Heavy Industry Meets the AI Power Crunch

The Shipbuilder That Wants to Power Your AI

HD Hyundai Heavy Industries announced on September 10 that it will invest roughly 1.1 trillion won — about $800 million — across two new manufacturing facilities at its Ulsan complex. One builds massive diesel engines. The other builds components for small modular reactors. Both are aimed at a single question: who powers the AI economy?

This is Chairman Jeong Ki-seon’s first major investment declaration since taking the helm. That alone makes it a signal. More importantly, it reveals something most English-language coverage of the AI energy crunch misses — the fight over power is no longer happening in server rooms or policy briefings. It’s happening on factory floors in South Korea, where industrial manufacturers are realizing they can pivot a 20-year-old ship engine design into a data center power plant.

What makes this move notable is not just the scale of the investment but the speed of execution. HD Hyundai has compressed what would normally be a multi-year feasibility process into a decisive capital commitment. The company appears to believe that the window for establishing manufacturing dominance in AI-grade power infrastructure is narrow and will not stay open indefinitely.

The HiMSEN Engine as a Data Center Micro-Plant

The HiMSEN engine was developed in 2000 as a marine power source. It powers ships. Each unit measures roughly 13 meters long and 4.7 meters tall, producing 20 megawatts. Stack 130 of them on a concrete pad and you have a one-gigawatt mini power plant — exactly the scale a hyperscale AI data center needs.

That insight landed quickly. This year alone, HD Hyundai signed contracts to export HiMSEN engines worth 15,831 billion won to American firms Amerithon Energy Group and Cobra Energy Group. Orders keep coming. The contracts represent a fundamental restructuring of the engine’s original purpose — these units will never propel a vessel. Instead, they will sit stationary, converting diesel into the electricity that runs training clusters and inference workloads.

The company originally planned a modest expansion of existing capacity. Instead, it is building a brand-new 83.36 billion won facility on 215,000 square meters in Onsan, Ulsan, designed to produce 3 gigawatts of engine capacity annually by 2028. By 2030, that output is projected to nearly double to 7.2 gigawatts — more than twice the current rate.

The IEA estimates global data center electricity consumption will exceed 945 terawatt-hours by 2030, more than double what it was last year. HD Hyundai’s logic is blunt: if the world needs more power, and ships already make good engines, why not sell those engines to people who need gigawatts, not propulsion?

There are second-order effects worth tracking. The conversion of marine engines into stationary power units creates a new category of energy asset that sits somewhere between diesel generators and gas turbines. It offers faster deployment than either — a full gigawatt can be installed in months rather than years — while avoiding the permitting delays that have stalled large-scale natural gas projects in the United States. But it also locks data center operators into a fossil fuel supply chain at a time when corporate carbon commitments make that a reputational risk. Companies signing these deals are likely weighing near-term delivery speed against long-term ESG exposure, and the math will shift as grid decarbonization policies tighten.

The SMR Play: Betting on Nuclear’s Next Chapter

If the engine factory is the present bet, the SMR plant is the future one. HD Hyundai is investing 238.6 billion won to build a facility that will manufacture the cylindrical reactor pressure vessels and other primary components for small modular reactors. That plant is scheduled to open in the first half of 2030.

The connection to America runs through Bill Gates. In November 2022, HD Hyundai took a stake in TerraPower, the SMR company Gates founded after growing frustrated with the pace of nuclear innovation. Five months ago, TerraPower selected HD Hyundai as its preferred partner for manufacturing and supplying key equipment for its next-generation reactor design. TerraPower is building the world’s first commercial SMR in Wyoming, targeting completion in 2030.

HD Hyundai’s timing suggests it is positioning itself as the global fabrication hub for a technology that could define the next decade of clean baseload power — particularly for data centers that need constant, carbon-conscious electricity.

The SMR play carries its own set of risks and rewards. TerraPower’s traveling wave reactor design remains unproven at commercial scale. Regulatory approval timelines are uncertain. And the economics of SMRs have been notoriously difficult to land — earlier generations of the technology struggled to demonstrate cost competitiveness against traditional light-water reactors. But the calculus has shifted. AI companies are now willing to pay a premium for guaranteed, round-the-clock power that does not rely on an aging electrical grid. That demand curve makes previously marginal SMR economics suddenly viable.

For HD Hyundai, the fabrication facility represents a longer-term hedge. If SMRs succeed, the company will sit atop a manufacturing moat built from decades of pressure vessel expertise. If they falter, the engine business provides a cash cushion that can sustain the nuclear bet until the market matures. Either way, HD Hyundai ensures it is in the conversation — and in many cases, at the table.

Who Wins, Who Loses

The winners are clear. HD Hyundai is leveraging the deepest cash reserves in its history to enter two markets that will only grow more competitive. Its shipbuilding subsidiary, HD Korea Shipbuilding & Offshore Engineering, reported a 17.6% operating margin in the first half of the year, with an order backlog of $76.1 billion. That profitability is funding the pivot.

Amerithon and Cobra Energy will get engines faster than any American manufacturer could produce them. TerraPower gets a proven heavy-industry partner with welding quality built over decades of shipyard work.

The losers are less obvious but real. Traditional power equipment manufacturers in the US and Europe who assumed the data center generation squeeze would be solved by natural gas turbines or grid expansion face a new competitor that can bolt together 130 ship engines and call it a power plant. Utilities may find themselves competing with tech companies for the same industrial fabrication capacity.

There is also a labor dimension. SMR fabrication requires specialized welders and engineers with nuclear-grade certification — a workforce that is already thin globally. HD Hyundai’s entry into this space could intensify competition for the same pool of qualified workers, driving up costs for everyone else attempting to build SMR components. The ripple effects could slow deployment timelines across the industry, even as they accelerate HD Hyundai’s own production ramp.

And there is a geopolitical dimension. South Korea, a country that imports nearly all its energy, is now positioning itself as a manufacturer of the equipment that powers the world’s most energy-hungry technology. That is a significant shift for a nation that has historically been a consumer, not a supplier, of power infrastructure. It also places South Korea in closer strategic alignment with US energy and technology policy, which could invite countermeasures from competitors like China, which views advanced energy manufacturing as a domain of national security importance.

What Happens Next

The engine factory comes online in 2028. The SMR component facility follows in 2029. Between now and then, HD Hyundai will likely sign more export deals as data center developers scramble for power solutions that can be deployed faster than a conventional plant.

The broader story here is not about AI or nuclear separately. It is about how the AI energy crunch is rewiring global manufacturing hierarchies. The companies that control the physical infrastructure of power — whether diesel or atomic — will sit in a position of unusual leverage over the companies that control computation. HD Hyundai’s trillion-won bet is a declaration that heavy industry intends to claim that leverage.

What distinguishes this moment from previous industrial pivots is the pace at which demand is converging. AI compute growth has outpaced every historical precedent for electricity demand, and the supply chain response has been sluggish by comparison. Grid interconnection queues in the United States now stretch five to seven years. New transmission lines face years of environmental review and local opposition. In that gap, HD Hyundai’s modular approach — deployable, scalable, and fast to install — becomes not just attractive but essential.

The question for observers outside Korea is simple: will this model — repurposing industrial manufacturing capacity for AI power infrastructure — become the blueprint other nations follow, or will it remain a uniquely Korean advantage in the energy race? The answer will shape not just who powers the next generation of AI systems, but who controls the economic rents that flow from doing so.