⚠ The Equipment Is Sold to EverybodyModerate threat
SK hynix (SKHY) — threat to the moat
Nothing in a fab is proprietary to its owner -- the same suppliers will sell the same tools to whoever is next in line.
A manufacturing advantage in semiconductors is unusual in that the means of production are not proprietary. The lithography, deposition, etch and bonding tools that SK hynix uses are bought from the same small group of suppliers who will happily sell the identical machines to Samsung, Micron and anyone else with the capital.
The roughly ₩15 trillion committed to M15X and ₩19 trillion to the P&T7 packaging plant1 buy machines any of the three producers can order. What cannot be bought is the recipe: the process integration, the tuning, the accumulated knowledge of which of ten thousand variables matters. That is genuinely hard and genuinely takes years. But it is knowledge held by people, and people move, and the physics is published.
This bounds how durable a yield lead can be. It is a rolling advantage that must be re-won at every node and every stack height, not a structural barrier. The industry's history is full of manufacturers who were decisively ahead and then decisively behind within one generation — including, on HBM, the competitor SK hynix took the lead from.
There is a second-order dependency worth naming: the tool suppliers are themselves a bottleneck, and in a boom the queue for equipment is the real constraint on adding capacity. That protects the incumbent's near-term position by slowing everybody down, and it equally means SK hynix cannot expand as fast as demand when it wants to.
The metric that tests this page is the cost per bit gap against the other two, disclosed by nobody and inferable from gross margins. If SK hynix's margin advantage in a comparable quarter is not visible, the manufacturing edge is a story rather than a number.
- ReportedAnnounced investment of approximately ₩15 trillion for M15X and approximately ₩19 trillion for the P&T7 advanced packaging plant.SK hynix Inc., semi-annual business report for the six months to 30 June 2026 (SEC Form 6-K, 18 August 2026) — 'The Company determines supply volumes and prices on a monthly and quarterly basis by mutual agreement with major customers, and there is no order backlog based on long term supply contracts.' Total equity W 262,693,228 million at 30 June 2026 against W 120,666,751 million at 31 December 2025; cash and cash equivalents, short-term financial instruments and short-term investment assets W 87,957,923 million against W 34,942,253 million; total borrowings W 18,586,634 million against W 22,247,905 million; debt-to-equity 32.80%, net borrowing ratio not disclosed because it is negative. For the six months to 30 June 2026, revenues of W 17,608,702 million (13.35%) and W 17,187,421 million (13.03%) were derived from external Customers A and B respectively; for the six months to 30 June 2025, W 10,890,639 million (27.31%) came from Customer A. IDC market share: DRAM 29.9% in 2023, 33.4% in 2024, 34.8% in 2025 and 29.1% in Q1 2026; NAND 19.6%, 21.4%, 20.9% and 18.5%. Gartner's world DRAM market: US$78.7bn in 2022 (-15.4%), US$49.9bn in 2023 (-36.5%), US$91.6bn in 2024 (+82.7%) and US$146.9bn in 2025 (+60.0%). Announced investments include approximately W 15 trillion for M15X in Cheongju, approximately W 19 trillion for the P&T7 advanced packaging plant and W 21,608.1 billion for Yongin, against long-term plans of roughly W 600 trillion for the Yongin cluster and W 100 trillion for Cheongju. The company developed the world's first 321-layer QLC NAND. China's State Administration for Market Regulation cleared the Intel NAND acquisition conditionally, obliging the group to maintain a reasonable pricing policy, increase production and support the entry of third-party competitors into the Chinese eSSD market for five years from December 2021. — six months to 30 June 2026 · publ. 2026-08-18 · source ↗