# AI Stocks Now Trade the Growth Rate of Spending, Not Its Level

*URL:* https://millionaur.com/millionaur-2026-11-21-daily-analysis/
*Updated:* 2026-11-21T07:56:28+00:00

Every Saturday I look back: what the week did to the portfolio, which written rules fired, what I got wrong, and what history says about the cycle. Earlier posts: [Friday evening’s update](https://millionaur.com/millionaur-2026-11-20-evening-update/) and [Friday’s plan on eToro](https://www.etoro.com/posts/f4f520f0-e517-11f1-8080-800171484502).

## The week in numbers

- S&P 500 −8.6%, Nasdaq-100 −10.4%, the semiconductor index −17%. Friday alone: S&P −3.7%, chips −10.7%, the VIX near 36.
- Taiwan −17.5% on the week, Korea −8%, Japan −7.7%. India +2.6%.
- Gold −9.5%, silver −16%, bitcoin around −13%.
- The US 10-year yield fell from 5.15% to 4.91%: its first weekly close below 5% this month.
- Millionaur: −5.7% on the week, with 40% of the portfolio in cash.

The cushion came from cash, Treasuries ([$IEF](https://finance.yahoo.com/q?s=IEF) +1.8%, [$TLT](https://finance.yahoo.com/q?s=TLT) +2.6%) and India. Gold and silver did not cushion anything. In a forced-selling week, hard assets behave like collateral, not like hedges.

## What my rules did

- **[$AVGO](https://finance.yahoo.com/q?s=AVGO):** my entry needed a trade above 345 in the US afternoon. It never came. Friday it fell 12% after an analyst report that one large customer trimmed its 2027 chip plan by about 20%. The rule cancelled itself before the drop.
- **[$INDA](https://finance.yahoo.com/q?s=INDA):** the India ETF reclaimed its long-term average while everything else fell. I added one standard clip on Friday. It closed flat on a −3.7% day.
- **Weekly-close rules:** Argentina ([$ARGT](https://finance.yahoo.com/q?s=ARGT)) cut by one clip. Silver ([$SLV](https://finance.yahoo.com/q?s=SLV)) confirmed its cut line; one clip goes Monday. The commodity basket ([$DJP](https://finance.yahoo.com/q?s=DJP)) did NOT trigger: Brent closed the week at 92.6, above my 90 line. My Friday evening read said otherwise, so the written rule saved me a wrong trade.
- Three planned entries were killed by their own weekly lines: a grid contractor ([$MTZ](https://finance.yahoo.com/q?s=MTZ)), gold miners ([$GDX](https://finance.yahoo.com/q?s=GDX)) and copper miners ([$COPX](https://finance.yahoo.com/q?s=COPX)). No money went in.
- Five trailing stops filled this week, four in profit, with almost zero slippage.

## What I got wrong

On Friday morning I called this a liquidation week: forced selling with no company news behind it. That was right for Taiwan, Korea, memory and gold. It was wrong as a statement about the whole market. By Friday’s close two of the largest AI names had their own specific claims: the customer report for Broadcom, and a news report that regulators are examining accounting questions at [$NVDA](https://finance.yahoo.com/q?s=NVDA) (unconfirmed; the regulator declined to comment). From here I judge company by company: forced selling where nothing changed, a dated test where something did.

## The cycle lesson: spending level vs spending growth

The closest precedent for AI spending is the 1996–2000 telecom build-out. Telecom spending peaked in 2000. The leading equipment maker’s stock peaked in March 2000, then fell about 87% by October 2002, while that year’s spending was still at its record. **The stocks priced the end of growth, not the end of spending.**

Today, big-tech data-centre spending is forecast at roughly 1 to 1.2 trillion dollars in 2027. On Goldman Sachs’ path, growth slows from nearly 100% this year to about 54% in 2027 and about 12% in 2028. The level still rises. The growth rate has already turned. That is the setting in which one customer trimming a plan can take 12% off a supplier in a day.

What is different from 2000: most of today’s spending is paid from the buyers’ operating cash flow, and parts of the chain are sold out into 2027. So I rank AI suppliers by how much of 2027 is already contracted: foundry ([$TSM](https://finance.yahoo.com/q?s=TSM)), memory and chip tools ([$ASML](https://finance.yahoo.com/q?s=ASML)) first; design wins that can move between vendors after their dated tests.

**What would change the view:** if the hyperscalers raise 2027 spending on their late-January calls and the stocks still fall, this is liquidation, not the cycle. If any one of the four cuts its 2027 plan, the bull case breaks.

## What I am watching next

- Monday in Asia: a lower open is likely after Friday’s US close. The first sign that sellers are exhausted is a close well off the lows on heavy volume.
- [$TSM](https://finance.yahoo.com/q?s=TSM): my zone is 295–315. I act on a daily reversal inside it, not on a fixed price.
- Broadcom: I wait for its 9 December results. Nvidia: I wait for the company’s response and its quarterly filing.
- [$BTC](https://finance.yahoo.com/q?s=BTC): the weekly close on Sunday night. Below 55k my value-buying plan ends.
- Sunday’s outlook decides whether to add more duration.

## Risks, openly

The sell-off could deepen before any reversal. Accounting questions can take weeks to resolve. A December rate hike is still priced. Cash protects capital but misses a sharp rebound if one comes fast.

## The process

One entry that earned its place. One that cancelled itself before a −12% day. Three setups killed by their own lines. Forty percent cash, waiting for the one thing it is there for: a reversal day.

*Snapshot: Friday 20 November 2026 closes (portfolio data Saturday 07:44 UTC).*

Sources: [Motley Fool on Nvidia’s 20 Nov drop](https://www.fool.com/investing/2026/11/20/nvidia-stock-plunge-wipes-out-nearly-usd500-billion-in-one-day-as-sec-reportedly-probes-its-accounting/) · [PYMNTS on Goldman’s 2027 AI capex forecast](https://www.pymnts.com/?p=4233514) · [Nasdaq on the 2000–2002 equipment-stock crash](https://www.nasdaq.com/articles/25-years-later-cisco-finally-recovers-dot-com-crash) · [Data Center Dynamics on Moody’s capex forecast](https://datacenterdynamics.com/en/news/moodys-hyperscaler-capex-forecasts-marked-up-by-85bn-to-close-in-on-1trn-by-2027/)

*This is not financial advice. Investing and copy trading involve real risk of loss, and past performance does not guarantee future results. Always do your own research.*

---

Human page: https://millionaur.com/millionaur-2026-11-21-daily-analysis/?vwao_human=1