SK hynix in Miyagi: Japan Has the Site, Korea Has the Stamp
A routine Korea-Japan chamber of commerce meeting in Sendai drew an unusual crowd on August 31. The Korean chamber's chairman is Chey Tae-won, who also chairs SK Group, and ten days earlier Hankyoreh had reported that SK hynix was weighing a memory fab in Miyagi prefecture. Nikkei's same-day report laid out the setup: Miyagi has offered a 300,000 square meter site and has been in quiet contact, but the biggest obstacle is Seoul's concern about technology leakage.
This post uses the Nikkei piece as a starting point to map how three governments see one fab differently, and checks where SK hynix's ability to fund it comes from, using the Q2 results Nikkei's own chart was pointing to. Chairman remarks, governor remarks and regulatory filings are kept separate.
KEY TAKEAWAYS
1. Miyagi has offered a 300,000 m² site in the Second Sendai North industrial park with water and power in place (Nikkei). In Sendai on August 31, Chey said SK is "reviewing various candidate sites for a joint-venture plant," the first time he has spoken of building one directly (Bloomberg, Korea Economic Daily).
2. The obstacle is Seoul. Advanced semiconductors are designated national core technology and an overseas fab needs government approval; the industry's common view, per Nikkei, is that AI-grade production abroad will not be approved. The US Commerce Secretary said in July he wants Samsung and SK building fabs in the US.
3. Funding is not the constraint. Q2 revenue was 79.3 trillion won, operating profit 60.5 trillion (76% margin), net profit 93.9 trillion including a 63.3 trillion gain on the Kioxia stake. Cash rose 33.6 trillion in one quarter to 88 trillion; net cash is 69.4 trillion.
Ten days of statements, by source
| Date | Who | What | Type |
|---|---|---|---|
| Aug 20-21 | Hankyoreh, Bloomberg | SK hynix weighing a multi-trillion-won fab in Miyagi; Chey visited the site | Anonymous sources |
| Aug 21 | SK hynix | "Reviewing various options including additional production bases; nothing decided" | Regulatory filing |
| Aug 26 | Governor Murai | "We have approached various chip firms; the prefecture knows nothing about this case" | Press conference |
| Aug 27 | SK hynix | Groundbreaking for Indiana HBM packaging plant (~$4B) | Official event |
| Aug 31 | Nikkei | Miyagi offered 300,000 m²; Korean approval is the main barrier | Report |
| Aug 31 | Chey Tae-won | "Reviewing candidate sites for a JV plant"; "will announce when discussions end" | Direct remarks |
The story moved up a level. Ten days ago it was anonymous sourcing and a filing that denied any decision. On the 31st the chairman named a structure (joint venture) and a timeline (an announcement). Partner, product, investment size and start date remain undisclosed, and the company's official position is still "under review." This post treats it as a plan under review, not a project.
Three governments, as Nikkei framed them
Japan has the site. Miyagi is one of three national chip hubs, with Kyushu (TSMC Kumamoto) and Hokkaido (Rapidus), backed by about 130.6 billion yen in government funding. The governor says there is no better place in the world. If built, SK hynix would be the third foreign chipmaker with a fab in Japan, after TSMC and Micron (Hiroshima).
The US has the leverage. Federal and local officials turned out for the Indiana groundbreaking on August 27, and SK framed it as a new chapter in US-Korea AI cooperation. Commerce Secretary Lutnick said in July he wants to bring Samsung and SK to build in America. Indiana is packaging, not a front-end fab, which is why it is about $4 billion. What Washington wants is the next step.
Korea has the approval. Advanced semiconductors are national core technology under the Industrial Technology Protection Act, and a new overseas plant requires government approval. Nikkei relays the industry's common view that AI-grade production abroad will not be licensed. No official Korean government position has been confirmed.
What Nikkei's chart was saying: the quarter net profit passed revenue
The chart in the Nikkei piece was headed "rapid earnings growth secures ample investment funds," with a footnote that Q2 net profit exceeded revenue because of gains on shareholdings. In numbers: revenue 79.3187 trillion won, operating profit 60.5426 trillion (76% margin), net profit 93.9226 trillion (118% net margin). Non-operating income of 62.17 trillion included 63.27 trillion of gains on investment assets, which is the Kioxia stake SK hynix holds indirectly through Bain's SPC convertible bonds, bought for about 3.9 trillion in 2018. The Bell estimates roughly 50 trillion of that is unrealized.
The number to read for investment capacity is cash, not net profit. Cash and equivalents stood at 88 trillion won at end-June, up 33.6 trillion in the quarter, with net cash of 69.4 trillion. A fab in the tens of trillions, the size Hankyoreh reported, is fundable from the balance sheet on arithmetic alone. But Yongin phase 1 (targeting end-2027) and the Honam cluster are already booked at home in the hundreds of trillions.
