The AI boom requires a lot of energy. Early predictions indicate that data centers behind AI technology will consume as much electricity as the entire Netherlands by 2027.
But for all the futuristic problems that AI promises to solve, there are clearly practical problems that prevent it from scaling up en masse. AI data centers rely on specialized transformers—refrigerator-sized units that convert electrical current to safe voltages—to integrate with power grids, networks of power plants, and wires that carry electricity across the country. However, these transformers are currently in such short supply that wait times for new units can be as long as four years. Due to demand, prices have also increased by about 70% since January 2020. Supply chain pressures and historical trends that have led to shortages show no signs of easing.
Benjamin Boucher, an energy analyst at consultancy Wood Mackenzie, said: “Unless we see significant investment, both at the commodity level and in transformer manufacturing, the shortages we currently face will continue to worsen. I think I’ll do it,” he says. luck. “It’s going to take a few years to actually get this in bulk.” [AI] online capacity. ”
Years of underinvestment and consolidation in this niche manufacturing sector have created transformer bottlenecks that are colliding with the surge in demand from the AI and renewable energy sectors. Analysts’ only hope is that innovation, or perhaps engineering assistance from AI itself, will create another way to power data centers without waiting years for new transformer capacity.
“AI provides what I describe as ‘self-growth bootstrapping.’ In other words, AI can help you maintain balance. [electrical] ” said Edward Wilford, a semiconductor analyst at consulting firm Omdia. luck. “On short notice, there’s always going to be some sort of shortage. So you just roll the dice and think that the industry you’ve built is enough to survive.”
This obviously could have a huge impact on the stock market, especially for the nearly $2 trillion NVIDIA, which is at an all-time high after the crash. But while many other analysts and market watchers are passionately debating whether the market is approaching bubble levels, this analyst said it was too early to tell.
What is a transformer and why is it important for AI development?
A transformer is an essential device in electrical projects connected to the power grid. To efficiently transmit power over hundreds of miles, Boucher says, transformers are needed to “step up” the voltage to a higher level on the transmission side. However, that high voltage current is unsafe for use in homes, offices, and data centers, so a “step-down” transformer is installed at the receiving end to bring the voltage back to a usable level. These transformers are bulky, expensive units that often look like large chest freezers. These are not replacements for safety reasons. Most insurance companies and utilities require a complete backup set for critical projects because using too high a voltage can be fatal and damage expensive electrical components.
Since 2020, the demand for both types of transformers has surged significantly. AI startups and data centers desperately need step-down transformers to connect to the grid and access power. Additionally, solar and wind farms require step-up transformers so that the electricity they generate can be transmitted to consumers.
“Recently, there has been a huge boom in demand for transformers,” says Boucher. “Right now, it’s a perfect storm for transformers in the sense that there’s so much demand and the supply just can’t keep up with it.”
Transformer supply problems go back a long way
The reasons for the slow supply of transformers go back much further than the pandemic. Consolidation of American transformer manufacturers began already in the 1980s, and after the boom years of his 60s and his 70s, when most of the power grid was built, the industry continued to shrink. These cuts mean that domestic manufacturers can now supply only about 20% of the U.S. transformer demand. (Imports make up the difference, with China and Mexico being her two biggest suppliers.)
Rising raw material and labor costs are also hurting production. Commodity prices for the copper and specialty electrical steel sheets needed to manufacture transformers have doubled since 2020 (AK Steel is the only company in the U.S. that makes electrical steel sheets).
Historical precedent has made transformer manufacturers reluctant to rapidly expand production. These companies were overactive during the pre-2008 housing boom to meet increased demand for electricity, but suffered significant losses when the market crashed. Transformer makers are also eyeing the semiconductor industry, which experienced its own bullish effects several years ago. By then, manufacturers were investing millions of dollars to ramp up production and demand, while buyers were desperate for more chips as supply chains slowed. It was withered. Transformer manufacturers need to be confident that the demand for AI and renewable energy is here to stay before investing in new production.
“One of the lessons the industry has learned from the semiconductor supply chain issues is that when there is an oversupply, there is no need to rush to make up for the shortage,” Wilford said. “No one wants to build infrastructure that will take him 10 years to pay off and then suddenly pay off. [worth nothing] It’s because I made too much. ”
Solutions that don’t exist yet
Recognizing that the surge in demand could overwhelm domestic transformer manufacturers, the Biden administration signed an executive order in June 2022 allocating aid to the industry through the Defense Production Act. But nothing has materialized yet, and the transformer industry has not received any funding. $1.7 trillion 2022 omnibus spending bill. Overseas suppliers are also struggling to keep up with supply.
“We need more investment, especially domestically, in terms of manufacturing and transformers,” Boucher said. “I think we need to thoroughly strengthen our supply chain to meet more than 20% of demand.”
AI is expanding rapidly, so if you want anything to improve, you need to act quickly. Global spending on data centers and other hardware behind AI development is expected to exceed $4 billion by 2030.
“When you factor in all the additional power that will be needed for AI and data centers, you can definitely see an even bigger gap in terms of transformer supply,” Bucher says. “That could definitely put the transformer industry in a bigger hole than we’re already in.”
But the transformer shortage will become a point of contention if the industry can find another way to power data centers. Analysts are hopeful that innovation will provide a solution, even though the exact problem of providing power in other ways has not yet been resolved.
“I’m not worried about disrupting the industry because I think we’ll find a solution,” Wilford said. “We will wait for the market to address this issue. The companies themselves will find a way. They are considering installing small nuclear reactors. [data centers]”
Alternatives include building an AI data center in the desert surrounded by solar panels, or using AI to more efficiently regulate power usage and integrating the data center into the existing power grid. include. Wilford pointed out that Taiwanese semiconductor giant TSMC leased an offshore wind farm in 2020 to solve its own electricity needs by generating its own electricity rather than sourcing it from the grid.
“People will continue to find solutions,” Wilford said. “I hope they’re good.”
