Now, korea’s Ministry of Science and ICT now projects data center power demand reaching 6.2 GW by 2030, a figure that overtakes earlier official estimates by nearly two years
Decision Focus
On July 22, 2026, South Korean reporting confirmed active legislative and ministerial momentum to expand direct PPAs beyond renewables to include LNG, specifically to serve AI data center load. The operational signal for Global Heads of Data Center Energy is clear: Korea’s procurement framework is under active redesign. Operators building or planning capacity in this market need to map their exposure to both current restrictions and pending rule changes before committing to a long-term power strategy.
90-Second Brief
Now, korea’s Ministry of Science and ICT now projects data center power demand reaching 6.2 GW by 2030, a figure that overtakes earlier official estimates by nearly two years. Under current law, direct PPAs are restricted to renewables, locking large AI data center operators out of stable, dispatchable generation contracts. The AI data center special law originally included an LNG direct-PPA provision, but it was stripped before passage due to opposition from the Ministry of Climate, Energy and Environment. The pressure has not dissolved; it has redistributed across multiple parallel tracks, a National Assembly amendment, and a city-level campaign in Ulsan, where SK Group and AWS are co-developing an AI data center campus.
What Is Really Happening?
The underlying tension is structural, not political. AI workloads running thousands of GPUs around the clock require baseload-grade power with high capacity factors. Korea’s renewable fleet — predominantly solar and wind — cannot meet that reliability profile without firm backup, and KEPCO’s utility tariff structure was not designed for hyperscale procurement. The result is a gap between what the grid offers and what AI operators need.
That gap has already produced workarounds in other markets. xAI installed natural gas turbines directly at its facility rather than wait for grid interconnection. Microsoft contracted the restart of Three Mile Island Unit 1, securing 20 years of dedicated nuclear output from 2028. US policy has moved toward a model that encourages operators to secure their own supply. Korea’s current framework moves in the opposite direction: self-supply structures are explicitly restricted, and distributed energy zones carry a 40 MW generation capacity cap — a ceiling functionally irrelevant to any large data center drawing hundreds of megawatts.
The National Assembly is now discussing an amendment to the Special Act on Distributed Energy that would remove the 40 MW cap for generators supplying AI data centers within designated distributed energy zones. Ulsan has simultaneously launched its own campaign to include LNG among eligible sources for direct PPAs. At the ministerial level, Deputy Prime Minister and Minister of Science and ICT Bae Kyung-hoon stated publicly that LNG must remain on the table to reach 8.4 GW of AI data center capacity by 2029 and 15 GW by 2035. Inter-ministerial disagreement with the Ministry of Climate, Energy and Environment remains the primary brake.
Why It Matters for Global Heads of Data Center Energy
The practical consequence for operators with Korean exposure is a procurement environment in transition but not yet resolved. Direct LNG PPAs are not currently available. The 40 MW distributed energy cap makes zone-based self-supply unworkable at scale. Operators are either sourcing through KEPCO’s standard tariff — not structured for hyperscale loads — or waiting on regulatory outcomes before finalizing power strategy for sites like Ulsan.
The ministerial target of 8.4 GW by 2029 creates a compressed planning window. If direct LNG PPAs are ultimately enabled, first movers with established generator relationships and site-ready interconnection will hold a structural advantage over those who waited for final legislative clarity. Conversely, operators who structured Korean procurement around renewables-only direct PPAs may face renegotiation pressure if the eligible source set expands.
There is also a portfolio-level signal worth reading: when a government revises its own demand forecast from 3.3 GW to 6.2 GW for 2030 — nearly doubling the figure — while simultaneously debating whether the current grid architecture can support that load, it indicates the power availability constraint is more severe than official planning documents had acknowledged. That revision carries direct implications for site selection confidence and interconnection timeline assumptions across Korea.
Forward View
Three fronts are worth tracking. First, the National Assembly amendment removing the 40 MW distributed energy cap: if it passes, large-capacity LNG direct contracts become structurally possible and operators should move quickly to identify generator counterparties. Second, inter-ministerial resolution between the Ministry of Science and ICT and the Ministry of Climate, Energy and Environment: until that standoff resolves, the LNG PPA pathway remains policy intent rather than legal fact. Third, the Ulsan corridor specifically: the SK-AWS campus is the most visible test case, and whatever procurement structure emerges there will likely set a precedent that other operators cite when negotiating with regulators.
What Is Still Uncertain
The timeline for any legislative amendment is not confirmed. The Ministry of Climate, Energy and Environment has already blocked this provision once, and no inter-ministerial agreement has been announced. It is also unclear whether a data center-specific tariff structure — which Minister Bae mentioned as a companion measure — would reduce urgency for LNG direct PPAs or simply run in parallel. The emissions accounting treatment of direct LNG contracts under Korea’s Scope 2 reporting framework has not been addressed in the public record, which matters to operators with 24/7 carbon-free energy commitments. Finally, whether the 6.2 GW demand projection for 2030 holds is itself uncertain — given that it already revised a prior official forecast upward by nearly a factor of two, further revision in either direction remains plausible as AI deployment timelines shift.
One Question for Your Team
If the 40 MW distributed energy cap is removed and LNG direct PPAs become legally available in Korea within the next 18 months, which of your current or planned Korean sites would benefit from immediate offtake negotiations — and do you have generator relationships positioned to move at that speed?
Sources
- Chosun — Korea weighs LNG direct PPAs as AI data centers demand steady power – CHOSUNBIZ (Link)
