Ulsan GPS on Korea Invest Insights

SK Gas LNG Cold Energy: How Much Of The AI Data-Center Cooling Option Can Become Earnings?

Sat, 04 Jul 2026 14:48:00 +0900

Related context This follows SK Gas vs SK Oceanplant, AI data-center power bottleneck map, and AI data-center CapEx bottlenecks. The related hubs are the Exclusive Analysis Hub and the Korean Semiconductor Value Chain Hub.

TL;DR

The right wording matters. The confirmed fact is not that SK Gas has disclosed a commercial AI data-center cooling project. The confirmed fact is that an SK Gas engineer published the technical logic for connecting LNG cold energy with AI data-center cooling through SK Group’s DevOcean channel.

The January 2, 2025 DevOcean post, authored by Byeon Jae-jun of SK Gas, says he works on a task force linking LNG cold-energy utilization technology with AI data centers. The post explains that LNG produces about 0.2 MWh of cold energy per ton when warmed from about -162°C liquid to about 5°C gas, and that most of this cold energy is currently discarded through seawater heat exchange.¹

The investment conclusion is precise:

Item	View
Confirmed	SK Gas-linked technical logic exists. Ulsan LNG infrastructure and the SK-AWS AI data-center direction are publicly visible.
Investment logic	LNG cold energy is not new cooling energy. It is the reallocation of otherwise wasted low-temperature energy from LNG vaporization into data-center cooling load.
Unconfirmed	Actual commercial contract, target PUE, realized PUE improvement, tariff, capex, payback, AWS/SK economics and SK Gas margin attribution remain undisclosed.

So SK Gas is not a pure “cold-energy theme stock.” The cleaner thesis is undervalued core energy-infrastructure earnings plus an under-modeled AI data-center cooling option.

1. Why Ulsan Matters

LNG cold energy is hard to move over long distances. The key condition is physical proximity between the LNG terminal and the data center. Ulsan may satisfy that condition.

KET, a joint venture between Korea National Oil Corporation and SK Gas, operates an energy storage terminal in Ulsan North Port and began commercial operation of its oil and LNG terminal in 2024.² SK Telecom has also announced that it is working with AWS on an AI data center in Ulsan, using SK Group land and energy supply capabilities.³

SK Gas also has Ulsan GPS, an LNG/LPG dual-fuel combined-cycle power plant. SK Gas sold a 49% stake in Ulsan GPS for KRW 1.2242 trillion while retaining 51% control.⁴ This makes Ulsan a combined power, LNG terminal, cooling and AI data-center campus candidate rather than a generic LNG story.

Cooling is important, but the common claim that cooling is 30-40% of all data-center power should not be applied to all modern hyperscale facilities. IEA data show that cooling and environmental control may be around 7% in efficient hyperscale data centers, while it can exceed 30% in less efficient enterprise data centers.⁵

That distinction matters. LNG cold energy may lower electric chiller load, but the actual PUE benefit depends on facility design, rack density, cooling loop efficiency, pumps, heat exchangers, backup chillers and utilization matching.

3. How The System Would Work

The simplified chain is:

Step	Meaning
LNG storage	LNG is stored near -162°C.
Vaporization	LNG must be warmed and vaporized before use. Cold energy is released in the process.
Recovery	Instead of discarding that cold energy into seawater, a heat-exchange system captures it.
Coolant loop	Refrigerant or brine loops move the cooling effect safely toward the data-center cooling plant.
Data-center cooling	Electric chiller load can be reduced.
Redundancy	Mission-critical data centers still need backup chillers, pumps and control systems.

The value is real only if cold recovery, reliability, capex and contractual economics work together.

4. Investment Idea One: A Cold-Energy Toll Road

If commercialized, SK Gas and KET could turn otherwise wasted cold energy into a cooling service.

Variable	Driver
Price	Avoided chiller electricity cost, water cost and carbon cost can partly become a cold-energy tariff.
Quantity	Ulsan AI data-center MW scale, rack density, LNG vaporization volume and cold-energy utilization rate.
Cost	Heat exchangers, coolant or brine loops, pipes, pumps, control systems, backup chillers and maintenance.

