AI boom drives fresh demand for steel as data centers, chip factories multiply

The AI boom is injecting new life into South Korea's steel industry, which had been weighed down by a prolonged construction slump. As data centers and semiconductor factories go up across the country, demand for structural steel products such as sections and heavy plates is climbing. With the government pushing a national infrastructure drive — the so-called three mega-projects — analysts expect the upswing in steel demand to continue.

Domestic sales of structural steel sections reached 1.03 million tons in the January–May period this year, up 9.7 percent from the same period a year earlier, according to data from the Korea Iron and Steel Association. The figure marks a rebound after full-year 2025 sales of 2.27 million tons came in below the prior year.

Heavy and medium plates, used mainly in shipbuilding, also rose 9 percent year-on-year to 2.71 million tons over the same period. Rebar, which depends heavily on construction activity, bucked the trend — falling 5.3 percent to 2.65 million tons, extending a 13.3 percent decline recorded last year.

The trend is showing up in corporate earnings as well. Dongkuk Steel Mill, the first major steelmaker to report second-quarter results, said operating profit rose 52.3 percent year-on-year to 45.6 billion won ($31 million). Cumulative operating profit for the first half reached 67 billion won, nearly doubling from the same period a year earlier. The company attributed the improvement to stronger demand from large-scale industrial infrastructure projects — including semiconductor facilities and AI data centers — which lifted margins on structural steel products. Its heavy plate division also posted modest gains in production and sales, supported by robust shipbuilding demand.

Securities analysts are also taking note of the data-center-driven demand surge. Choi Yong-hyeon, an analyst at KB Securities, said incremental steel demand from domestic data center construction is estimated at around 2 percent of total consumption. "Given that the steel industry has long suffered from weak downstream demand, this is a meaningful rebound," he said.

Steel demand from AI data center and semiconductor factory construction is expected to keep growing as the government presses ahead with its three mega-projects. The plan calls for building out 8.4 gigawatts of AI data center capacity in the first phase, expanding to 18.4 gigawatts by 2035. Each 1-gigawatt data center is estimated to generate between 60 trillion and 70 trillion won in investment, putting the total at an estimated 1,100 trillion won.

Semiconductor investment commitments are equally substantial. Samsung Electronics plans to expand its existing production complexes in Pyeongtaek and Yongin, and will invest 400 trillion won to build two new semiconductor fabs in the Honam region. In the Chungcheong region, the company will invest 140 trillion won to add its fifth and sixth HBM fab lines in Cheonan and Onyang. SK Hynix will expand its DRAM-focused fab at the Yongin Semiconductor Cluster and invest 400 trillion won to build two new memory fabs in the Honam region. It also plans to invest 100 trillion won to scale up its NAND production facility in Cheongju.

Park Seong-bong, an analyst at Hana Securities, estimated that building out 18.4 gigawatts of AI data center capacity through 2030 will generate roughly 860,000 tons of steel demand. He added that Samsung Electronics' Pyeongtaek P5 and SK Hynix's Cheongju P&T7 projects, both currently under construction, are expected to generate up to 450,000 tons of structural steel demand — including heavy plates and H-section steel.

"If we factor in the construction of four new fabs, structural steel demand could reach around 900,000 tons — roughly double the volume from projects currently under way," Park said. He added that non-structural steel demand for piping and exterior cladding, along with additional requirements from power grid expansion, means the overall steel demand effect from new fab construction is likely to be substantial.