📌 X Insight Update[x_fin] (2026/07/19 23:01)

📈 Original Insights Summary

TSMC raising CY26 capex to $60-$64B from $52-$56B is not just a capacity story. Tool inflation is becoming a real driver, and that should feed directly into stronger semicap gross margins. ASML already flagged room for price hikes, while AMAT has been pushing value pricing for a while. 1
ASML still has elite pricing power. A single $ASML EUV machine now costs about ~$284M. Average EUV pricing compounded around 13% annually for the last decade, then surged another 45% in a single year with no order cancellations. That is not normal cyclicality. That is monopoly-like leverage. 2
Semi valuation dispersion is getting stretched. PEG < 1 screens as mispriced growth, while PEG > 2 starts entering the danger zone. Current stack: $LRCX ~2.3x, $ALAB ~2.1x, $INTC ~1.9x, $ANET ~1.8x, $ARM ~1.6x, $KLAC ~1.3x, $AMAT ~1.2x, $ASML ~1.0x, $TSM ~0.9x. On this lens, $TSM looks cheapest while $LRCX and $ALAB look priced for perfection. 3
The $NBIS bull case is a hard infrastructure ramp, not a soft AI narrative. The path points to $30B revenue and $21B EBITDA by 2030, with the ramp starting in 4Q26 as Vineland comes online. The setup is backed by $46B of signed $MSFT and $META deals, including a $27B Meta agreement with $12B fixed. 4

$SPX still has an upside-first path before the rug pull. The roadmap: 7,455 → 7,700, then 7,700 → 6,545-6,930, then 6,545-6,930 → 8,000. Translation: possible pump into around August, sharp 10-15% drawdown around Sept-Oct, then a monster rebound leg. 5

$AAPL has opened an extreme lead over $MSFT. Apple is now 67% larger, about $2T ahead in market cap, equal to roughly two Berkshire Hathaways. The spread between two all-time great compounders looks historically stretched in absolute terms. 6

$ETH has a clean trigger level. Getting back over $2,000 would start to make the chart interesting again. Until then, it is still just a watchlist setup, not a confirmed move. 7

The DeepSeek debate is not as simple as “open-source weights equal open moat erosion.” After reviewing expert-call transcripts, the counterpoint is that Dean Ball’s argument looks partly wrong, with DeepSeek used as the key example. Open weights alone do not fully explain competitive dynamics in frontier AI. 8