Dear Reader,
Good morning,
Happy Friday!
Today I’d like to talk about Nvidia’s earnings:
Nvidia CEO Jensen Huang came out and rejected and destroyed any talk of an AI bubble…
Basically a smackdown of me and everybody else who says we’re in a bubble.
And that’s okay.
He said, “we see something very different.”
So I want to touch on the numbers and really talk big picture here.
First of all, Nvidia’s $57 billion in Q3 revenue is $2 billion more than expected.
That’s great, but here’s the really shocking part: $19 billion in net income.
33% to the bottom line.
Those are abnormal profits.
What that tells you is that there’s not much competition in this space.
The average profit in America is about 12%.
When you see 33% to the bottom line, competitors start introducing competitive products and driving those profits down with competitive strategies.
Huang also said he expects Q4 revenue to be $65 billion, and explains why we are not in a bubble.
He says areas like data processing, search systems and engineering are all turning to GPUs because they need AI.
That means older computing infrastructure based around CPUs will transition to new systems running on Nvidia chips.
Basically, the whole world is getting upgraded on Nvidia chips.
Remember leading up to the year 2000, the Y2K bug? I know you’re old enough to remember.
In 1999 they had to upgrade all the computer systems in the world because everybody was afraid that when it turned to the year 2000, programs that only used two digits for the year would see 00 as 1900 instead of 2000, causing mayhem in everything from banking to power grids.
It was all very amusing, but it drove a massive increase in tech spending.
So number one, Huang says, is that all current systems are being replaced with GPUs.
That’s a big deal.
The second big driver, and reason we’re not in a bubble according to him, is that AI isn’t just being integrated into current apps…
It’s going to enable entirely new apps.
So, existing infrastructure is being replaced with AI…
AI is going to enable new types of apps…
And finally, according to Huang, “agentic ai,” applications that can act as your agent without significant input from the user, able to reason and plan, will require even more computing power.
So he says these are the three huge megatrends driving the AI trade.
What we’re looking at here is like a “reindustrialization,” a re-buildout of everything out there.
Here’s an interesting statistic I want to share with you:
Huang says, of their $57 billion in revenue, $51.2 billion came from GPUs in data centers.
That’s over 90% of Nvidia’s revenue.
Two things strike me about that:
1. Wow - that’s a big number.
It’s like the CPU era as we understood in the nineties is over.
What they’re doing now is creating big, football field-sized data centers tying all these GPUs together to create one massive “brain.”
So, data centers are a huge deal, obviously.
But…
2. Now we are preparing for “embodied AI” - robots - including self-driving cars, Tesla’s plan to roll out tens of millions of Optimus robots, etc. - the world is going robot, basically…
Right now, 90% of AI happens in the cloud.
But for AI to become physical, these devices need to think, themselves.
Imagine a robotaxi in Austin, Texas is driving…
It has to look at everything around it - not just a green light, but the mother walking her child across an intersection - the car itself has to think, “stop.”
Nvidia chips are for the data center (cloud).
We need chips that allow the device itself to think.
This is what Nvidia’s “secret supplier” does.
This is basically going to allow physical AI, or EI - “embodied intelligence” to take off.
Nvidia is a data center company - that’s great.
When these self-driving cars connect to the internet at night and get new mapping or image systems - when they’re pulling on the cloud - that will touch the data center.
But when the internet or satellite goes down because it’s raining or whatever, and the car is driving 50 miles per hour down the road and sees a mother pushing a stroller, it can’t reach out to the cloud and say, “should I go or not? What kind of obstacle is this?”
That processing needs to be in the device itself.
This is the next big opportunity.
The first part of the AI trade? Done. Here we are. Data center infrastructure getting revamped.
The second part, the bigger part, the device part, is coming.
And that’s something we feel we’re really early on, and I’m excited about that.
Enjoy!
“The Buck Stops Here,”
Found this helpful? Share it with others.
Written by Dylan Jovine