TOKYO -- Japanese trading house Marubeni is betting on physical bottlenecks such as demand for energy and natural resources stemming from the artificial intelligence boom to drive its growth, expanding its exposure to businesses that supply and manage the power, metals and communications infrastructure needed to support rising demand for computing power.

AI is "the biggest growth engine of this century," Masayuki Omoto, Marubeni's CEO, told Nikkei Asia in a recent interview. But as the technology develops, "The physical resources supporting it are becoming scarce," he said. "Data centers require considerable amounts of copper for equipment such as cables, transformers and electrical wiring. Above all, there is energy. AI is driving an overwhelming increase in electricity demand." Strains are already appearing. Global data-center electricity consumption rose 17% in 2025 versus the previous year and is set to double by 2030, according to the International Energy Agency.

Copper also faces a squeeze. Research company BloombergNEF expects a shortage to emerge in 2028, widening to 20 million metric tons by 2050, equivalent to 44% of forecast global demand, partly due to demand from data centers.

Although Marubeni, one of Japan's five large trading houses backed by prominent U.S. investor Berkshire Hathaway, does not plan to make data-center development a centerpiece of its strategy, the company is expanding in parts of the AI value chain where it already has exposure and sees potential for growth.

For example, in upstream businesses, it acquired substantial natural gas production capacity in the U.S. as demand for energy rises. Its U.K.-based power supplier and trader, SmartestEnergy, has entered markets as far-flung as Texas and Iberia, where the company expects electricity demand to rise, partly because of data-center growth.

The trader also holds interests in three Chilean copper mining operations. One mine, Centinela, is expanding production. In communications, its subsidiary Arteria Networks operates fiber-optic infrastructure across Japan that Marubeni says is positioned to handle rapidly increasing data traffic in the AI era.

Marubeni President and CEO Masayuki Omoto poses for a photo in Tokyo on Oct. 27.

"The businesses linked to these physical constraints are major growth areas in their own right, and we intend to pursue them," Omoto said.

Marubeni was founded in 1858 as a linen trading business, and gradually moved onto general trading. Although still known as a trading house, Marubeni, like its peers, has increasingly shifted from being a simple trader of goods and commodities to investing in, owning and improving businesses where it sees long-term competitive advantages.

In addition to investing more in AI-related sectors, Omoto said the technology can create additional value in businesses where it has proprietary data, specialized expertise and strong relationships with customers.

Omoto cited the example of Helena Agri-Enterprises, a U.S. subsidiary that supplies agricultural inputs such as fertilizer, seeds and herbicides. Even before AI caught on, the company had a proprietary system that combines information about soil, crops, fertilizer and timing to make recommendations to farmers. Such analysis will be "dramatically faster" with AI, Omoto said.

But AI alone will not give Marubeni an advantage, Omoto said. Any competitive edge will come from proprietary data accumulated over decades, and Helena's knowledge of its customers, which will allow it to identify the problems AI should address.

"Ultimately, AI is only as good as the questions it is asked. It cannot produce an answer better than the question," he said. "That is why companies need close customer relationships that enable them to identify the questions that really matter." Marubeni has also embraced AI in-house. The group has deployed around 4,700 AI agents -- software that independently carries out tasks based on user instructions -- across its 150 companies, which employ 26,000 people in all, Omoto said. This has reduced menial tasks equivalent to roughly 1,000 jobs on an annualized basis.

The AI agents also help with investment due diligence. Employees proposing and scrutinizing deals use these agents to collect public information on a scale that a human working alone would struggle to match.

"The amount of information we can consider when making a decision, and the speed at which we can consider it, is dramatically greater," Omoto said.

One agent even allows Marubeni employees to retrieve lessons from past investments that have gone sour, as "AI neither forgets nor defers to anyone," the CEO added.

Businesses that address AI's physical bottlenecks "roughly speaking, account for around half" of Marubeni's profit, but Omoto added: "You could say, since we are looking at improving the competitive advantage of all our businesses through AI, that all our businesses are AI-related." The emphasis on AI is not without risks. Some heads of AI companies, such as Dario Amodei of Anthropic, are calling for a slowdown in AI development, given the risk that the technology could escape human control and cause havoc. The funding methods for data center construction are also increasingly seen as a risk to the global financial system.

The International Monetary Fund, meanwhile, has warned that the technology's impact on the labor market could be "profound." Kristalina Georgieva, its managing director, said in a recent blog post that "unsettling incidents involving frontier models or cyberattacks could dent confidence and raise financial stability risks." Omoto acknowledged there may be a phase where AI development slows down, but he believes the underlying demand for AI will hold up.

"When a technology creates added value for customers and society, people naturally continue pursuing that value," he said. "As long as that remains the case, demand for computing resources will not disappear. Over the medium term, I believe the advance of AI is irreversible."