📌 X Insight Update[x_fin] (2026/06/23 21:00)

🧠 AI & Compute Infrastructure

SoftBank / Masayoshi Son vs. Elon Musk’s $SPCX orbital data center idea: the near-term AI race is still an Earth-based compute land grab. Launch, maintenance, and communication costs can eat up the theoretical power savings of orbital data centers. Speed matters more than elegance here: “He who strikes first wins.” 1

$TSM / TSMC is leaning into pricing power. A 5–10% price increase across advanced nodes, including 7nm, signals management wants to capture the same pricing upside memory players are enjoying. This is less about one-off pricing and more about foundry leadership monetizing tight supply and AI-driven demand. 2
DRAM looks structurally tight through the end of the decade. Deutsche Bank’s model has demand running above supply every year through 2030. The core squeeze is HBM, because it burns far more wafer capacity per bit than standard DRAM. As HBM rises from 18% to 35% of total capacity, it tightens both HBM and commodity DRAM at the same time. 3
$MU earnings are set up for a high-volatility event. After the last couple of days of massive swings, the print is unlikely to be a quiet one. 4

$WMT signing its first nuclear power agreement with $CEG for about 176MW from Constellation’s Dresden facility is another confirmation that large corporations are moving toward nuclear for reliable, long-duration, carbon-free power. The read-through is clear: AI-era power procurement is shifting from ESG optics to baseload reliability. 5

$ZETA + $PLTR is more than a headline partnership. $ZETA was up over 8% premarket and reclaimed $20 on the catalyst, but the bigger point is architectural. Zeta’s Data Cloud is being rebuilt on Palantir Foundry, putting the Athena AI decision layer directly on top of enterprise-grade ontology and governance. Management says the partnership could add over $100M in annual revenue in coming years, but the deeper bull case is stronger enterprise AI infrastructure, not just the revenue line. 6
SoFi Composer is not a fully agentic AI trader. The better read: it turns natural-language inputs into rules-based strategies, then lets users backtest before execution. That matters because the value is not “AI trades for you”; the value is strategy construction, historical testing, and understanding how a rule set behaves across market regimes before money goes live. 7

The Korean market’s crash needs context. KOSPI -10% sounds brutal, but it only took the index back to where it traded ten days ago. That makes the move ugly but not necessarily trend-breaking. The better read: violent reset, not automatic bear-market signal. 8
Korea’s selloff was not random tape noise. The index fell nearly 10%, marking the biggest one-day drop since March 4, and an intraday drop of more than 8% triggered a 20-minute trading halt. The key catalyst was tax reform discussion around taxing unrealized gains, which hit sentiment hard and turned an already fragile market into a forced de-risking session. 9

The tape is opening with real risk-off pressure: $QQQ was down nearly 3%, with $MU earnings still ahead this week. Dip-buying has worked so far, but this setup needs level discipline. No blind knife-catching. Let the bounce prove itself around key levels first. 10