Elon Musk has a message for American business leaders: Get ready for the AI revolution or start writing your company's obituary.
This year alone, Tesla's CEO is ramping up AI training and inference and positioning Tesla at the forefront of the industry in non-automotive real-life applications, at a time when Tesla's CEO is reducing investment in new vehicle production capacity. We're spending $10 billion. Generative AI.
“Companies that spend at this level and do not operate efficiently cannot compete,” he wrote on X Sunday.
Spending on AI inference will primarily target his range of cars, as he prepares to custom design the next generation of fully self-driving (FSD) computers known as HW5. It may be showing.
The distinction between training and inference is important. Because it turns out that Musk is currently working on another major AI project. His humanoid robot is called Optimus, after his 1980s comic book vehicle that transforms into a sentient robot.
This bold and risky AI pivot marks a departure from his previous focus on increasing auto sales 10x to 20 million EVs a year, leading to questions about whether Tesla is an automaker or not, favoring the latter. It definitively answers the age-old question: Are you a technology company?
The typical car executive has long invested in rejuvenating one of the auto industry's oldest product lines. For example, BYD, Tesla's biggest EV rival, is rolling out new models across its portfolio of brands with the help of a small army of 90,000 automotive engineers.
Tesla plans to spend about $10 billion this year on a combination of training and inference AI, the latter used primarily in cars.
Companies that do not spend and perform efficiently at this level cannot compete.
— Elon Musk (@elonmusk) April 28, 2024
But Musk seems to think of his car as an iPhone on wheels, a premium device for distributing high-margin software, since revenue is collected by providing vehicle-related services. , and can be sold at lower profits.
So far, that approach hasn't worked. Tesla has found itself forced to repeatedly cut prices to stimulate enough demand to keep its factories open. Musk recently even resorted to cutting the price of FSD software by a third.
Just 18 months ago, the idea that Tesla was having trouble finding customers seemed laughable, to borrow Musk's favorite adjective. But China's new generation of EV rivals are in a league of their own when it comes to value for money, hurting his personal brand.
Musk's latest answer is to become the first global company to move away from direct car-to-vehicle matching and instead tap into the future market of self-driving ride-hailing networks.
It's true that Tesla's CEO predicted every year that his cars would be able to drive fully autonomously without supervision and would fail every time, but his new FSD software v12 is a potential game-changer. Thing. Unlike all previous attempts, it runs entirely on his AI without relying on hard-coded commands, and initial feedback from customers has been positive.
New deal expands FSD to China
Emboldened by this success, Musk quickly acquired all the AI chips he could find.
In the first quarter alone, Tesla spent more than $1 billion to double its computing power to the equivalent of 35,000 Nvidia H200 chips, the benchmark for AI processing. Musk last week promised that number would reach 85,000 by the end of the year.
Mr. Musk is concerned that his direct competitors will not heed his advice and that Tesla will be the first to solve large-scale unsupervised autonomous driving and beat Waymo to the lucrative business of licensing its self-driving technology to rivals. I hope we can beat them.
An important step in that direction is proving that v12 works just as well overseas as it does in the United States, where the software was trained.
On Monday, he managed to close a deal that could allow FSD to eventually gain approval in China. To appease the Chinese government's demands, Mr. Musk partnered with local internet search giant Baidu, a major competitor in self-driving cars, to license the company's mapping and navigation software.
“We see Musk winning FSD approval in the key Chinese market as a turning point in Tesla's story,” Wedbush senior technology analyst Dan Ives said Monday. Tesla stock is expected to rise more than 9% at the start of trading.
Meanwhile, the Tesla community has raised suspicions that Tesla will nominally cancel its $25,000 low-cost EV and instead launch a hatchback version of its Model 3 sedan that can be built using existing production lines. I'm strengthening it. That means Musk doesn't have to splurge even more cash to build new capabilities.
Investors like billionaire Ron Baron welcomed the decision because Tesla already has many factories and could easily shut one down. Baron said Musk's company can currently produce 3 million cars this year at its four car manufacturing plants, but it probably won't sell more than 2 million this year.