OpenAI and Nvidia’s giant AI project will take a lot of foreign-made parts

Supply chain data into the OpenAI and Nvidia $100 billion investment commitment this week shines a glaring light on the United States’ heavy reliance on foreign suppliers for components to complete the projects.

Brandon Daniels, CEO of AI-powered risk and supply chain management solutions company Exiger, told CNBC that building traditional power plants, whether gas-fired or nuclear, requires an array of specialized parts that the U.S. doesn’t produce in large numbers.  

“There are four major categories of equipment that are both extremely expensive and largely foreign-sourced, and this dependency comes at a time when power construction is already grappling with severe backlogs and shortages,” Daniels warned.

The first is gas turbines.

Exiger supply chain data analysis shows the global market for heavy-duty, utility-scale turbines is dominated by three Original Equipment Manufacturers (OEMs): GE Vernova of the U.S., Siemens in Germany and Japan’s Mitsubishi.

The three control new turbine supply, and nearly 50% of that supply is foreign-sourced, Daniels said. “These companies will be indispensable for powering this scale of [artificial intelligence] infrastructure,” he said.

Nuclear components 

The second category is ultra-large nuclear plant forgings and components.

Daniels said the massive, one-piece reactor pressure vessels and associated hardware are no longer manufactured in the U.S. For example, South Korea-based Doosan fabricated the pressure vessels and steam generators for the recently-built reactors at the Vogtle plant in Georgia, the first new nuclear plant to open in the U.S. in a generation.

Large transformers essential for distributing power into the electrical grid from generating plants are a third part of the supply chain relies on overseas suppliers.

More than 80% of such high-voltage transformers are made by suppliers in countries ranging from South Korea to Germany to Canada, Daniels said.

Steel is a fourth element in this supply chain that’s likely to add to costs.

“While the U.S. and allies like the U.K. remain significant producers, project developers frequently rely on imports to meet both cost and capacity needs,” Daniels said.

Tariff effect

Higher tariffs on imported goods only add to the complexity of these projects.

“The tariff impact isn’t uniform, because the scale of the buildout and planning will take years, not months, and U.S. domestic capacity may expand during that time,” said Daniels. “Some allied suppliers are also under lower tariff regimes, with the U.K., for example, currently exempt from the 50% U.S. steel tariff. But even with these carve-outs, the scale of material and equipment required means costs will rise.”

On multi-billion-dollar energy projects, a 3%-6% budget increase due to tariffs translates into hundreds of millions of dollars, Daniels said.

“This is just in steel and aluminum alone,” said Daniels. “Additional pressure comes from foreign-sourced components like turbines, reactor vessels and transformers, where backlogs and import reliance amplify the exposure. Together, tariffs and sourcing realities make this AI infrastructure build-out not just an engineering challenge but also a significant supply chain and trade policy challenge.” 

Add to this the amount of labor support needed to bring plans to completion.

“This requires an almost wartime-like labor expansion of the industrial base,” said Daniels. “We simply don’t have the skilled workforce in place to scale this fast. U.S. [building] trades are already in decline, with a shortage of welders, machinists and electricians. This could be just as big a pinch point as the hardware itself. That’s the reality we’ll have to confront over the next five years to make this even remotely possible.”