As the AI industry continues to surge, so does its demand for energy. By 2027, data centers for AI technology will use as much electricity as the Netherlands. However, a crucial yet practical problem threatens to hinder the industry’s scalability: the shortage of electrical transformers.
The Transformer Shortage
AI data centers use refrigerator-sized transformers to change high-voltage electricity to lower voltage. These transformers connect to the power grid.
There are very few transformers available right now. It takes up to four years to get one, and prices have increased by 70% since January 2020. The supply chain pressures causing this shortage show no signs of easing.
“Unless we see considerable investment at both the commodity level and in the manufacturing of transformers, the shortage we’re currently facing is only going to worsen,” says Benjamin Boucher, an energy analyst at Wood Mackenzie. This bottleneck is a significant challenge for scaling AI capacity.
Importance of Transformers
Transformers are vital for transmitting power over long distances. They “step up” voltage for transmission and “step down” to usable levels for homes, offices, and data centers. These large, costly units are essential for safety and reliability. Using too much voltage can be dangerous or harm electrical parts.
Since 2020, demand for both types of transformers has surged. AI startups and data centers use step-down transformers, while solar and wind farms use step-up transformers to send their power.
“We’ve seen a huge boom in demand for transformers recently,” Boucher notes. “The supply just hasn’t been able to cope.”
Historical Context
The transformer supply issue dates back to the 1980s when manufacturers began consolidating. Higher costs for materials and labor have slowed production, with copper and electrical steel prices doubling since 2020.
Manufacturers are cautious about increasing production quickly. They want to avoid making the same mistakes as in the past, such as the housing boom before 2008 and the recent semiconductor supply chain issues.
“We’ve learned from the semiconductor supply-chain issues that there will be no rush to make up a shortage if it results in oversupply,” says Edward Wilford, a semiconductor analyst at Omdia.
Potential Solutions
To address this bottleneck, innovative solutions are being explored. AI itself could help balance electrical loads more efficiently. Some analysts suggest constructing AI data centers in deserts with solar panels or using small nuclear reactors for power.
Wilford demonstrates how TSMC, a Taiwanese chip company, rented a wind farm in 2020. This allowed them to generate their own electricity and avoid relying on the grid. “People will keep finding solutions,” Wilford says. “I hope they’re good ones.”
Finding reliable and scalable power sources will be crucial as the AI industry expands. The lack of transformers is a big problem, but new ideas and careful planning can help create a strong and lasting future for AI.
- If you want to learn more about the topics discussed in this blog, click on the following links. Wood Mackenzie provides insights into energy markets and discusses the challenges and solutions related to the transformer shortage.
- Copper Prices gives updates on recent trends in copper prices. These trends have greatly affected the cost and availability of transformers.
- Omdia covers the semiconductor industry and its supply chain issues, drawing parallels to the current transformer shortage.
- TSMC has rented an offshore wind farm to produce its own electricity. This helps them deal with energy supply issues in a unique manner.
These resources provide additional information on the discussed issues. They help in understanding the bigger picture and potential solutions for the transformer shortage in the AI industry.