One caution: more than a third of that cash appeared in a single quarter, and the Kioxia gain can reverse in Q3 depending on Kioxia's share price (The Bell). Eighty-eight trillion is a peak-of-cycle number.
Three reasons SK hynix wants Japan
First, the length of the shortage. Chey said at GTC 2026 that the memory shortage runs to 2030, and as the UBS supply-demand chart covered earlier shows, 2027 is drawn as the year consumption passes shipments. If domestic new fabs cannot carry 2027-2030 demand, overseas capacity is needed.
Second, Japan's equipment and materials ecosystem. Chey told Nikkei in June that Japan is a fully capable candidate because its semiconductor ecosystem is in place. Miyagi sits near Tokyo Electron's home base in Iwate and Miyagi, and near Kioxia's Kitakami fab. SK hynix is Kioxia's largest indirect shareholder, which is one way to read the word "joint venture," though no partner has been named.
Third, geographic diversification. As Nikkei notes, having production concentrated in Korea carries peninsula-contingency and natural-disaster risk, and single-country dependence is a mark against a supplier in customer supply-chain audits. Indiana diversifies packaging; Japan would be the first front-end candidate.
What it makes decides whether it is approved
Korea's approval barrier turns on what the fab produces, not on whether there is a fab. National core technology designations cover DRAM and NAND design and process technology below specific generations, so the difficulty depends on the product mix.
Leading-edge DRAM (1c, HBM core dies) faces the high wall the industry describes. One-generation-older DRAM would be judged case by case, with no precedent. NAND in a joint venture with Kioxia changes the direction of technology transfer and therefore the picture. HBM back-end and packaging already has an approval precedent in Indiana. "A multi-trillion-won JV fab" does not tell you which of the four this is, and until that is settled, neither is approval.
The US variable overlaps. Hankyoreh's read was that a concrete Japan investment could intensify US demands for a front-end fab. For SK hynix, a Japan fab could either accelerate the US ask or serve as an "already building abroad" argument.
What I actually watch
| When | What | Why |
|---|---|---|
| Weeks | The announcement Chey flagged: partner, product, size | Which of the four scenarios sets the approval difficulty |
| After that | MOTIE national-core-technology approval process | First official Korean government position |
| Late Oct | SK hynix Q3: direction of Kioxia mark-to-market, cash balance | Whether 88 trillion holds or the gain reverses |
| Ongoing | US Commerce specifics on Samsung and SK front-end investment | Sequencing of Japan versus US |
Value chain read-through
| Segment | Direction | Evidence |
|---|---|---|
| SK hynix | First overseas front-end attempt; approval, US pressure and investment sequencing as three variables | Chairman remarks, filing, Nikkei |
| Japanese equipment and materials | Direct beneficiary if Miyagi proceeds (Tokyo Electron nearby) | Chey's "Japan ecosystem" remark |
| Kioxia | Expanded NAND cooperation if named as partner; unconfirmed | SK hynix's largest-indirect-shareholder position |
| Korean back-end and tools | Need to confirm Yongin and Honam priority holds | Hankyoreh: domestic plans proceed, overseas is additional |
Risks to this view
- Review stage. Partner, product, size and timing all undisclosed. The company's position is "nothing decided"; the governor says the prefecture knows nothing.
- Approval risk. If it is leading-edge DRAM, the common industry view is that Seoul will not approve. No official position yet.
- Nature of the cash. A large share of the 88 trillion arrived in one quarter, and about 50 trillion of the Kioxia gain is unrealized and can reverse in Q3.
- The cycle. Nikkei itself writes that the memory outlook is hard to read and the durability of the AI boom is the big variable. An overseas fab built with peak-cycle cash comes online in the next downturn.
Bottom line
Japan has land and infrastructure, the US has demands, Korea has the stamp. SK hynix has 88 trillion won and must decide what to build with it and where. In the announcement Chey has promised, the thing to look for is not the investment figure but the product. Once that is set, the other three governments' responses follow.
Sources: Nikkei, "SK chip plant: Miyagi eager to host, Seoul wary over technology leakage," Aug 31, 2026 / Kyunghyang, Etoday, Seoul Economic Daily, Aju Business Daily, Aug 31, 2026 (citing Nikkei and the governor's press conference) / Money Today, Korea Economic Daily, Aug 31, 2026 (Chey remarks via Bloomberg) / Global Economic, Aug 21, 2026 (citing Hankyoreh) / SK hynix Newsroom, Q2 2026 results, Jul 29, 2026 / Yonhap Infomax and The Bell, Jul 29, 2026 (non-operating and cash detail) / The Elec, Q2 earnings call transcript, Jul 30, 2026. Everything here is from public sources.
Comments
Post a Comment