When the same bidder wins every auction for five years, the room stops being an auction. The price is set not by competition but by the depth of a relationship, and nobody outside the room has tested it.
That is the structural question behind SK Horizon, the data center and submarine cable company being demerged from SK Broadband, SK Telecom’s wholly owned subsidiary, in South Korea. KKR is paying KRW 1.82 trillion (approx. USD 1.33 billion; all conversions at KRW 1,368 per USD as of Aug. 31, 2026) for 29%, the latest step in a five-year run of transactions with SK Group. An IMM Investment–Stonebridge consortium takes another 20%.
Read as news, the SK Horizon–KKR deal is a bet on Korean AI data centers. Read through the filings, it is narrower: investors pay a premium for a scarce asset, while the right to extend that scarcity, and the cash to fund it, stays with the seller.
The Deal
SK Telecom (SKT) is splitting SK Broadband through a demerger (a spin-off in which the shareholders of the existing company receive shares in the new one; here SKT is the sole shareholder). The new company, SK Horizon, inherits eight domestic data centers and the submarine cable business. The split ratio is approximately 0.84 to 0.16 on net book value.
- Capital: KKR pays KRW 1.82 trillion for 29%; the IMM Investment–Stonebridge consortium pays KRW 1.26 trillion for 20%; SKT keeps 51% and management control. Total consideration is KRW 3.08 trillion (approx. USD 2.25 billion).
- Price: KRW 76,839 per share, identical for existing and newly issued shares.
- Assets: data centers in Seocho, Ilsan (two sites), Bundang, Gasan, Centum, Yangju and Pangyo, plus new AI data centers under construction in Ulsan and Guro. Total infrastructure is stated at 318MW.
- Scale: latest-year revenue of KRW 347.7 billion, 7.7% of SK Broadband’s KRW 4.54 trillion; net book value of KRW 475.8 billion.
The deal does not stand alone. KKR bought SK E&S redeemable convertible preferred shares (RCPS, preferred equity the holder can redeem or convert) in November 2021 and December 2022; all KRW 3.135 trillion of that issuance came from KKR. In 2022 it was named preferred bidder for SK Materials Airplus’s industrial gas facilities.
In 2026, SK Group picked KKR as its renewables partner without a competitive process, pooling generation assets from SK Innovation, SK Ecoplant and SK Discovery. In July 2026, KKR took control of the resulting USD 1.3 billion platform, built to supply clean power to data centers and chipmakers. SK Horizon followed a month later.
A price set by bidding and a price set by a relationship can be the same number and still be different objects.
Structure Breakdown
The KRW 3.08 trillion is reported as one number; the filings show two. SKT sells 24,481,427 existing shares for KRW 1.8811 trillion, and SK Horizon issues 15,617,069 new shares for KRW 1.2 trillion. Same price per share, different destination: 38.9% of the money enters the company, and 61.1% goes to the parent.
SKT’s filing gives the purpose as funding growth businesses. A month earlier, its board had approved SK Hyper, a wholly owned AI data center developer receiving KRW 750 billion through 2030 and targeting 5GW by 2029 and 15GW by 2035. Cash leaving an asset in which investors own 49% becomes development capital for an asset in which they own nothing. Nothing about this is improper; the two filings were simply not read side by side.
Then the price. Pre-money, 66,216,598 shares at KRW 76,839 imply KRW 5.09 trillion of equity value, or KRW 6.29 trillion post-money: 10.7 times book and 14.6 times revenue on the pre-money figure. More than 90% of the price buys something absent from the balance sheet.
That something is grid connection in the Seoul capital region. Per local reports, 71% of data center applications target the region and more than half have been judged unserviceable, while inference workloads reward proximity to users. The sites SK Horizon inherits cannot be rebuilt today; the price is for the line into the substation, not the concrete.
