The global AI boom has just triggered what may be the largest infrastructure investment in tech history. In late June, the South Korean government announced a staggering, coordinated initiative alongside memory chip titans Samsung Electronics and SK hynix to drastically expand the nation's semiconductor manufacturing capabilities. Specifically, Korean President Lee Jae Myung stated that the goal is to double Korea's DRAM production capacity within 5 years. Public and private commitments as part of South Korea's chip push have so far totaled around 1,350 trillion KRW, which is about $880 billion USD.

So what's up with the $518 billion number in the headline? Well, that's "just" the value of SK hynix and Samsung's commitments to build four new mega-fabs in the southwestern part of South Korea. See, the scale of the investment is so large that it's difficult to even track. The core, government-backed mega-project officially pledges 800 trillion Korean won, which converts to roughly $518 billion USD, but the South Korean government's larger AI infrastructure initiative, which includes the construction of massive multi-gigawatt datacenters, comes out to roughly 1,350 trillion won (about $870 billion) when accounting for all of the projects included.

Today we're just talking about the new chip fabs, though, and those are going to be built by Korea's memory titans, SK hynix and Samsung. Now, those two firms are bitter rivals, locked for decades in a fierce battle for dominance in the memory chip market, so it's fair to wonder how the firms are pulling this off together. The short answer is that they aren't actually collaborating on the tech. There will be no shared facilities or joint chip designs; instead, this is a government-orchestrated regional cluster. By coordinating under a national umbrella, both competitors benefit from fast-tracked building permits, heavy government subsidies, and a shared regional infrastructure grid, but they will each independently build, own, and operate two of the four new fabs.

This infrastructure element is exactly why the expansion is happening in South Korea's southwestern Honam region, near the city of Gwangju. Historically, both companies have clustered their mega-factories near the South Korean capital of Seoul. However, modern chipmaking, specifically for the High Bandwidth Memory (HBM) chips required by advanced AI processors, requires unfathomable amounts of power and water. The Seoul region has simply run out of the utility capacity required to support these massive new plants, prompting the government to shift the tech hub to the southwest where renewable energy and water are still abundant.

To get these facilities online, the South Korean government is slashing red tape to pull the construction schedule forward by up to a decade. However, building a state-of-the-art semiconductor fab from the ground up is a painstakingly slow process. Even with an aggressively expedited schedule, these four new plants are not targeted for completion until the mid-2030s, which means it will be a minimum of eight to nine years before they actually start pumping out silicon despite Lee Jae Myung's goal of doubling production in 5 years.

Because of that long timeline, even this half-trillion-dollar investment isn't going to save us from the "RAMageddon". Massive data center construction has been impeded by various factors, but the companies continue to buy up the world's supply of RAM, driving up costs for everyday consumer electronics. South Korea's 800 trillion won mega-project is definitely a bold long-term play to secure global dominance in the AI era, but everyday consumers and PC builders will still be left to weather the current RAMpocalypse on their own for the forseeable future.

_{Top image: South Korean President Lee Jae-myung, 2018 photo by Gyeonggi Provincial Government (KOGL Type 1)}