Doosan Enerbility - The Bottleneck of Power, the Present of Gas Turbines, and the Future of SMR

POWER · AI ENERGY · 2026.06.15

The bottleneck called power,

The present called gas turbines,
The future called SMR

Doosan Enerbility

034020 · KOSPI

The power supercycle driven by AI — Earnings from gas turbines, optionality from nuclear and SMR

June 15, 2026 · Narrative constructed after research and multi-round discussions by the Goguma Team
Power bottleneck · The present of gas turbines · The future of nuclear and SMR · The shadow of Bobcat · Bull/Super-Bull take-profit prices

Onion (Macro/Energy) · Acorn (Industry/Reality) · Peanut (Risk/Value) · Carrot (Chart/Supply-Demand) · Sweet Potato (Summary)

Prologue

AI is eating up all the electricity

The battleground of the AI race is shifting from chips to electricity. Even if you stack hundreds of thousands of the world's most powerful graphics processing units (GPUs), data centers won't turn on without the power to run them. And power is built much more slowly than chips — it takes years to build a single power plant. Thus, the phrase 'the real bottleneck of AI is power' has emerged at the forefront of the market, and Doosan Enerbility stands right at that crossroads.

Doosan Enerbility is Korea's leading power generation equipment company. It is a global top-tier manufacturer of core nuclear power components, a 'foundry (contract manufacturer)' partner for small modular reactors (SMRs), and above all — what AI data centers need right now — a third supplier that has localized gas turbines, penetrating the market vacuum previously split between GE and Siemens. On June 15, 2026, the company's stock price rebounded by over +7% in a single day, recovering to around 100,000 KRW.

This report answers two questions. Why is power the bottleneck for AI, and how does that benefit flow to Doosan — the answer lies in viewing 'the present called gas turbines' and 'the future called nuclear and SMR' separately. And how justified is the valuation of this stock, which has a price-to-earnings ratio (PER) of 188x — from the level supported by gas turbine earnings to the level where the nuclear renaissance fully unfolds. Based on multiple team discussions and the data, we even map out the final take-profit price.

Chapter One

Power is AI's real bottleneck

The market talks about a GPU shortage, but if you take a step back, a deeper bottleneck becomes visible — it is power. Chips can be scaled up in a year, but power plants take several years.

According to the International Energy Agency (IEA) and industry forecasts, data center power demand will double by 2026 and increase nearly fourfold to approximately 1,600 terawatt-hours (TWh) by 2034. The problem is what will meet this surging demand. Nuclear and renewables take a long time to build, and it is projected that gas and coal will meet over 40% of the new power demand through 2030. In other words, there is a structure where gas power, which can be built quickly, fills the 'gap' before next-generation nuclear power like SMRs fully matures.

This is exactly where Doosan Enerbility's three businesses fit perfectly — the immediate gap is filled by gas turbines, and the main game after 2030 is filled by nuclear and SMRs. As long as AI consumes electricity, the company building the equipment to generate that electricity stands at the center of structural demand.

2034 data center power demand forecast (approx. 1,600TWh) — Power runs out before GPUs, and gas turbines fill that gap

Chapter Two

The multipolar power demand created by Sovereign AI

AI data center power demand is no longer just a U.S. story. As each country builds redundant GW-class infrastructure domestically in the name of 'data sovereignty (Sovereign AI),' power demand is geographically dispersing while the total scale is amplifying multiplicatively. This is why Doosan Enerbility's integrated supply position, capable of catering to a 'multipolar power infrastructure' rather than just a single U.S. market, is being highlighted anew.

The essence of the Sovereign AI race is the 'global replication' of power demand. While past AI data center demand was concentrated in the U.S., now the Middle East, Europe, and Asia are building infrastructure of the same scale simultaneously, citing their own data sovereignty. In the Middle East, Saudi Arabia's HUMAIN is adding 1.9GW by 2030 (expanding to 6.6GW by 2034), and its joint venture with AMD and Cisco aims to start with 100MW in 2026 and reach up to 1GW by 2030. In the UAE, the 5GW 'Stargate' campus in Abu Dhabi, involving G42, OpenAI, Oracle, NVIDIA, and SoftBank, is being pursued, with the first 200MW phase of its 1GW cluster scheduled for completion in Q3 2026. Market research firm Wood Mackenzie expects the UAE's data center power consumption to double from about 3 terawatt-hours (TWh) in 2025 to over 6TWh by 2030.