The upside is a potential infrastructure-style toll road without directly buying GPUs. The missing proof is a long-term cold-energy supply contract, minimum volume, tariff, capex payback, target PUE and SK Gas margin attribution.

5. Investment Idea Two: Firm Power Call Option

The cold-energy idea sits on top of a more basic thesis. AI data centers need firm power. They do not only need cheap power. They need power that does not stop.

Ulsan GPS gives SK Gas a firm-power option. The key variables are power-market price, capacity payments, LNG/LPG fuel spreads, plant utilization and SK Gas’s retained 51% exposure after the stake sale.

The KRW 1.2242 trillion price for 49% of Ulsan GPS implies a simple 100% value of roughly KRW 2.5 trillion and a retained 51% value near KRW 1.27 trillion. This is only a simple cross-check, not a full valuation, but it supports the idea that the market may still be applying a large discount to SK Gas’s energy infrastructure assets.

6. What Is Still Missing

Item	Status
Target PUE for the Ulsan AI data center	Not disclosed
LNG cold-energy share of total cooling load	Not disclosed
Cold-energy supply tariff	Not disclosed
Long-term contract structure	Not disclosed
Cold-recovery capex and payback	Not disclosed
AWS/SK economic allocation	Not disclosed
SK Gas EBITDA or net-income contribution	Not disclosed
Proprietary technology or patent scope	Not disclosed

This table is the core of the analysis. The technical logic is credible, but the numbers required for an EPS model are not public yet.

Final View

SK Gas’s LNG cold-energy option is not empty. It has technical logic, a relevant Ulsan location and a broader SK-AWS AI data-center context. But it is not yet a confirmed earnings driver.

The right investment stance is Watchlist / conditional buy. SK Gas should be read as an undervalued energy infrastructure company with an AI data-center cooling option. The cold-energy thesis becomes investable only when the market sees a supply contract, PUE improvement, capex payback and SK Gas margin capture.

SK Gas vs SK Oceanplant: Which Second-Line AI Data-Center Power Candidate Is Cleaner?

Sat, 04 Jul 2026 05:54:00 +0900

Context This is a follow-up to AI data-center capex and Korea’s bottlenecks, VinaTech and Bloom Energy, GNC Energy and AI data-center backup power, construction and EPC rerating and H1 2026 AI infrastructure review.

TL;DR

After SK Group’s AI data-center investment announcement, the market did not treat every SK affiliate as a clean winner. It looked for the names that are closest to the actual power bottleneck. On that basis, SK Gas is the more direct second-line AI data-center power candidate than SK Oceanplant.

View	Summary
SK Gas	Directly linked to Ulsan GPS, a 1.2GW LNG/LPG combined-cycle power plant, the KET LNG terminal, fuel trading capability and SK Group’s Ulsan AI data-center power structure.
SK Oceanplant	Not a direct power supplier to AI data centers. It is an offshore-wind foundation manufacturer, and works better as a second-line candidate if the power-shortage theme broadens into renewables and offshore-wind infrastructure.
Current priority	SK Gas > SK Oceanplant. SK Gas has valuation and asset-revaluation support. SK Oceanplant has early flow-turn signals, but the chart and earnings visibility are still weaker.

0. Basis

Item	Detail
Companies	SK Gas, SK Oceanplant
Topic	Second-line power candidates after SK Group’s AI data-center plan
Price, consensus and flow basis	July 3, 2026 close and recent flow snapshot
Public-source check	July 4, 2026 KST
Key public events	SK Group AI data-center plan, Ulsan AI data-center project, SK Gas sale of 49% of Ulsan GPS, SK Oceanplant’s Taiwan offshore-wind foundation delivery
Conclusion	SK Gas first, SK Oceanplant conditional watchlist

This is not a claim that either company has a confirmed AI data-center power contract. The point is to rank exposure, valuation and flow quality.

