#TemasekInvestsSKHynixJapanPlant

Temasek and SK Hynix's Japan plant. Few headlines this year carry more weight for the AI memory story, and the market is reading them the way long term investors should: sovereign capital and manufacturing capacity moving together into the same shortage. It is a confidence signal built on three legs, memory leverage on AI demand, a running shareholder return programme, and the world's leading supplier of high bandwidth memory.

Two August 2026 reports built that story. On August 12, Korea's Asia Business Daily reported that Singapore's state investor Temasek was preparing a direct investment in Samsung Electronics and SK Hynix through its own internal team. On August 21, Hankyoreh, picked up by Reuters, reported that SK Hynix is considering a memory chip fab in Miyagi prefecture in northeastern Japan, with investment potentially reaching tens of trillions of won, after SK Group chairman Chey Tae-won visited the region, which Tokyo has designated a semiconductor hub. If it proceeds, it would be the first large scale semiconductor manufacturing investment in Japan by a Korean chipmaker, exactly the supply diversification the AI build out needs. The company says the matter is still under review and has flagged September 18, 2026 for an update, which gives the story a dated catalyst rather than an open ended hope.

The first reaction shows how badly investors wanted it. On August 12, SK Hynix closed up about 5.5 percent at roughly 1,504,000 won, Samsung finished up 6.7 percent, the KOSPI added 3.7 percent, and both chipmakers briefly gained more than 8 percent intraday. On August 21, the Miyagi report lifted SK Hynix 2.3 percent to 1,730,000 won and it outperformed a KOSPI that closed 0.67 percent lower. Temasek has since said it first invested in both companies more than two years ago, capital that has already stayed through a cycle.

Now the level that anchors everything. SK Hynix closed September 15 at 1,690,000 Korean won, about 1,250 dollars per share in dollar terms, with a market value near 1,234.5 trillion won, or roughly 918 billion dollars, across 730.49 million shares. From that 1,250 dollar area, the arithmetic favours patience. A return to the September 9 zone near 1,380 dollars is about 10 percent higher, the September 10 intraday high of 1,890,000 won is about 1,405 dollars or 11.8 percent above, and the June 2026 high of 2,987,000 won, which is around 2,221 dollars, is roughly 76.7 percent higher. The 52 week range runs from 325,000 won, about 242 dollars, to that 2,987,000 won peak, with the shares still up about 150 percent year to date and about 379 percent over twelve months.

The recent tape shows how much real money moves through this name even during a shakeout. September 7 jumped 8.26 percent to 1,783,000 won on 3.47 million shares and about 6.10 trillion won of turnover, after reports that memory inventories had fallen below ten days. September 8 added 0.56 percent to 1,793,000 won on 4.12 million shares and 7.58 trillion won, the heaviest volume of the run. September 9 gained 3.51 percent to 1,856,000 won on 3.52 million shares and 6.51 trillion won. September 10 was flat, down 0.16 percent to 1,853,000 won, but printed the largest turnover of the stretch at 8.09 trillion won on 4.37 million shares. September 11 eased 2.21 percent to 1,812,000 won on 2.62 million shares and 4.69 trillion won. September 14 slipped 6.35 percent to 1,697,000 won between 1,740,000 and 1,686,000 on 3.74 million shares and 6.41 trillion won. September 15 closed 0.41 percent lower at 1,690,000 won, with a high of 1,729,000 and a low of 1,671,000, on 2.55 million shares and 4.33 trillion won.

Two numbers there should encourage anyone watching liquidity. Average volume is about 3.46 million shares a day over twenty sessions and 3.83 million over thirty one, so positions of size can be built and exited freely, with daily cash turnover between 4.33 trillion won and 8.09 trillion won in the past week alone. September 15 traded only 74 percent of the twenty day average, which usually means the sellers who wanted out are gone. A pullback on shrinking volume is a healthier picture than the headline suggests.

Support underneath is unusually visible, from three sources. First, the buyback: on August 19 the board approved a 40 trillion won, roughly 28.6 billion dollar repurchase and full cancellation, about 24.07 million shares or 3.3 percent of the issued total, running from August 20 to November 19, the largest treasury share cancellation in Korean listed history, and a return target above 50 percent of cumulative 2025 to 2027 free cash flow. Samsung's own programme lifts combined commitments close to 55 trillion won. Second, the sovereign bid, with Temasek reported to be moving in directly and foreign ownership still near 50.47 percent. Third, the passive bid created by index weight: the Bank of Korea notes the two chipmakers are about 51.2 percent of KOSPI market capitalisation, and the BIS flagged that they exceeded half of Korean equity trading value by mid 2026, up from 12 percent in 2025. Any fund tracking Korea has to own this in size.