Europe has also joined in. The European Union (EU) is investing 20 billion euros through 'InvestAI' into up to five AI gigafactories (each with over 100,000 advanced AI chips), with formal bidding opening in early 2026. SoftBank announced in May 2026 that it would invest up to 75 billion euros in France to build a 5GW data center, with the first 45 billion euro phase constructing 3.1GW in the Hauts-de-France region by 2031 alongside EDF and Schneider Electric. Currently, Europe's data center power load is led by Germany at 4.26GW, followed by the UK at 3.69GW and France at 1.72GW. In Asia, SoftBank is starting construction on a 1GW campus in Tomakomai, Hokkaido, Japan, in April 2025 to commercialize Sovereign AI GPU cloud in October 2026, while India's Reliance is creating a 3GW campus in Jamnagar, where Meta will lease a 168MW facility. Johor, Malaysia, has surged from 10MW in early 2021 to 1.3GW in November 2024 and approximately 5.8GW by mid-2025, absorbing even the overflow demand from Singapore's data center construction moratorium.

Korea is no exception. The government is pushing a large-scale policy drive to foster Sovereign AI, aiming to expand data center capacity by 50 units and a total of 2GW by 2030. The key is that all these countries are redundantly building the same GW-class infrastructure 'one set per country.' A single 1GW data center absorbs power equivalent to approximately 300,000 to 880,000 households when operating continuously. Distributed construction by country means multiplicative amplification of power demand, which is the background for the IEA's forecast that global data center power demand will double from about 415TWh in 2024 to about 945TWh in 2030 (about 3% of global electricity).

945TWh

2030 global data center power demand (2x 2024's 415TWh, IEA) — It's not just the U.S., the whole world is driving it up

For Doosan Enerbility, this multipolarization is a structural long-term momentum that goes beyond dependence on the single U.S. market. The company has become a symbol of AI power benefits through gas turbines for North American big tech (12 units cumulative, including 7 380MW-class units for xAI's Memphis data center), but it has equally continued a relay of gas combined-cycle EPC (Engineering, Procurement, Construction) orders in markets outside the U.S. Cumulatively from 2025 to 2026, it won Saudi Luma 1/Nairiya 1 (approx. 2.2 trillion KRW), Jafurah Phase 2 (approx. 840 billion KRW), Oman Duqm (approx. 530 billion KRW), and Vietnam O Mon 4 (approx. 900 billion KRW), and directly manufactures and supplies core equipment like steam turbines and generators for these projects. In Europe, it secured the Czech Dukovany project (total approx. 5.6 trillion KRW) and is expanding its nuclear pipeline with Westinghouse AP1000 in Poland.Amid a bottleneck where global gas turbine orders (110GW) overwhelm manufacturing capacity (60-70GW) and prices have risen 195% compared to 2019, Doosan has equipped itself with an integrated position to capture power demand occurring outside the U.S. as the 'third global supplier' after GE, Mitsubishi, and Siemens.

Key — Sovereign AI 'replicates' power demand. It forces the construction of the same GW-class data center in every country, dispersing and amplifying power demand not just in the U.S., but in the Middle East (Saudi 1.9GW, UAE 5GW), Europe (France 5GW, 5 EU gigafactories), and Asia (Japan 1GW, India 3GW, Korea 2GW, Johor 5.8GW). The IEA expects global data center power demand to double from 415TWh in 2024 to 945TWh in 2030. Doosan Enerbility has secured an integrated position to capture power demand outside the U.S. as a third global supplier amidst gas turbine supply bottlenecks (110GW in orders vs 60-70GW supply).

Chapter Three

Gas Turbines — Current Earnings

In the Doosan story, the side that is 'making money right now' is gas turbines. And the core of that business is simple — there are only three companies in the world that can make them, and the first two are already full.

