$SKHY $SSNLF Anthropic’s IPO Push and Altman’s Return to Korea Anthropic publicly filed its S-1 five days before Sam Altman returned to Korea. SK hynix had just disclosed long-term agreements covering more than 50% of its 2027–2028 production capacity. That puts an important question alongside the AI fundraising race: how much future memory supply can the money actually secure? 1. What happened in Korea On October 31, Altman met about 50 Samsung executives in Yongin. Samsung’s account, reported by Yonhap, described a discussion of AI technology, business opportunities and corporate transformation. The company also supplied photographs of the event. Edaily separately reported that Altman had lunch with SK Chairman Chey Tae-won, joined by senior SK hynix and SK Telecom executives. Its report explicitly said the details of the discussion were unknown. Those reports have not disclosed a new memory order, a reservation of additional capacity or agreed pricing. There is a substantial existing relationship. On October 1, 2025, OpenAI announced Stargate partnerships with Samsung and SK, including a target of 900,000 DRAM wafer starts per month. That was a capacity ambition, rather than a disclosure of delivered volume. Anthropic’s October 26 announcement, meanwhile, concerned a proposed Nasdaq IPO under ANT. The offering’s share count and price range were still unset in that announcement. There is no evidence that its IPO filing prompted Altman’s visit. 2. Why the capacity matters SK hynix’s October 28 release provides the stronger investment evidence: 14 major customers have signed long-term agreements, covering more than half of total production capacity for 2027 and 2028. Its pricing mechanism allows adjustments reflecting market conditions within an agreed range. The release does not identify an allocation to OpenAI or Anthropic. My interpretation is that AI expansion increasingly needs financing and supply commitments to advance together. Additional funding can support larger infrastructure plans, while actual deployment still depends on memory arriving in the required volumes and on schedule. Advance commitments can improve a memory supplier’s demand visibility and its basis for investing. The eventual earnings benefit still depends on contract prices, production costs and the capacity brought online. That is why I would read Altman’s return against the existing Stargate relationship and the latest capacity commitments. A CEO meeting can indicate strategic attention; the investment case becomes stronger when quantities, delivery dates and commercial terms follow. For Samsung and SK hynix, the next useful disclosure is how those relationships translate into contracted deliveries. For the AI labs, the question is how much deployable infrastructure their capital can buy, and when.