But 318MW includes Ulsan and Guro, still under construction. Depreciation starts at completion while lease-up typically takes 24 to 36 months, an accounting loss window. On the desk’s estimates (KRW 10–13 billion per MW, roughly 120MW unfinished), remaining capex is KRW 1.2–1.56 trillion, almost exactly the primary proceeds. That is completion capital, not growth capital.
Post-money value is KRW 19.77 billion per MW, or an estimated KRW 24.18 billion (roughly USD 17.7 million) with remaining capex added. Blackstone’s AirTrunk, across five markets with 800MW contracted and over 1GW of owned land, reportedly traded at around 23.5x EBITDA. SK Horizon, single-country, largely unfinished and with its pipeline owned elsewhere, carries an estimated effective multiple of about 24x. That is a premium, not a discount.
Finally, the submarine cables. Bundling a permit-based, long-life, contracted business with colocation is about classification, not synergy: it turns an IT operator into digital infrastructure, which KKR signaled by booking the deal under its Asia Pacific infrastructure strategy. It is a variant of The Portfolio Paradox: two businesses with different physics priced through one multiple, used here less to blend quality than to change which kind of capital may buy the asset.
The Governing Logic
Read the sequence. KKR took control of the power platform first and entered the data centers a month later. In that order, this is a position on two layers of one value chain.
Power is the largest variable in data center margins. When prices rise, the data center loses and the generator gains; when they fall, the reverse. Holding both hedges the cost risk inside one fund’s book, visible only across the portfolio.
The hedge casts a shadow. If the two assets sign a long-term power contract, someone sets the price. KKR controls the generator while SKT holds 51% of the data centers, so the tariff would decide where profit lands, and SK Horizon’s minority holders would not be on the favorable side.
Then the cap. The 51/29/20 split stops investors at exactly 49%, leaving SKT the 51% it needs for consolidation and control; with a telecom-type cable asset inside, the line reads as drawn by accounting and regulation, not negotiation. Unlike The Sub-Majority Key, where the capped buyer takes control back as seniority, the disclosed terms show one price and no preferred return. The private equity route of control and restructuring is closed from day one.
For SKT, a partial sale without loss of control is generally an equity transaction: no disposal gain, a credit to capital surplus. The KRW 1.8811 trillion equals 14.52% of end-2025 consolidated equity, so one-seventh of equity was monetized without a P&L hit or loss of control. The price was 49% of future data center earnings.
Downside Protection
One explicit hedge is public: the new shares are issued sequentially. Each tranche can carry conditions such as megawatts leased, completion or permits. Control the pace of capital, and the counterparty’s pace of delivery follows.
It covers only 38.9% of the money. The secondary 61.1% goes to SKT unconditionally, with recourse limited to representations and warranties (R&W), special indemnities and escrow. R&W insurance excludes known risks, and grid shortfall, permit delay and local opposition are all known.
The relationship cuts both ways: a counterparty you will deal with again is one you rarely sue. The defense likely rests more on relationship than on documents, collateral that cannot be quantified.
The exits are narrow. An IPO with SKT at 51% meets aversion to duplicate listings; an infrastructure fund or REIT pays a low multiple; an SKT buyback lets the buyer set the price; a global strategic faces licensing and foreign ownership limits. Exit rests on a shareholders’ agreement that has not been disclosed.
What Must Be True
Three propositions carry this read. Each comes with the observation that would falsify it.
1. Capital-region grid scarcity must survive the lease-up of Ulsan and Guro. SK Horizon’s largest potential competitor is SK Hyper, whose 5GW target is more than fifteen times its 318MW, partly funded by the investors’ own secondary proceeds. Falsifier: SK Hyper’s first-phase capacity reaching the market in the window when Ulsan and Guro must fill. On the desk’s sensitivity, an 18% rent decline cuts EBITDA by about 30% and equity value by 40%, and by more than 60% if multiples compress, because both moves stem from the same supply wave.