Large gas turbines have extremely high technical barriers, and the global market has effectively been split between GE Vernova (U.S.) and Siemens Energy (Germany). However, due to the surge in AI data center demand, both companies' order backlogs are full for the next 2-3 years. Increasing capacity is not easy. Doosan Enerbility is effectively the world's only 'third supplier' that has localized H-class large gas turbines, and it is penetrating that vacuum. When there are only three suppliers and two are at capacity, the third holds pricing power.

Indicator	Value	Significance
Cumulative Gas Turbine Orders	2025 3 units → 2026.3 7 units (380MW-class)	Cumulative approx. 12 units, approx. 1.2 trillion KRW
Key Customer	xAI (Elon Musk), etc. data centers	Delivery within 2026
Q1 earnings acceleration	1Q26 Revenue 4.26 trillion KRW (+13.7%) · Operating Profit 233.5 billion KRW (+63.9%)	Gas turbines driving profit acceleration
Competitive Landscape	GE Vernova/Siemens Energy order backlogs full for 2-3 years	Doosan = Third supplier, pricing power

The 64% increase in Q1 operating profit shows the power of gas turbines. The big picture identified by the IEA that 'gas fills the gap' and the industrial structure where volumes GE and Siemens cannot take flow to Doosan — this is the real engine boosting Doosan's earnings now.

Chapter Four

Nuclear/SMR — Future Option

Doosan's second story is the 'future that has not yet arrived.' Large-scale nuclear power and SMRs — they are massive, but the point where they ripen into revenue is after 2028. Therefore, this should be read as an 'option,' not 'earnings.'

Large-scale Nuclear — Stepping stone with Czech

Doosan is a global top-tier manufacturer of core nuclear power components (reactors, steam generators, turbines). In December 2025, it won the main contract for core equipment for the Czech Dukovany nuclear plant (approx. 5.6 trillion KRW, ~2038), and revenue recognition began in Q1 2026. It is a direct beneficiary of the global nuclear renaissance extending to Poland, the Middle East, and the U.S.

SMR — The 'Foundry' of Nuclear Power

The bigger future is SMR (Small Modular Reactor). Doosan has established itself as an 'SMR foundry' that contract-manufactures core equipment for design firms like NuScale, X-energy, and TerraPower in the U.S. — a structure identical to TSMC contract-manufacturing designs in semiconductors. It has received reservations for 16 sets of X-energy components, started construction on a dedicated plant with a 20-unit annual capacity, and has also penetrated the UK's Rolls-Royce SMR market. The SMR order target by 2030 reaches NuScale 10.2 trillion KRW + X-energy 13.8 trillion KRW + TerraPower, etc. 4.1 trillion KRW = approx. 28 trillion KRW.

Approx. 28 trillion KRW

2030 SMR order target (NuScale 10.2T + X-energy 13.8T + TerraPower, etc. 4.1T)

Main point of Chapter Three — Options must be valued as options. The potential of nuclear/SMR is massive, but the time when it ripens into revenue is far off — SMR revenue is under 100 billion KRW in 2026, approx. 700 billion KRW in 2027, and meaningful profit is after 2028. Moreover, SMR is a field prone to delays in licensing and commercialization (there are cases in the U.S. and Europe where they retreated due to cost and construction issues). Therefore, this massive number is not 'confirmed earnings' but an 'option that time must prove,' and it is already significantly reflected in the current stock price.

Chapter Five

Numbers — Order backlog 24 trillion KRW, and the shadow of Bobcat

Doosan's fundamentals have both light and shadow. The light is the 24 trillion KRW order backlog, and the shadow is the sluggishness of Doosan Bobcat, which accounts for half of the consolidated earnings.

