Everyone says: don't put all your eggs in one basket.
I have two baskets. That's it.
Not because I'm reckless. Because I've looked at what actually generates returns — and it's almost never the broad basket.
Value doesn't spread evenly. It concentrates.
1% of companies generate 99% of returns. That's not a theory. That's how markets work. And AI is making it more extreme, not less.
Owning everything means you own the winners. And 1,590 things that aren't.
The MSCI World has 1,600 companies. Every sector. Every continent. The S&P 500 has 500. Europeans go even broader — they add the MSCI for emotional Europe exposure. More eggs. More baskets. More dilution.
It protects you from being wrong.
It also guarantees you never fully benefit from being right.
The alternative most people reach for: pick the winners yourself.
I've tried. I bought Tesla on the way down — made 100% in a few months. Sold. It kept going. Same with Netflix. I looked at Palantir when it was trading at €6. The numbers didn't look great, so I passed.
Palantir is at €156 today.
Here's what those stories have in common.
I can identify the right companies. My problem is holding them. Dark clouds appear — I move. Uncertainty creeps in — I sell. That works in business. It destroys compound returns.
And I don't want to live inside quarterly reports. I don't want to track every management change. Not my skill. Not where I want to spend my attention.
So I stopped guessing at the company level.
I ask a different question instead: what does every possible AI future have to run through?
Not which company. Not which sector. Which bottleneck — the universal prerequisites that no version of the future can skip.
I was too young and too broke to ride the internet wave. I watched from the outside.
That's not happening again.
I want to be fully in on AI. But not as a stock picker — I'd lose that game. As someone who owns the infrastructure. The layer every AI future has to run through.
Those bottlenecks right now are two:
Chips. Every model, every agent, every application needs compute. No workaround.
AI Software. The orchestration layer — where intelligence becomes usable. Models are commoditizing. The systems that deploy them aren't.
Robotics comes next — when that wave arrives. Not yet.
I own both layers via concentrated ETFs. Semiconductor indices. AI infrastructure funds. Less than 70 positions total. Not one broad index. The right structural layers.
The historical case is already written.
Semiconductors (SOXX) have compounded at 22% per year over the last 15 years. The NASDAQ 100: 16% per year. The S&P 500: 8%.
Same period. Roughly double — or triple — the returns. Not by picking winners. By owning the right structural layer.
The internet didn't lift all boats. It built a few enormous ones. Right at the infrastructure layer.
AI won't be different. It will be more.
That's not diversification. That's structural concentration.
Diversification spreads your ignorance.
Concentration builds on your conviction.
I can live with concentration. I can't live with average.
This isn't financial advice. It's how I think. Your situation is different. Your conviction has to be your own.