Korea and Japan put their US investment pledges into power plants beside data centers

Japan's Pennsylvania plant is the only deal in this group with both a price and a size on the record: $17 billion for 4.3 gigawatts ^[11], or about $3.95 billion per gigawatt ^[19]. The Seoul Economic Daily reported Encinal's total cost of $23.3 billion ^[23] but did not report its capacity. At Pennsylvania's rate, that sum would buy about 5.9 gigawatts ^[24]. Treat that as an estimate until the developers publish a figure.

The strongest evidence that Washington is steering the money is where the plants are going. According to the paper, most of the selected sites are places where data centers are already clustered or planned ^[8]. Texas, which the paper called the world's most closely watched data-center hub, gets gas plants from both Korea and Japan ^[7]. GE Vernova Hitachi will lead small modular reactor projects in Tennessee and Alabama, two states that already host large data centers run by SpaceX, Google and Meta ^[6]. Citing analysts it did not name, the paper concluded that the US is using allied investment as leverage to ease its power bottleneck ^[9].

A plant built next to its customer also needs less transmission. By the paper's account, the US grid has a long way to go to catch China's ^[16]. Citadel Securities expects the US to struggle with grid interconnection, permitting and local opposition to data centers ^[13]. China had 370 gigawatts of inter-provincial and inter-regional transmission capacity last year, according to its National Energy Administration ^[14]. The Department of Energy is aiming for a 16% increase in long-distance transmission capacity by 2030 ^[15]. I'd expect allied money to keep going to plants sited next to their customers for as long as that gap holds.

On scale, the deals look small. Our World in Data put US generation last year at 4,520 terawatt-hours against China's 10,583 ^[12], so the US produced 42.7% as much ^[21]. Citadel expects China to add nearly six times as much capacity as the US over the next five years ^[13]. If the Pennsylvania plant ran every hour of the year, it would produce about 37.7 terawatt-hours, or 0.83% of last year's US total ^[22].

The Texas plant can go one of three ways. Its output can be contracted to data-center tenants near Encinal, which is what the steering argument assumes. It can sell into the wider Texas market, in which case it is a power plant that happens to sit in a data-center region. Or data-center demand can arrive after the plant does, and the Korean money pays for capacity before its customers exist. The paper did not report who buys Encinal's power or what the investment fund earns on it. In my view the siting supports the steering thesis, but the contracts have not tested it yet. Two things would count against it: a purchase agreement with the general Texas market, or a first phase that slips past its 2029 date ^[4].