Indicator	Value	Note
2025 Consolidated Revenue / Operating Profit	17.06 trillion KRW / 762.7 billion KRW	Operating profit -25% (impact of Bobcat sluggishness)
Doosan Bobcat weight	About half of consolidated revenue (52% stake)	Construction equipment cycle slowdown is a drag
Enerbility separate revenue	Approx. 3.33 trillion KRW	Nuclear/Gas Turbine/Generation core business
1Q26 Operating Profit	233.5 billion KRW (+63.9%)	Core business acceleration clear
Order backlog (end of 1Q26)	24.13 trillion KRW (+46%)	New orders 2.79 trillion KRW (+62%)
2026 New order guidance	13-14 trillion KRW	Gas Turbine · SMR · Nuclear Power · EPC
Global Data Center Power Demand (IEA)	415TWh (2024) → 945TWh (2030)	Approx. 2x, driven by AI. Distributed construction across countries outside the US is driving the aggregate growth.
Global Gas Turbine Supply & Demand	Orders 110GW vs Supply 60–70GW	Price +195% compared to 2019, lead time 5–6 years. Doosan = Opportunity as the 3rd supplier
Non-US Gas Combined Cycle EPC (25–26)	Saudi Arabia, Oman, Vietnam, etc. Cumulative approx. 5 trillion KRW+	Jafurah 2 (840 billion KRW), Oman Duqm (530 billion KRW), Vietnam O Mon 4 (900 billion KRW), direct supply of main equipment
North American Gas Turbine / xAI	Cumulative 12 units / xAI 380MW class 7 units	H-Class, direct supply to data centers. 380MW class is 5th in the world and 1st in Korea

The key is to separate 'connections' from the 'core business.' The 25% drop in 2025 operating profit is due to the sluggish performance of the construction equipment subsidiary, Doosan Bobcat, while the core businesses of nuclear power and gas turbines are accelerating with a 64% increase in Q1 operating profit. The 24 trillion KRW order backlog supports the visibility of future earnings for the core business. In short, whether you see Doosan as a 'company held back by Bobcat' or a 'company whose core business is accelerating alongside its order backlog' determines your valuation.

The truth about the 188x P/E ratio. The reason Doosan's P/E appears to be 188x is that 2025 earnings are depressed by Bobcat's poor performance — it is an optical illusion caused by a temporarily smaller denominator (earnings). As the core operating profit accelerates to about 2 trillion KRW in 2026, the forward multiple will drop sharply. However, there is definitely a portion reflecting future profits created by SMRs (after 2028), which is the essence of why this stock 'looks expensive.'

Chapter 6

Four Perspectives — Supercycle or Priced-In?

In the midst of a power supercycle, opinions are divided on the same company. The team debated, 'Will it go further from here, or has the future already been priced in too much?'

Onion · Macro/Energy (Bullish)

"Power is the real bottleneck for AI — data center power demand will increase 4x by 2034, and gas turbines will fill the gap before SMRs arrive. Doosan has pricing power as the only 3rd supplier in a market saturated by GE and Siemens. With a 24 trillion KRW order backlog and a 28 trillion KRW SMR goal, the supercycle supports the P/E. There is room for further upside."

Acorn · Industry/Reality (Balanced)

"Gas turbines are earnings, nuclear/SMRs are options. The 188x P/E is a distortion based on 2025 Bobcat weakness; if core operating profit accelerates to 2 trillion KRW in 2026, the forward multiple will plummet. The current price of 99,800 KRW is below the lower end of securities firms' target prices (122,000 KRW) — however, those targets account for both gas turbine earnings and SMR options."

Peanut · Risk/Value (Cautious)

"SMR revenue won't materialize until after 2028, but the stock price has already absorbed the 2030 scenario. SMRs are notorious for licensing and construction delays, and the governance issue regarding the 52% stake in Bobcat has only been sealed, not resolved. Gas turbines are performance, SMRs are expectations — buying on dips in the 90,000 KRW range is reasonable, but chasing above 110,000 KRW is overheating."

Carrot · Chart/Flow (Practical)

"After pulling back from the June high of 105,600 to the 94,000s, today's +7% move back toward 100,000 is a V-shaped recovery. The trend is alive (IB targets trending upward to 130k-160k). However, volatility is high, so discipline is required — for new entries, buy on dips to 94,000; avoid chasing; stop loss at 88,000."