1. What The Market Selected After SK Group’s Announcement

SK Group’s June 29, 2026 announcement framed AI infrastructure as a full-stack system that combines semiconductors, data centers, power and telecom. Local reporting said SK is already working with AWS on an Ulsan AI data center, combining SK Telecom, SK Broadband and SK Gas capabilities, and plans to expand similar models toward 15GW of AI data centers.¹

Al Jazeera also reported South Korea’s broader chip and AI data-center investment drive. The reported plan includes large-scale data-center investment involving SK Group, GS Group and Naver, and mentions a 10GW additional AI data-center target by 2035.²

But stock-market reaction was selective.

Item	Jun 30-Jul 3 return	Interpretation
KOSPI	-3.6%	Broad market correction
SK Inc.	-12.1%	Holding-company and funding discount
SK Hynix	-7.7%	Crowded memory leader pulled back
SK Telecom	-4.2%	AI data-center opportunity offset by cost concerns
SK Square	-3.1%	NAV discount partly cushioned the move
SK Gas	+2.6%	Power and gas infrastructure directness stood out
SK REITs	+3.6%	Infrastructure and asset-value defense

The signal is not the absolute gain, but the dispersion. The market did not simply buy the SK Group label. It preferred the affiliates tied to power, land, infrastructure and asset value.

2. Core Comparison

Item	SK Gas	SK Oceanplant
Price	KRW 218,500	KRW 15,320
Jun 30-Jul 3 return	+2.6%	-0.8%
1-week return	+7.6%	+6.5%
1-month return	-2.0%	-5.4%
2-month return	-24.4%	-37.6%
2026F PER	3.9x	21.6x
2026F PBR	0.52x	1.15x
2026F ROE	15.3%	5.4%
Consensus target price	KRW 367,000	KRW 26,167
Target-price upside	+68.0%	+70.8%
AI data-center directness	High	Low to medium
Investment character	Power and gas infrastructure plus undervaluation	Offshore-wind infrastructure plus turnaround option

The target-price upside looks similar. That is not enough. The better question is whether the company is close to the bottleneck, whether valuation is already stretched, and whether flow is turning from a damaged chart into real accumulation.

On those tests, SK Gas is ahead.

3. SK Gas: The More Direct Power Candidate

SK Gas has a clearer link to SK Group’s AI data-center power structure. The Ulsan project reportedly combines SK Telecom, SK Broadband and SK Gas capabilities.¹ That makes SK Gas more than a generic group affiliate.

The relevant assets are straightforward.

Asset or capability	Investor meaning
Ulsan GPS	1.2GW LNG/LPG dual-fuel combined-cycle power plant, the closest physical power asset in this comparison.
KET LNG terminal	LNG procurement, storage and supply infrastructure.
LPG/LNG trading	Fuel procurement and pricing capability that can affect generation economics.

SK Gas describes its LNG and power business on its official site.³ The Fact reported that Ulsan GPS started commercial operation in December 2024 as a 1.2GW LNG/LPG dual-fuel combined-cycle power plant. On June 30, 2026, SK Gas sold a 49% stake in Ulsan GPS for roughly KRW 1.2242tn while retaining a 51% controlling stake.⁴

That transaction matters because it confirms external asset value, releases capital and preserves strategic control.

SK Gas Investment Points

Point	Interpretation
AI data-center power directness	Potentially tied to SK Group’s Ulsan AI data-center power stack.
Valuation	2026F PER 3.9x and PBR 0.52x are not demanding.
Asset revaluation	The Ulsan GPS stake sale validated infrastructure value.
Shareholder return	2026F DPS KRW 10,500, roughly a 4% yield.
Flow	Latest four trading days: institutions +KRW 0.52bn, real-money-style flow +KRW 0.86bn.

The stock is not yet a strong momentum name. It is above the 20-day moving average, but still 8.9% below the 60-day line, with RSI around 40. This looks more like a low-valuation infrastructure name trying to base after a drawdown than a confirmed trend stock.

Level	Meaning
KRW 214,000-218,000	First support zone to monitor.
Reclaim of KRW 222,000	Initial strength confirmation.
Break above KRW 231,500	A stronger trend-reversal signal.
Break below KRW 203,000-209,000	Weakens the rebound thesis.