The Japan plan also makes industrial sense, which is why the market is treating it as a positive rather than a cost. Tokyo's METI has allocated about 1.23 trillion yen for semiconductors and AI in its 2026 budget, Rapidus is receiving more than 1 trillion yen, and TSMC's Kumamoto venture has drawn up to 476 billion yen, so a Miyagi project would sit inside a subsidy framework that lowers the net cost of capacity. Micron already runs a DRAM plant in Hiroshima, proving the ecosystem works. The site also puts SK Hynix closer to Japanese suppliers and customers, and diversifies production away from Korea and China, what AI customers want after three years of supply shocks. It layers on top of a 54 trillion won Yongin and Cheongju plan, a Dalian NAND restart and a packaging site in Indiana above 4 billion dollars, all aimed at the same demand wave.

The financial engine is still firing. In the second quarter of 2026, revenue reached 79.3187 trillion won, up 51 percent quarter on quarter and 257 percent year on year, and operating profit hit 60.5426 trillion won, up 61 percent and 557 percent, a 76 percent operating margin. EBITDA was 64.56 trillion won, cash stood at 88 trillion won against 18.6 trillion won of debt, net cash near 69.4 trillion won, and DRAM average selling prices rose about 30 percent quarter on quarter. Net profit of 93.9226 trillion won was flattered by roughly 62.2 trillion won of non-operating income, so operating profit is the number to hold onto. Revenue and operating profit landed just under the most aggressive consensus, a rounding issue against the scale of the cash generation, not a change in direction.

Memory pricing is the real bull story here. HBM3E contract prices were raised about 20 percent for 2026, HBM4 stacks are expected to price between 500 and 600 dollars per unit, roughly 55 to 70 percent above the previous generation, DRAM makers have been pushing about 20 percent increases into the third quarter, and inventories have reportedly collapsed below ten days, which is a genuine shortage rather than an inventory cycle. The HBM market is projected to grow from about 35 billion dollars in 2025 to 54.6 billion in 2026 and toward 100 billion by 2028. SK Hynix enters that window as the leader.

Valuation is where the opportunity is clearest. At that 1,250 dollar equivalent, with first half equity of 262.69 trillion won, book value is roughly 359,000 won per share, so the stock trades near 4.7 times book. Annualising second quarter operating profit gives about 242 trillion won, so the company sits near 5.1 times that, with EV/EBITDA around 4.5 times on net cash and price to sales near 3.9 times annualised revenue. The quoted trailing ratio of about 7.4 times earnings per share of 227,525 won understates the multiple because of one off gains. Micron at about 1,016 dollars screens near 11 to 12 times current fiscal year earnings, while Samsung trades near 5.4 times 2026 earnings at 248,500 won. Published targets cluster between 2.8 million and 3.2 million won, roughly 2,080 to 2,380 dollars per share, well above the current level.

My own view is constructive. The Temasek report and the Miyagi plan are two halves of the same message: the capital and the capacity are both being pointed at the same shortage, and both are arriving as the cycle tightens rather than eases. Add a buyback that removes 3.3 percent of the shares outstanding, a balance sheet with nearly 70 trillion won of net cash, a leadership position in the product that AI data centres cannot build without, and a valuation in the mid single digit multiples of cash earnings, and the risk of missing the move looks larger than the risk of a wobble. The September 14 selloff was not about this company; it was a global AI sentiment reset that took SoftBank down 10.72 percent and the KOSPI down 3.26 percent the same day, and SK Hynix stabilised the next session.

The catalysts are lined up and dated, which turns a story into a path. September 18 is the company's update deadline on the Japan review, and a confirmed plan with scale, timeline and subsidies would be the cleanest possible news. September 30 brings Micron's results, the first hard read on memory pricing for the complex, followed by Samsung's preliminary figures in early October, SK Hynix's third quarter report in late October and the buyback running to November 19. If memory pricing keeps climbing and the buyback keeps absorbing supply, the path back toward the September highs near 1,380 to 1,405 dollars, and eventually toward the June peak near 2,221 dollars, stops looking ambitious. Worth watching: capital spending discipline, a stronger won and the 2027 to 2028 supply question, the normal risks of owning the leader in a boom. From 1,250 dollars, with a shortage underneath it, the case for patience is stronger than the case for doubt.
#GateSquareMidAutumnReunion