While everyone is still watching NVIDIA’s stock price and wondering when the AI bubble will burst, the “Iron Man of Silicon Valley,” Musk, wouldn’t be Musk if he didn’t stir up some big news.
Just now, Musk made a bold statement on social media that has left the entire semiconductor and automotive industries restless. This is not just bragging; it’s practically a declaration of war!
The situation is as follows: Musk officially announced that:Tesla’s AI inference chips will enter an “annual upgrade mode”! That’s right, just like your iPhone, a new version every year, without any delays. Moreover, he added a kicker:Tesla’s chip production target is to exceed the total production of all other chips combined!
With this declaration, one wonders if Jensen Huang (Old Huang) at home might be sneezing. Today, let’s delve into just how audacious Musk’s move is and what ambitions lie behind it.
01 This iteration speed is “life-threatening”
First, let’s talk about this “annual upgrade mode”.
In the traditional automotive industry, it takes at least two to three years for a chip to go from design to being installed in a vehicle. Many car manufacturers are still using chips from years ago, claiming they are “automotive-grade stable.” But in Musk’s eyes, this is simply a snail’s pace.
Musk’s current strategy is: treat car manufacturing like making smartphones and treat chips like fast-moving consumer goods.
Currently, Tesla’s HW4.0 (Hardware 4.0) is still warming up, and HW5.0 is already on the way. Now he tells you, no need to guess anymore,there will be new models every year! What does this mean? It means that Tesla’s computing power growth will exhibit exponential explosion. While competitors are still struggling to adapt to the previous generation of chips, Tesla will already be using the next generation with double the performance.
This is not just technical suppression; this is adimensional strike. It’s like two people racing: while you’re tying your shoelaces, the other person has already switched to a rocket booster. This fast-paced hardware iteration is designed to keep up with the crazy evolution of Tesla’s FSD (Full Self-Driving) software.When software is lacking, hardware compensates; when hardware is strong, software takes off. This is Musk’s violent aesthetics.
02 “Production exceeds the total,” has this boast been broken?
Next, let’s look at the most controversial point:annual production must exceed the total of other chips.
Many netizens laughed at this: Old Musk is talking nonsense again. You have to take on all the chip manufacturers in the world by yourself?
But guys, don’t rush to mock Musk.Remember when he said he would build reusable rockets, who believed it? When he said the Model Y would become the global sales champion, who believed it? And what happened? They got slapped in the face.
Musk’s “exceeding the total” likely refers to theedge AI inference chips, which are the chips installed in cars and in the future Optimus robots.
Think about it, if Tesla’s annual sales reach 10 million vehicles, each requiring top-tier AI chips; plus potentially billions of humanoid robots in the future, the demand will be astronomical.Musk doesn’t want to be constrained by TSMC’s production capacity, nor does he want to pay “toll fees” to NVIDIA; he wants to control core productivity himself.
If this production truly comes to fruition,Tesla will no longer just be a car manufacturer; it will become one of the largest AI hardware manufacturers in the world. This scale effect will bring cost advantages that will be a nightmare for all other competitors. When your chip costs one-tenth of others but performs twice as well, how do you play this game?
03 The drunken man’s intention is not in the wine, but in the “cloud”
In fact, Musk’s insistence on chips has a deeper logic:to break free from dependence on NVIDIA.
In today’s AI circle, everyone is working for Huang Renxun. An H100 graphics card is hard to come by, and prices have skyrocketed. Although Musk has also purchased tens of thousands of NVIDIA graphics cards to train models, he is fundamentally unwilling to lose.
Tesla not only has FSD chips on the vehicle side but also Dojo supercomputer chips for cloud training.Musk’s statement is actually about building a completely closed-loop ecosystem:
-
Data: Millions of vehicles running on the road every day, collecting massive amounts of real-world data.
-
Computing power: Self-developed Dojo chips training models in the cloud.
-
End devices: Annual upgrades of AI chips perfectly executing trained models in vehicles.
This entire process is fully self-developed, fully self-determined. This is the real moat! In contrast, other car manufacturers are still stuck in the patchwork model of “buying chips + buying solutions + buying data,” and the gap will only widen.
Some say Musk is a madman, others say he is a fraud, but it is undeniable thathe is one of the few people on this planet who can turn “blown-up claims” into reality.
This time, Tesla’s AI chip “annual upgrade” declaration is not only a self-imposed challenge but also a warning to the entire semiconductor and automotive industries:In this era of AI explosion, being a step slow means being a step behind the times.
We don’t know if Musk can fulfill his promise by next year, but one thing is certain:the future battlefield of smart cars will no longer be about sofas, TVs, and refrigerators, but about the “strongest brain” hidden beneath the chassis.
What do you think? Can Musk pull this off, or is it just another “pie in the sky”?
Feel free to leave your comments below, and let’s discuss!