KHNP's Dukovany subcontract collapse: a fixed-price gamble that Korea's own construction model could not absorb

KHNP has walked away from its own construction consortium at Dukovany. On 29 October the company said it had begun terminating the joint cooperation agreement with Daewoo Engineering & Construction and Doosan Enerbility, and would tender for a replacement contractor, aiming to name one by July 2027. Bidders will be briefed in January and bidding opens in April.

The headline reads like a Korean corporate falling-out exported to Central Europe. It is more accurately a dispute about who carries quantity risk on a fixed-price nuclear build — the same question that has wrecked EPC margins from Olkiluoto to Vogtulbe, and one that KHNP must now resolve in public, against a 2039/2040 completion schedule it has just re-committed to.

What actually broke

KHNP holds the KRW26 trillion (about USD19.4 billion) main contract signed in June 2025 with Elektrárna Dukovany II (EDU II), the Czech state project company. Daewoo E&C and Doosan Enerbility had been nominated construction partners as far back as 2018 and had bid alongside KHNP. Both were named in the "Team Korea" scope split announced at contract signature: KEPCO E&C on design, Doosan on major equipment and construction, Daewoo on construction, KEPCO NF on fuel, KEPCO KPS on commissioning.

According to World Nuclear News, the talks failed across five interlinked points: construction cost, contract structure, scope of work, baseline quantities, and how the main contract's terms would flow down into the construction subcontract. Reduced to its core, the argument was:

KHNP's logic is defensible on its own terms: it has taken a fixed-price posture toward the Czech state and wants the same discipline below it. The consortium's logic is equally standard: nobody prices a nuclear island off basic design and eats the difference. The two positions are not reconcilable by goodwill — only by one side accepting the overrun exposure.

Doosan's statement is notably unhostile, noting that existing consortium members will be allowed to rebid and that it will "prepare diligently, submit a bid." Daewoo, by contrast, says it wants to keep discussing — language that reads as a party that does not expect to be excluded permanently, but does want the record to show it was not the one who walked. Analysis: that asymmetry suggests Doosan's dual equipment-plus-construction role gives it more commercial leverage and less to lose from a rebid, while Daewoo's exposure is concentrated in the contested construction scope.

Why the schedule claim is the real test

KHNP reaffirmed site preparation for January 2028, full construction start in November 2029, and completion in 2039 and 2040. A contractor selected by July 2027 would have roughly eighteen months to mobilise before site preparation and about 28 months before full construction.

That is achievable on paper, but it compresses the wrong activities. A replacement contractor inherits a basic design it did not develop, a quantity baseline it has publicly disputed the logic of, and a client that has just demonstrated it will terminate rather than concede on risk allocation. Two consequences follow:

• The rebid itself will be a negotiation about the same five points. If KHNP holds the lump-sum-on-basic-design line, it should expect higher pricing, more qualification, or thin participation. If it relaxes, the original consortium's position is vindicated and the three-month termination process looks expensive.
• The July 2027 date is an intention, not a contractual milestone. WNN notes the process and timeline are not finalised. Any slippage moves the whole 2028–2029 mobilisation window, and with it the materials procurement that depends on detailed design.

Equipment is insulated; civil works are not

The most important structural fact is what this dispute does not touch. Doosan confirmed that its reactor equipment supply contract and turbine generator equipment contract with KHNP, both signed 16 December 2025, "remain in effect without any amendment or termination." Long-lead manufacturing — reactor pressure vessels, steam generators, turbines — sits on its own contract architecture and its own multi-year forging and fabrication slots.

So the damage is concentrated in civil construction and the quantity-sensitive bulk materials that flow from detailed design: rebar, structural sections, liner plate, embedments, and the concrete-embodied steel that dominates a two-unit APR1000 site. That is a timing problem before it is a volume problem. The bulk of Dukovany's civil steel demand falls in roughly 2029–2033 regardless of who pours the concrete; what has been lost is the 2026–27 detailed-design quantity baseline that would have been priced and procured. Analysis: the practical risk is not cancelled tonnage but a later, lumpier, more volatile procurement curve, awarded under schedule pressure to a contractor with less time to optimise — a recipe for buy-side urgency in a European rebar and sections market that is otherwise soft.

The Czech angle nobody in Seoul controls

Dukovany is a Czech state project with explicit localisation expectations: Industry and Trade Minister Lukáš Vlček cited initial Czech involvement of about 30%, rising toward 60% by the end of construction, with Škoda Power's turbine hall scope alone worth up to 15–20% of contract volume. A Korean construction consortium was always going to subcontract heavily into Czech and wider European civil capacity.

That cuts two ways. It limits how much a Korean contractor change actually alters local supply chains — the Czech subcontractor base was going to do this work either way. But it also means EDU II and Prague have a legitimate interest in how the rebid is run, and a new entrant — potentially a European contractor with Central European nuclear civil experience — would fit the localisation narrative better than the incumbent. Analysis: expect the Czech side to press for a wider bid field than a straightforward Korean re-tender, even if KHNP nominally controls procurement.

Worth flagging as unconfirmed: the currency basis of the headline figure. WNN's June 2025 coverage put the project at CZK407 billion (USD18.6 billion); the October 2026 article uses KRW26 trillion (USD19.4 billion) for what is described as the same contract. Different currencies, different FX dates, and possibly a different scope basis (EPC plus a decade of fuel versus EPC alone). The two should not be read as a price revision without confirmation — the implied rates are roughly KRW1,340/USD and CZK21.9/USD respectively.

What to watch

• January's bidder briefing. Who attends is the single best signal of market appetite. A field limited to the two incumbents means KHNP has won the argument but not the market; a wider field, including European civil contractors, means the risk terms have softened.
• Whether KHNP publishes the contract form before bidding. Releasing the lump-sum/remeasurement architecture ahead of the April bid would show the dispute has been genuinely resolved rather than deferred.
• The 2028 site preparation date. It is the first hard, physical milestone and the first place slippage will become visible in steel and concrete order books.

The broader lesson for the nuclear supply chain is uncomfortable but simple: a fixed-price main contract is only as bankable as the subcontract architecture beneath it. KHNP has spent a year and a half building the top of that structure and has now discovered the base does not fit. The cost of that discovery will appear either as a higher replacement bid or as a later start — and possibly as both.