I’ve had a very good run with AI stocks. Better than I probably deserved.
I’m keeping a few.
But I’m also taking some money off the table because I’ve reached that uncomfortable point where I still like the companies, I still believe the story, and I increasingly dislike the price.
That distinction matters.
Take Palantir around $190.
I can make the bull case myself. Excellent business. Huge operating leverage. Real demand. Government moat. Commercial growth.
But I get an economic FCF yield of roughly **0.5%**.
To clear my **12.5% return hurdle**, even assuming PLTR is still worth around **40x FCF five years from now**, I need roughly **50%+ annual FCF/share growth**.
For five years.
AMD around $630 is similar. Current economic FCF yield is below **1%**. Because semiconductors are cyclical, I want about **15% annually**.
The maths wants roughly **55% FCF/share growth**.
Marvell gets close to **60%**.
Could they do it? Absolutely.
But there is a difference between *possible* and *what I want to pay for today*.
Meanwhile:
**Adobe \~$235:** roughly 9% economic FCF yield. Needs only about **6% FCF/share growth** to clear my hurdle.
**Intuit \~$276:** roughly 7% normalized yield. Needs **7–8%** growth.
**Uber \~$70:** roughly 6% yield. Needs around **10–11%**.
**FICO \~$864:** roughly 5% yield. Needs about **11%**, although the debt makes me twitch slightly.
So I’m rotating some AI winnings into those kinds of setups, while watching LVMH, Rightmove, Novo and UNH.
Nothing revolutionary.
Just businesses where I don’t need the future to arrive wearing a cape.
Of course, I could be completely wrong.
AI cash flows may explode. Multiples may stay at 40–60x. Treasury yields may moon and nobody cares. Zcash may do 800%. Chillguy may do 1,400%. The next Black Bull Anselm may casually print +2,000%.
And I’ll be somewhere in Europe drinking a **€19 glass of Burgundy**, explaining “normalized owner earnings” to absolutely nobody who asked.
That is the risk.
I’m not leaving AI.
I’m just no longer willing to pay today for every miracle AI might perform tomorrow.
If that makes me boring, fine.
**Boring is survivable. Paying 60 years of good news upfront usually isn’t.**