The global AI boom runs on two things: algorithms and electricity. Software advances at a blistering pace. However, the physical constraints of power and real estate are creating a brutal bottleneck. In Korea, a new partnership offers a surprisingly pragmatic answer. AI full-stack company Elice Group has secured a strategic investment to build a colossal 1-gigawatt AI data center — a project made possible by tapping into the nation’s industrial giants. For foreign investors watching Korea’s AI ambitions, this deal is a blueprint worth studying.
The investment comes from two sources that, at first glance, seem mismatched: Dongkuk Holdings and GS Ventures. For foreign observers, however, this pairing is immediately revealing. Dongkuk Holdings is the holding company of Dongkuk Steel Group, a foundational name in Korea’s heavy industry. Meanwhile, GS Ventures is the corporate venture capital (CVC) arm of GS Group — a sprawling chaebol, or family-controlled conglomerate, with deep roots in energy, construction, and retail. In Korea, chaebol groups wield enormous influence over physical infrastructure, making them natural — if unconventional — partners for a tech buildout.
Therefore, their involvement is not about software synergy. It is about securing the raw materials of the digital age: power and land. This is a marriage of old-world assets and new-world technology. And in a country where grid connections can take years to secure, that marriage carries real strategic weight.
Dongkuk Holdings possesses a large-scale power-receiving facility inside its steel mill properties. In particular, the existing electrical infrastructure at its Pohang factory — a city on Korea’s southeastern coast known as the heart of domestic steel production — provides a ready-made foundation for a power-hungry AI data center. By contrast, GS Group, through its affiliate GS Donghae Electric Power, offers a total package: both the physical site and the full 1GW of power needed to run it.
That figure deserves context. One gigawatt is an enormous amount of electricity. For reference, the most ambitious next-generation AI data center projects currently under development in the United States target around 2GW. In other words, Elice Group’s project sits in the same league as the world’s largest planned facilities. Furthermore, this alliance between a tech firm and industrial titans demonstrates a creative strategy to de-risk infrastructure projects. It bypasses the lengthy, costly process of acquiring land and securing new power grid connections — the two biggest hurdles in data center construction, solved at once.
Elice Group brings critical technology to the table. The company has a proven track record in designing and operating liquid-cooled data centers — a capability that has shifted from niche to essential. The latest AI GPUs consume between 700 and 1,200 watts per chip. As a result, power density per rack has surged from a conventional 10–15 kilowatts to as much as 50–150 kilowatts. Air cooling simply cannot keep pace.
Liquid cooling — and its next-generation variants such as direct-to-chip and immersion cooling — is therefore becoming the industry standard. Elice has also commercialized container-based GPU cluster technology, which allows high-density computing resources to be packed into a compact footprint. This is ideal for training large-scale AI models. Five years of operating high-density data centers has given the company the operational know-how that convinced its industrial partners to write the check. The global market for liquid cooling in AI data centers is projected to grow from $6.6 billion in 2025 to $55.8 billion by 2034 — a compound annual growth rate of 28.7%. Elice is positioning itself at the center of that curve.
“AI infrastructure is a core element directly linked to national competitiveness,” said Kim Jae-won, CEO of Elice Group. “Its starting point is power reception and AI cloud technology.” The remark is not mere corporate rhetoric. It reflects a genuine crisis in Korea’s AI buildout.
The Korean government has set an ambitious goal: becoming one of the world’s top three AI nations — a target often described locally as “AI G3.” However, the country faces a stark infrastructure paradox. The metropolitan Seoul area, where most tech firms cluster, suffers from a saturated power grid. Transmission line construction has fallen badly behind demand. In addition, a special legislation bill — the so-called AI Data Center Special Act — which would allow power purchase agreements (PPAs) and streamline permitting, remains stalled in the National Assembly due to inter-ministry disagreements. PPAs, for context, are long-term contracts that allow large power consumers to buy electricity directly from generators, bypassing the public utility — a common tool in the West but still heavily restricted in Korea.
Lee Hae-min, a lawmaker and former Google engineer, put it bluntly: “Simply securing 260,000 GPUs does not make AI services possible. Data centers, cooling systems, and above all, power are essential.” His warning cuts to the heart of Korea’s dilemma. The nation has the software talent and the government ambition. Nevertheless, without a physical power foundation, those assets sit idle.
For a country determined to be a global AI leader, solving the power problem is the first, most critical step. The Elice–Dongkuk–GS partnership sidesteps the regulatory gridlock by using industrial sites that already hold the necessary power infrastructure. It is a workaround, not a systemic fix — but in the current environment, workarounds win.
The global data center industry is undergoing a quiet but profound shift in site-selection logic. Proximity to fiber networks used to be the primary criterion. Increasingly, however, proximity to large-scale power sources is becoming the overriding factor. The International Energy Agency projects that global data center electricity consumption will more than double — from 415 terawatt-hours in 2024 to 945 terawatt-hours by 2030 — driven almost entirely by AI demand.
This creates a structural tension. Korea, like many major economies, is simultaneously pursuing net-zero carbon targets and aggressive AI infrastructure expansion. In response, analysts expect governments to accelerate the development of microgrids combining nuclear, LNG, and renewable energy — alongside regulatory reform to make power access faster and more flexible. Korea’s next legislative move on the AI Data Center Special Act will therefore be watched closely, not just domestically but by foreign capital evaluating where to deploy in Asia.
As a result of this strategic investment, Elice Group has secured a stable power foundation and a credible path to hyperscale capacity. The project aims to bolster Korea’s self-reliance in AI infrastructure and expand the domestic industrial ecosystem. For investors, the more interesting signal is structural: when a steel mill and an energy conglomerate back a tech company’s data center ambitions, it tells you exactly how physical the AI race has become.
The next frontier of AI is not just about smarter models. It is about who controls the power to run them.
A skincare gadget sold in a pharmacy chain is not where most people expect a…
Only about 1% of South Korea's farmers are under 40. The fields are aging, the…
In 1893, the Joseon Dynasty sent silk, ceramics, and ornate ceremonial flags to Chicago. The…
China's vast domestic market has long attracted Korean entrepreneurs — and just as often, repelled…
Korea once measured national ambition in semiconductor yields and broadband speeds. Now it measures it…
South Korea has one of the world's most wired populations — and, until recently, one…