The key follow-ups are the second-half strategy update, actual power-supply structure for the Ulsan AI data center, use of the Ulsan GPS sale proceeds and the 2027-2029 shareholder-return policy.

4. SK Oceanplant: Offshore-Wind Infrastructure Option

SK Oceanplant should be framed differently. It does not directly sell electricity to AI data centers. It manufactures offshore-wind foundations. So the right thesis is not “direct AI data-center power exposure,” but “second-line exposure if the power-shortage theme broadens into renewables and offshore-wind infrastructure.”

Recent business news is constructive. Yonhap reported that SK Oceanplant delivered the first foundations for Taiwan’s Fengmiao 1 offshore-wind project on July 2, 2026. The project uses 33 15MW turbines for a 495MW wind farm, and SK Oceanplant signed a KRW 393.4bn contract in June 2024 to supply 21 foundations.⁵

The company is also investing KRW 1.153tn in what it describes as the world’s largest specialized offshore-wind foundation production base, targeting completion by the end of 2026.⁶

SK Oceanplant Investment Points

Point	Interpretation
Theme	Benefits if AI power shortages broaden into generation buildout and offshore wind.
Recent event	First Fengmiao 1 delivery gives visibility to the KRW 393.4bn contract.
Flow	Jun 30-Jul 3: foreigners +KRW 1.67bn, institutions +KRW 3.07bn, retail -KRW 5.02bn.
Program flow	Jun 30-Jul 3 program buying of KRW 3.00bn.
Burden	2026F PER 21.6x, ROE 5.4% and a still-damaged chart.

The recent four-day flow is constructive, but not enough to call structural accumulation. On a 20-day basis, foreigners plus institutions were still net sellers by KRW 18.69bn.

The chart is also still weak. The stock is 4.6% below its 20-day moving average, 21.8% below the 60-day line and 48.5% below its two-month high. The July 1-2 rebound faded again on July 3.

Level	Meaning
KRW 14,300-14,700	Aggressive support zone.
KRW 16,200-16,650	First strength-confirmation zone.
Break above KRW 17,800	Potential trend reversal if turnover holds.
Break below KRW 13,900	Invalidates the early turn.

For SK Oceanplant, the decisive variables are additional offshore-wind orders, the new yard schedule and upward revisions to 2027 earnings estimates.

5. Final View

SK Gas is the better second-line candidate today.

It is closer to the AI data-center power stack.
It trades at a lower 2026F PER and PBR.
The Ulsan GPS stake sale validated asset value and provided capital.

SK Oceanplant is still worth watching, but it is not the same kind of exposure. The Taiwan delivery and large production-base investment are meaningful. Still, this is an indirect offshore-wind infrastructure option, not a direct AI data-center power supplier.

Rank	Name	Stance	What to confirm
1	SK Gas	Watchlist / Pullback Buy	KRW 214,000-218,000 support, reclaim of KRW 222,000, Ulsan AI data-center power structure, shareholder-return policy.
2	SK Oceanplant	Watchlist	KRW 14,300-14,700 support, reclaim of KRW 16,200-16,650, new offshore-wind orders and 2027 earnings revisions.

The simple conclusion is this: in the AI data-center power trade, SK Gas is the closer infrastructure asset. SK Oceanplant is a higher-beta candidate for a later phase if the market expands the theme into offshore-wind generation buildout.

Ulsan GPS on Korea Invest Insights

SK Gas LNG Cold Energy: How Much Of The AI Data-Center Cooling Option Can Become Earnings?

TL;DR

1. Why Ulsan Matters

2. The Cooling Share Should Not Be Overstated

3. How The System Would Work

4. Investment Idea One: A Cold-Energy Toll Road

5. Investment Idea Two: Firm Power Call Option

6. What Is Still Missing

Final View

SK Gas vs SK Oceanplant: Which Second-Line AI Data-Center Power Candidate Is Cleaner?

TL;DR

0. Basis

1. What The Market Selected After SK Group’s Announcement

2. Core Comparison

3. SK Gas: The More Direct Power Candidate

SK Gas Investment Points

4. SK Oceanplant: Offshore-Wind Infrastructure Option

SK Oceanplant Investment Points

5. Final View