2. Power cost must remain a policy advantage. Korean tariffs still under-reflect cost, and the government is reportedly reviewing discounts for AI data centers. Falsifier: Korea adopting a minimum-billing rule like those in US states such as Ohio, which reportedly bill new data centers above 25MW for at least 85% of contracted capacity regardless of use. The desk estimates that a 20% tariff rise, with half of leases unable to pass it through, cuts EBITDA margin by 5–8 points and enterprise value by KRW 1.0–1.6 trillion.
3. The chain must stay parallel, not serial. Lease-up losses consolidate into SKT, which has paid KRW 830 per share for two consecutive quarters while funding AI expansion. Falsifier: a dividend cut, which runs through share price and group funding to an SK Hyper delay, and with it the investors’ last upside of one day absorbing that pipeline.
The Architect’s Takeaway
What worked here was not the size of the capital but a path built by repeated transactions and a few clauses placed on it. Capital cannot be copied; paths and clauses can. If you were building your own deal room:
- Repeat small deals with the same counterparty. Make each engagement accumulate as history; by the fourth, the counterparty calls you before opening an auction.
- Secure the adjacent layer first. Hold the step just before or after your own, where cost or demand is set: distribution, data sources, the room where requirements are defined.
- Release your scarcest resource in tranches. Split time, labor or reputation into stages, each gated by a condition the counterparty must meet.
- Ask where the money lands, not how large it is. Company results, a manager’s reputation and your own portfolio are three different accounts.
- Trade control for veto. Related-party approvals, budgets and exit timing: hold those three and 51% becomes unnecessary.
- Write the exit door into the first document. Defer the exit terms, and years later you discover you cannot leave.
One cost travels with the method. Long relationships open doors and dull clauses; in a room without competition, neither the price nor the terms get tested.
Syndicate’s Note
The asset is real. The capital-region sites cannot be rebuilt, and grid access is the scarcest input in Korean digital infrastructure. But the right to extend that scarcity stayed with the parent, while the risk of its erosion passed to the investors, and their own secondary proceeds help finance the vehicle most likely to erode it.
Entry was built by repetition and defense by clauses, yet the clauses visible to the public protect less than two-fifths of the money. Ownership of an asset and ownership of its growth were separated quietly inside a contract.
The desk files the mechanism as a candidate, The Growth Carve-Back: a controlled minority sale in which investors buy the scarce asset while the seller keeps, and funds with their money, the pipeline that will compete with it. One case does not make a pattern; it enters the declassification queue pending a second.
House Verdict: TRACK. The verdict classifies structural learning value, not investment merit. This read will be tested in public: by whatever is disclosed of the conditions on each primary tranche, by the timing of SK Hyper’s first-phase capacity against the Ulsan and Guro lease-up, and by any power contract between KKR’s platform and SK Horizon. Falsifier: if SK Hyper’s capacity is contributed to, or offered first to, SK Horizon, the growth was never carved back and this read is wrong.
FAQ
What is a proprietary deal in private equity?
A deal negotiated bilaterally, without a competitive auction, by a buyer with privileged access to the seller. The seller gains speed and certainty, the buyer uncontested entry; the cost is a price no market has validated.
How is the SK Horizon investment by KKR and IMM structured?
Investors pay KRW 3.08 trillion for 49% at KRW 76,839 per share. KRW 1.8811 trillion buys existing shares from SK Telecom and KRW 1.2 trillion buys new shares issued sequentially, so 61.1% goes to the parent and 38.9% to the company.
Why is the investor stake in SK Horizon capped at 49%?
Fifty-one percent is the minimum for SK Telecom to keep consolidation and control, and the telecom-type cable asset reinforces that line. Investors have no path to control, and exit depends on an undisclosed shareholders’ agreement.
This article is for informational purposes only and does not constitute investment advice, an offer, or a solicitation. Figures are based on public reports and may be estimates.
For more structural insights and deep-dive video breakdowns, visit Structure Syndicate on YouTube.