generation capacity
The article argues that some of these bottlenecks cannot be quickly resolved by placing additional orders, as with purchasing chips. Large cloud providers can continue buying more GPUs, but they cannot immediately gain new high-voltage connectivity capacity.
Texas grid connection applications exceed 700 GW
Texas is considered a typical case. The article cites data showing that power connection requests for proposed data centers in the region have exceeded 700 GW, far surpassing the current actual power consumption of data centers in the United States. Due to the possibility that some projects may never materialize, regulators have begun tightening connection rules to prevent grid resources from being tied up by "phantom demand."
This also creates direct financial challenges. Utility companies may invest in infrastructure for projects that have not yet materialized, while actual AI parks may face years of delays and struggle to gain timely access to electricity.
Mining companies and power companies benefit.
The article argues that this bottleneck is reshaping the beneficiaries across the AI industry chain. Recently, Vertiv, a supplier of data center power management and cooling equipment, agreed to acquire the microgrid company Utility Innovation Group in a transaction worth up to $2.6 billion, signaling rising value in on-site power generation and infrastructure independent of the main grid.
The importance of microgrids lies in their ability to integrate grid power, on-site generation, and energy storage, reducing reliance on utility connection timelines. Meanwhile, power providers such as NextEra Energy and Dominion Energy are also benefiting from long-term demand generated by cloud providers and data center developers.
The article also notes that AI expansion is changing the economic logic of Bitcoin mining. Mining companies typically control access to electricity in regions with low power rates, and such connections themselves are already scarce. If AI companies are willing to pay higher prices for the same electricity, some mining companies may shift part of their capacity toward high-performance computing services.