Seoul's $22 Billion Bet on Texas Gas—and 8 US Nuclear Reactors

The Power Behind the Chip

South Korea’s answer to the American AI energy crunch is not another chip fab. It is a 6.3-gigawatt gas-fired power plant in Ensign, Texas, and possibly eight large nuclear reactors built on American soil. The two projects sit at the center of Seoul’s $350 billion investment package with Washington, a deal that is quietly redefining what Korea exports to the United States.

The Ensign project is the first domino. According to a closed-door party-government meeting reported on September 7 by Yonhap, South Korea’s Ministry of Trade, Industry and Energy briefed lawmakers that the two sides have aligned on the Texas gas combined-cycle plant as the opening move of the broader investment framework. The plant will sit in a county just north of Austin, in a region where electricity demand has surged alongside a wave of new semiconductor fabs and AI data centers.

The numbers tell the story. The plant’s full capacity will reach about 6.3 gigawatts, built in two phases. The first stage—1.4 gigawatts of gas turbine generation—will go up quickly so operators can test actual demand and economics. A second 4.9-gigawatt combined-cycle block will follow once the initial phase proves its worth. That phasing matters. It signals that Korea is not writing a blank check. It is buying a seat at the table of Texas electricity markets, with an exit ramp if the load factor does not materialize.

The Price of Admission

The price nearly derailed the deal. Korea’s initial offer sat around $20 billion. The American side, according to the report, abruptly pushed for $25 billion—a 25 percent increase. After behind-closed-doors bargaining, the two sides converged near $22 billion. That midpoint, roughly $2 billion per gigawatt of planned capacity, is competitive for a modern gas combined-cycle project in the US but far below the cost of new nuclear build.

The compromise also reveals something about the asymmetry of the negotiation. Korea came in low, expecting a partnership. Washington came in high, knowing Ensign sits in the only US electricity market where developers can actually sell power without waiting decades for regulatory approval. ERCOT, the Texas grid operator, does not regulate prices the way public utility commissions do elsewhere in America. If you build it, you can sell it. That optionality is worth a premium.

The Nuclear Question

The more consequential part of the package is what comes after gas. The United States proposed that $1.2 trillion of the broader $2 trillion investment footprint—excluding $1.5 trillion already earmarked for shipbuilding—be directed toward constructing up to eight large nuclear reactors in the US. That figure, if taken literally, would imply roughly $150 billion for eight reactors, or about $19 billion per unit. No one is calling that realistic under current US construction costs, where each new large reactor has run well north of $20 billion in recent projects. The number is likely a negotiating anchor, not a budget.

Korea’s side appears to be pushing back against that framing. Government sources indicate Seoul is negotiating for a broader, less specific commitment—language around “cooperation on the construction of nuclear reactors” rather than a fixed dollar amount tied to eight units. That wording gives Korea flexibility. It also lets Korea avoid committing capital to a project type that still carries construction risk in the United States, where only two large reactors have broken ground in the past decade.

Why This Matters for the AI Race

The Ensign plant is not an energy play in the traditional sense. It is an AI infrastructure play disguised as power generation. Texas is where Google, Amazon, Microsoft and a growing list of AI-focused developers are building data centers at a pace that has strained the regional grid. The Electric Reliability Council of Texas has warned repeatedly that peak demand is outpacing new generation permits. A 6.3-gigawatt gas plant—roughly the output of six to eight average US nuclear reactors—adds meaningful capacity to a market that is already pricing in future scarcity.

Korea is not just buying electricity. It is buying optionality. Every gigawatt of Korean-owned generation in Texas gives Seoul leverage in future negotiations with American tech companies that will need firm power contracts to justify their own capital spend. The play mirrors how Gulf states built petrochemical complexes near oil fields: not because they lacked other options, but because proximity to the resource creates long-term margin advantage.

The Semiconductor Gap

Perhaps the most surprising detail in the report is what is excluded. Samsung Electronics and SK Hynix—Korea’s two most valuable companies—were not part of this investment package. The US had reportedly requested that they expand memory chip production on American soil, but Seoul kept that demand separate. That decision is strategically significant.

Samsung and SK Hynix already carry enormous capital expenditure commitments at home and in the United States. Adding a government-negotiated expansion mandate on top of voluntary investment would compress their already thin free cash flow and weaken their competitive position versus Taiwan’s TSMC in advanced logic. By carving semiconductors out of the deal, Korea is protecting its crown jewels from political bargains it did not orchestrate.

The trade-off is clear. Korea gains a foothold in American energy infrastructure without locking its champions into a politically dictated expansion schedule. The US gets a $22 billion check and a partner willing to build critical capacity. Both sides walk away with something, though the asymmetry favors Washington, which controls the demand side.

What Comes Next

The next milestone is a memorandum of understanding, expected as early as September 18. Before that, the investment plan must clear two committees: the Korea-US Strategic Investment Project Management Committee, chaired by the minister of trade, industry and energy, and the Korea-US Strategic Investment Operations Committee, chaired by the minister of finance. Once both sign off, the plan goes to the National Assembly for reporting—a procedural step, not a veto point, given the current governing majority.

The Ministry of Trade, Industry and Energy issued a standard clarification on September 7, saying the reports were inaccurate and that specific details remain under negotiation. That is standard bureaucratic positioning before an MOU. The substance, however, is already leaking through.

The Bigger Picture

Seoul’s energy bet in Texas is a signal that Korea’s post-chip strategy is taking shape. For decades, Korean industrial policy chased export volume in semiconductors, ships and batteries. The AI era is changing the arithmetic. Chips are only as valuable as the electricity that runs them, and the United States is the only major market where Korean capital can deploy at scale into generation assets that American AI builders actually need.

The Ensign plant will come online years before any Korean-owned reactor breaks ground in America. That sequencing is deliberate. Gas funds the nuclear conversation. The $22 billion buys Korea a seat at the energy table; the eight-reactor proposal is the long game. If Washington ever reforms its Nuclear Regulatory Commission timeline and brings large reactor costs down, Korea will already have a Texas reference project and a relationship with American utilities and developers.

The deal does not solve Korea’s structural challenges—demographic decline, domestic investment gaps, competition from China in midstream manufacturing. But it does give Seoul a new tool: the ability to export capital into American energy infrastructure and capture a slice of the AI boom’s most overlooked bottleneck. In a decade when energy is the new geopolitics, that may be the most valuable export Korea has.