The sum of the four perspectives is clear. Doosan is a core stock in the power supercycle that holds both the 'proven present of gas turbines' and the 'massive future of SMRs,' but the current stock price has significantly priced in that future. Therefore, the path depends on two speeds — if gas turbine orders continue in the second half, the consensus (125,000–160,000 KRW) is justified, and if SMRs shift from 'expectation' to 'performance,' the upside will open up further. Conversely, if orders slow down or SMR schedules slip, there will be a retracement proportional to the future that was priced in.

Chapter 7

Valuation — Bull/Ultra-Bull Exit Prices

The target prices below depend on 'which business supports the value.' Gas turbine performance sets the base, while nuclear/SMRs open up the Bull and Ultra-Bull scenarios.

Doosan Enerbility (034020) — Currently approx. 99,800 KRW

Scenario	Target Price	Rationale
Bear	85,000 KRW -15%	Failure to meet 2026 order progress, SMR schedule delays, continued Bobcat weakness. Reflects only gas turbine performance.
Base	125,000 KRW +25%	Acceleration of gas turbine performance (core OP 2 trillion KRW) + 24 trillion KRW backlog. Lower end of brokerage targets (Meritz 122K).
Bull	160,000 KRW +60%	Gas turbine supercycle + Czech revenue realization + SMR visibility. Upper end of brokerage targets (IBK 160K).
Ultra-Bull	200,000 KRW +100%	SMR 2028+ full-scale launch (X-energy/NuScale 28 trillion KRW reality) + power supercycle + multiple large-scale nuclear plant orders.

The current price is below the lower end of brokerage targets (122K) — however, that consensus includes the SMR option. Split-buying on dips to 94,000; avoid chasing (above 110K). Stop loss at 88,000. Two triggers for further upside are 'H2 gas turbine/nuclear order announcements' and 'SMR shift to earnings.'

Buy the present and the future separately. Doosan Enerbility is a stock that overlays a 'massive future of nuclear/SMRs' on top of a 'proven present of gas turbines.' Up to the Base (125,000), it is the 'value of the present' supported by gas turbine performance and the 24 trillion KRW order backlog, while the Ultra-Bull (200,000) above that is the 'value of the future' as SMRs shift from expectation to performance. Therefore, the strategy is — buy the present (gas turbines) for their performance, hold the future (SMRs) as an option, and accumulate in parts rather than chasing all at once.

Chapter 8

Signs of a Diverging Narrative

The power supercycle is structural, but the stock price has priced in the future. We define the signals that the future is delayed or going off track.

Signal	Meaning	Response
① Failure to meet H2 order progress	Progress rate in H1 is low compared to 2026 new order guidance (13–14 trillion KRW). Disappointment if large Czech/US orders don't materialize in H2.	Monitor gas turbine/nuclear order disclosures. Reduce weight if targets are missed.
② SMR schedule delay	SMRs are prone to licensing/construction delays. If full-scale revenue is pushed beyond 2028+, the priced-in expectations will evaporate.	Adjust Ultra-Bull timeline. Base remains valid due to gas turbines.
③ Bobcat governance reignition	The 52% stake/merger issue was only sealed, not gone. If reignited, controversy over shareholder value dilution.	Check governance news. Potential discount factor.
④ Valuation overheating correction	188x P/E is vulnerable if core business performance doesn't keep up. Chasing above 110k is the overheating zone.	Split-buy on dips to 94,000, avoid chasing >110k, stop loss 88,000.

Conclusion. One firm truth of the AI era is that no matter how intelligent AI becomes, it won't turn on without electricity. Power is the real bottleneck for AI, and Doosan Enerbility, which builds the equipment to solve that bottleneck, stands at the center of structural demand. Currently, gas turbines are converting that demand into earnings, and after 2028, nuclear/SMRs will take the baton. However, since the stock price has already priced in that future — buy the present (gas turbines) for performance, hold the future (SMRs) as an option, watch the two signals of orders and performance, and follow the discipline of scaling in and scaling out.

This report is a narrative analysis summarizing the research and multiple discussions of the Gogu-ma team (Onion, Acorn, Peanut, Carrot, Gogu-ma).
The suggested target prices are scenarios based on brokerage consensus and the latest industry trends, not investment advice. Investment decisions and responsibilities lie with the investor.
As of June 15, 2026 · Doosan Enerbility (034020 · KOSPI)