Dear Fellow Investor,
On July 6th, I wrote you a sentence that got me some mail.
“Microsoft looks stupid cheap to me.”
A few of you thought I’d lost my marbles.
A $3 trillion company. Stupid cheap. Sure, Dylan.
On Thursday, Microsoft closed up more than 16%.
Best single day in nearly eighteen years.
Cloud revenue hit $59.3 billion.
Azure grew 43% in the quarter.
Azure crossed $100 billion in annual revenue for the first time in its history, and the backlog of contracts customers have already signed but haven’t paid for yet climbed 84%, to $678 billion.
Six hundred and seventy-eight billion dollars of work already on the books.
That is what “stupid cheap” looked like four weeks ago.
But I’m not writing to take a victory lap.
I’m writing because of what almost nobody noticed underneath that number.
Amazon and Apple both reported after Thursday’s closing bell.
Same evening. Same test Microsoft had just passed.
The entire market is holding its breath over whether the money these companies are pouring into AI is ever coming back.
And here’s the thing about those three companies.
Microsoft. Amazon. Apple.
They are three of the most valuable businesses on earth…
They compete with each other in nearly every direction…
And every one of them has quietly stopped being just a customer for chips.
They now design their own.
Amazon has Graviton and Trainium. Andy Jassy said the custom chip business crossed a $20 billion revenue run rate.
Microsoft has Cobalt.
Google has Axion, sitting as the host processor for its AI systems.
Apple has been designing its own silicon for years.
Four different companies. Four different chips. Four different strategies.
One architecture underneath all of them.
Not one of them designed their chip from a blank sheet of paper.
They all licensed the same foundational blueprint, and they all pay a royalty on every single chip they ship.
Let me tell you why that matters more this week than it did a month ago.
On Wednesday night — the same night Microsoft reported — the company that licenses that blueprint reported, too.
Almost nobody covered it.
Record quarter. Revenue up 22%. And the number that dropped my jaw: royalties from data centers more than doubled year over year.
They’ve now shipped over 1.5 billion of these data-center cores. The first billion took about six years. The most recent five hundred million took nine months.
Then this, from the research firm IDC: spending on servers built around this architecture has nearly doubled in two quarters and has now passed spending on the old Intel-style x86 platform.
Not caught up. Passed.
Here’s the lesson I want you to take from this past week, whether or not you ever buy a share of anything.
In every gold rush in history, there were two ways to make money.
You could dig. Or you could sell the shovels, the maps and the mules.
Digging is high-stakes, thrilling, some might say “glamorous.” And some diggers got unbelievably rich.
Most went home broke, because digging costs a fortune up front and pays off only if you picked the right hill.
That’s how I see the AI trade right now.
Microsoft picked the right hill and just got rewarded with its best day since 2008.
Meta reported that same week, grew revenue 28%, and got punished — net income down 14%, free cash flow crushed to $784 million after $31 billion of capital spending.
Same industry. Same quarter. Opposite outcomes.
Selling the maps is different. You don’t care which digger wins.
You get paid on every chip, from every hyperscaler, in every direction the technology goes.
Amazon beats Microsoft?
You get paid.
Apple’s new processor takes over the desk?
You get paid.
Nvidia’s next generation ships? Its main processor is built on the same blueprint. You get paid.
There is a name for that kind of business.
A toll booth.
You do not have to guess which car wins the race. You only have to stand at the exit.
So what do I do with this?
Going into Thursday night, I was watching Amazon and Apple like everybody else.
My marker was AWS growth above 31% — clear that, and the AI trade would have its second validation in twenty-four hours, with legs into September.
But my own money was never riding on which of those two printed the better number.
And that is the whole point I want to leave you with this weekend.
I didn’t have to sweat those numbers on Thursday night. I won’t have to sweat the next set either.
Because I already know who gets paid either way.
I’ve been a big fan of this company for a long time — long before this past week made it obvious.
It’s not a household name, and almost nobody outside the semiconductor business can tell you what it does.
It is simply the one supplier that every single one of these giants cannot build their AI chips without.
Which is why I put the whole thesis, the company, and the specific reason its royalty rate is rising into one briefing for you.
If Thursday’s move made you feel like you missed something — this is the part you missed.
If you bought Microsoft on July 6th when “Microsoft looks stupid cheap to me” may have sounded ridiculous to some readers, congratulations on your gains.
“The Buck Stops Here,”
Found this helpful? Share it with others.
Written by Dylan Jovine