The competition in artificial intelligence (AI) is shifting from semiconductors to power infrastructure. As AI data centers grow to gigawatt (GW) scale, the speed at which companies can secure transformers and switchgear is becoming a critical factor in business operations, surpassing the need for GPU acquisition.
According to the power equipment and semiconductor industries, major global tech companies are prioritizing the procurement of high-voltage transformers and power distribution equipment alongside GPUs and servers for their AI data center projects. Since AI data centers consume significantly more power than traditional data centers, delays in establishing power infrastructure could postpone the completion of these facilities.
Microsoft, Amazon Web Services (AWS), Meta, and Google are expanding their investments in AI data centers across the United States, focusing on securing long-term power purchase agreements (PPAs) and early delivery schedules for transformers. The 'Stargate' project, led by OpenAI and SoftBank, also emphasizes the importance of securing power infrastructure alongside data center construction.
Industry analysts suggest that the bottleneck in the AI data center supply chain is shifting from GPUs to power equipment. While GPU supply, primarily from Nvidia, is gradually increasing, the production capacity for transformers and high-voltage switchgear is difficult to ramp up quickly, often resulting in delivery times extending to several years.
For instance, reports indicate that delivery times for high-voltage transformers in the U.S. power industry have stretched to 2 to 4 years. The International Energy Agency (IEA) has warned that without concurrent investments in power infrastructure like transmission networks and transformers, the growth of the AI industry could be constrained due to the surge in data center power consumption.
The scale of AI data centers is also rapidly increasing. Nvidia projected at this year's GTC that future AI factories will exceed 1 GW in power usage. This marks a significant shift from the past, where data centers operated at several megawatts, now consuming power equivalent to that of an entire city.
This transformation presents new opportunities for domestic power equipment manufacturers. LS Electric is expanding its supply of high-voltage distribution equipment and power management solutions for data centers, while HD Hyundai Electric and Hyosung Heavy Industries are seeing a significant increase in transformer orders, particularly in North America. Analysts predict that if investments in AI data centers continue to grow, the surge in orders for power equipment manufacturers will also persist for an extended period.
The government's initiative to establish a national AI computing center and private AI data center projects also highlight the critical need for securing power infrastructure. Industry experts believe that the competition to attract AI data centers will ultimately hinge on how quickly transmission networks and substations can be developed.
As interest in the AI industry shifts from GPU performance to the construction of AI factories, the strategic value of power equipment is increasing. Even if semiconductors and servers are secured, delays in power supply could prevent data centers, which require investments in the trillions of won, from becoming operational.
An industry insider stated, "In the past, the key competitive factor was how many Nvidia GPUs could be secured, but now it is about how quickly transformers and power networks can be obtained that will determine the completion timeline of AI data centers. The core of the AI supply chain is expanding from semiconductors to power infrastructure."
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.