On July 2 HD Hyundai Electric Co., Ltd. (KRX:267260) told the exchange it had signed a long-term supply agreement for data centre power infrastructure. The value is $721,293,753, which the filing converts to ₩1,121.2bn at ₩1,554.40 to the dollar, the first quoted rate at KEB Hana Bank that day.
Against FY2025 revenue of ₩4,079.5bn, that single agreement is 27.5%.
The customer is described as a global big tech company. It is not named, and the structure of the disclosure means the contract you can read is not the contract that matters.
Read the counterparty field carefully. It says HD Hyundai Electric America Corporation.
That is the company's own Atlanta subsidiary. The explanatory note sets out the chain: the Atlanta entity signed an agreement with a global big tech company, described as a product-specific supplementary agreement, and then placed a back-to-back order with the Korean parent.
So what has been disclosed to the Korean market is the intercompany leg. The customer contract sits one step removed, inside a US subsidiary, and its terms are not in any filing.
This is not evasion; Korean disclosure rules capture what the listed entity signed, and the listed entity signed with its own subsidiary. But it does mean an investor cannot see the things that determine whether $721m is real: whether volumes are committed or indicative, what the termination rights look like, whether pricing is fixed or indexed, and what happens if the customer's data centre plans change.
For a framework running to January 2029 with a hyperscaler, those terms are the whole substance.
The filing is explicit about the mechanism, and it is the most important line in the document.
Actual purchase orders under this agreement will be concluded individually, several times, over the contract period. Any PO that meets the disclosure threshold will be filed separately.
That tells you two things. First, the $721m is a framework value rather than committed work, so it belongs in a different mental category from a signed order. Second, and more usefully, the company has committed to disclosing the POs as they come. Which means investors get a running scoreboard on whether the framework converts.
If large PO filings appear through late 2026 and 2027, this was real. If the framework sits quietly with only small orders against it, the number in the July headline was a ceiling rather than a forecast.
The contract is denominated in dollars and reported in won, and the won has moved a long way since July 2.
Korean filings date-stamp their exchange rates, so the trail is easy to follow. This contract used ₩1,554.40 on July 2. Samsung Electro-Mechanics used ₩1,475.60 on July 22. Doosan Enerbility used ₩1,402.50 on August 20.
From ₩1,554.40 to ₩1,402.50 is a 9.8% appreciation in the won across seven weeks. Applied to $721.3m, the same contract signed at the August rate would have been reported at roughly ₩1,011bn rather than ₩1,121bn, a difference of about ₩110bn.
Nothing has changed about the work. HD Hyundai Electric earns most of its profit from exports and reports in won, so a strengthening won compresses reported revenue and margin on every dollar contract it holds. The half-year report does not describe hedging on this specific agreement.
The commercial logic on the customer's side explains why this company is worth what it is worth, and it is worth spelling out for a US reader.
Large power transformers and the switchgear that goes with them are the binding constraint on connecting a data centre to a grid. Lead times went from months to years after 2022, because global manufacturing capacity did not expand during the decade of flat electricity demand that preceded the AI build-out, and because the certification and qualification process for high-voltage equipment takes years rather than quarters. You cannot improvise a 500kV transformer.
A hyperscaler facing that constraint does the only rational thing: it locks in capacity with a qualified manufacturer years ahead, accepts a price it would have laughed at in 2019, and signs a framework rather than shopping order by order.
That is also why HD Hyundai Electric's economics changed so completely. Operating margin was negative 8.85% in FY2019. It was 24.40% in FY2025, and the second quarter of 2026 ran at 25.14%. Gross margin went from 13.02% in FY2021 to 34.14% in FY2025. Those are not the numbers of a heavy electrical equipment maker in normal conditions. They are the numbers of a scarce supplier in a shortage.
Two things, and both are structural rather than about this contract.
The first is that shortages end. Every major grid equipment manufacturer is adding capacity, and the current pricing is precisely what funds that expansion. The company's own capital spending has doubled, from ₩121.6bn in FY2024 to ₩233.5bn in FY2025 and ₩139.9bn in the first half of 2026 alone. When enough capacity arrives, a 34% gross margin becomes a memory, and frameworks signed at today's prices become the customer's good deal rather than the supplier's.
The second is valuation. At the August 27 close of ₩821,000, up 12.01% on the day, across 36,047,135 shares, the market capitalisation is ₩29.59tn against total equity of ₩2.26tn at June 30. That is roughly 13 times book value. Annualise first-half net income of ₩413.8bn and it is about 36 times earnings. The shares already price a long shortage.
The counter, and it is a fair one, is that the shortage has lasted four years and the order book keeps extending. This company raised its 2026 order guidance from $4,222m to $5,185m in June, an increase of 22.8%, before this contract was even signed. Customers are not behaving as if capacity is about to arrive.
The individual PO filings. The company committed in writing to disclose them when they cross the threshold, and they are the only way to verify a framework signed inside a US subsidiary. A first large PO before year-end would confirm the customer is moving.
The second marker is the FY2026 order figure against the raised $5,185m guidance, which appears with full-year results. Guidance that gets met or raised again means the backlog is still building faster than revenue, which at a 25% operating margin is exactly the condition that justifies an expensive multiple. Guidance that gets missed would be the first evidence that the queue for transformers is shortening.
kstock reads DART every morning and writes up what moved — the contract, the buyback, the number that does not add up. The daily post and a Saturday roundup, by email.