Thursday morning set-up. Yesterday the market stopped treating “AI” as one trade. Here is what split, why it matters, and the one number that decides my next move today.
What split yesterday
- NVIDIA (NVDA) fell 2.9% to about $151, its fifth loss in a row and a 17-month low, even after a strong quarter.
- Broadcom (AVGO) rose about 3.6% to about $354, its second up day, closing near the high.
- AMD (AMD) gained about 5%.
- Palo Alto Networks (PANW) jumped 9.9% after a beat-and-raise, and the cyber security group rallied with it.
- Overnight, Taiwan’s Taiex fell about 4% and TSMC about 3%, while Korea was roughly flat.
Same theme, opposite days. The market is no longer paying for “AI exposure”. It is paying for a specific layer, and asking who funds the demand behind it.
Why I think this is healthy
For weeks the whole complex sold off together: chips, memory, optics, power. When everything falls at once, nothing is being priced, it is being liquidated.
Yesterday was different. Money moved toward the custom-chip and networking supplier and the alternative GPU vendor, and away from the incumbent whose leadership and financing commitments are now open questions. The foundry in the middle was sold with the incumbent.
Meanwhile the fundamentals did not move: memory makers say most of next year’s supply is already contracted, and order books at the tool makers stretch into 2028. When fundamentals hold and prices split, that is a market starting to discriminate again. That is when a written plan beats a gut feeling.
The one number that matters today
My rule for adding to AI names is tied to the price of money, not to any single company’s results.
- If the US 10-year yield closes today below 5.10%, the rule allows one AI addition from Friday.
- Between 5.10% and 5.20%, nothing changes and I re-read on Friday.
- Above 5.20%, the pause stays and I lean toward a “higher for longer” scenario.
Yesterday’s close was around 5.07% to 5.10%, depending on the source. Interesting detail: the Fed minutes said almost all members expect more hikes, and long yields still fell. Bond markets are starting to price growth risk, not only inflation.
What this means for the portfolio
No trades overnight. Cash stays about 38% of the portfolio, by design: it is dry powder, and it earns interest while it waits.
- Broadcom (AVGO) is first in line. It is the only AI name with all three parts of my entry rule: intact fundamentals, a long-term support level (back above its 100-week average, about $314), and a daily turn that has already printed. If the rate check passes, one standard position from Friday, after the opening auction. The plan is wrong if it closes a day below the low of the signal day (about $341).
- TSMC (TSM): no action on the Taiwan drop. It is still well above its long-term support zone. I act on a daily close back above about $386, or on a daily reversal in the $295 to $315 zone.
- China large caps (FXI): the trailing stop sits about 1% below price, and Hong Kong fell about 2% overnight. If it fires, the line goes back to my watchlist with the re-entry rule already written. A stop that fires is the system working, not failing.
A blind spot I am opening tonight
The book has no cyber security exposure. Yesterday’s move does not make me chase a gap, but the thesis deserves a proper look: if AI makes attacks cheaper and faster, security becomes spending that companies cannot postpone. Tonight I research which layer actually gets paid, and at what valuation.
Today (Bucharest time)
- Before the US open: Walmart results, a read on the US consumer
- 15:30 US jobless claims and the Philadelphia Fed survey
- 16:30 US open, then the 10-year close that decides the rule above
- Friday: the weekly close that confirms or cancels every signal
Risks, stated openly
- A December hike of 50 basis points instead of 25 would push yields higher and keep the pause in place.
- The dollar is at a 19-month high. That weighs on metals, emerging markets and China, where the portfolio has exposure.
- Splits like yesterday’s can reverse quickly. One day of dispersion is a signal to watch, not a trend.
The process
I do not need to know which AI company wins. I need a rule that tells me when the price of money allows me to act, a zone that tells me where, and a daily signal that tells me when. Everything else is patience. The plan for today was written before the open.
Snapshot: portfolio data as of 19 November 2026, 07:51 UTC; market data from the 18 November US close and the 19 November Asian session.
Sources: Motley Fool, market close 18 Nov · FXStreet, commodities feed · Morningstar, ASML bookings into 2027
Disclaimer: This is not financial advice. Copy trading carries real risk of loss, and past performance does not guarantee future results. Always invest only what you can afford to lose.

