Tesla just shared more about its AI vision. These are the 3 biggest takeaways.

By William Gavin

A lot is happening with Tesla's automotive and energy businesses. But the company used its earnings call to further hammer a future centered on AI.

Tesla has gathered 6 billion miles of training data from drivers with electric vehicles equipped with the company's supervised Full Self-Driving system.

Elon Musk loves to say that Tesla Inc. is more of an artificial-intelligence and robotics company than a traditional automaker. His company's most recent earnings call emphasized that even further.

Despite record electric-vehicle sales and energy deployments in the latest quarter, Tesla (TSLA) executives spent most of the call focused on AI, which is at the core of several major products set to define the company's future.

And Musk says he's ready to get the ball rolling.

"We are at a critical inflection point for Tesla," the chief executive said on the company's earnings call Wednesday. "It's honestly going to be like a shock wave."

Ramping up robotaxis

A fully driverless robotaxi service across the U.S. is a major part of Tesla's growth plans, as the company looks to get a slice of an industry that Goldman Sachs estimates could generate $7 billion in annual revenue by the end of the decade.

Tesla took a big step toward that outcome in July, when it began offering ride-hailing services in Austin, Texas, using Model Y vehicles operated by its Full Self-Driving software. However, as a safety precaution, every vehicle is still required to be accompanied by a Tesla employee who can react in the case of an emergency. But that might change soon.

"We are expecting to have no safety drivers in at least large parts of Austin by the end of this year," Musk said, referring to Tesla's pilot launch.

Besides Austin, the only area where Tesla offers supervised ride-hailing services is California's San Francisco Bay Area. But by the end of 2025, Tesla expects to be operating in between eight to 10 metro areas, including in Nevada, Arizona and Florida, pending regulatory approval.

And as the company ramps up, Musk said Tesla could shorten the time it keeps safety drivers in the car to "maybe three months" before taking off the training wheels.

"We probably could just let it loose in these cities, but we just don't want to take a chance," he said.

See more: This underrated Tesla business deserves more attention - and it's not AI

Although Tesla's timeline in Austin is one measure of confidence, Morningstar's Seth Goldstein noted that Musk's new outlook for expansion is more conservative than his prior forecasts. "The earnings call confirmed our view that a full launch would likely be delayed" until after next year, Goldstein said.

But Musk maintains that he is "100% confident" that Tesla will solve unsupervised Full Self-Driving. It helps that the company has an ocean of training data: Vehicles equipped with the widely available, supervised Full Self-Driving had logged 6 billion miles driven as of September.

An 'oversupply' of chips

Semiconductors are among the most important components of Tesla's product lineup. The small silicon wafers power everything from EVs to smartphones, and Tesla has big ambitions for the AI5, its next-generation chip.

"Our explicit goal is to have an oversupply," Musk said on Wednesday. "If we have too many AI5 chips for the cars and robots, we can always put them in the data center."

Tesla uses a combination of in-house chips and Nvidia Corp. (NVDA) graphics processing units to train its AI models. Initially, Tesla had tapped Taiwan Semiconductor Manufacturing Co. Ltd. (TSM) (TW:2330) to make its next-generation AI5 chips, while Samsung Electronics Co. (KR:005930) would make the current-generation AI4 chips and the AI6 chips meant to succeed the AI5.

Now, Tesla plans to put Samsung to work on the AI5 alongside TSMC, Musk said, noting on the call that he has nothing but "great things" to say about the company. In July, Samsung signed a $16.5 billion supply deal with Tesla to make chips at its Taylor, Tex., facility. That deal lasts through 2033.

Representatives for Samsung and TSMC declined to comment.

The AI5 chip, first announced in 2024, will be 40 times better than the existing AI4 chip, Musk said. The chip could have the best performance per dollar for AI "maybe by a factor of 10," according to the Tesla CEO.

"Tesla only has to satisfy requirements from one customer," Musk said. "That makes the design job radically easier and means we can delete a lot of complexity from the chip."

Tesla is just one of several companies that are moving to lessen their reliance on Nvidia. Meta Platforms Inc. (META), for one, is working on custom chips designed to fit its unique needs, and recently acquired chip startup Rivos Inc. Elsewhere, Amazon.com Inc. (AMZN) and Alphabet Inc. (GOOG) (GOOGL) also have their own custom chips.

Read: Google shows off its semiconductor prowess with a new Anthropic deal

A 'robot army'

On Wednesday, Musk took the chance to make what was likely his final verbal appeal to Tesla investors still on the fence about his compensation package, which will be voted on in less than two weeks.

The money, he said, isn't the driving force behind his interest in the approximately $1 trillion package. Rather, it's about making sure he has a major say in the company.

"It's called compensation, but it's not like I'm going to go spend the money," Musk told investors. "It's just, if we build this robot army, do I have at least a strong influence over that robot army?"

Ahead of Tesla's upcoming annual meeting, investors have to decide whether to approve or reject a few proposals related Musk's pay, including one that guarantees he will get the benefits of his contested 2018 compensation plan.

See More: Here's what Elon Musk must do to become a Tesla trillionaire

Also on the table is a controversial plan that would allow Musk to earn some $1 trillion as long as the company hits certain milestones over about a decade, including delivering 1 million Optimus humanoid robots.

The market for humanoids could be worth $5 trillion by 2050, with the technology picking up speed in the mid-2030s, according to Morgan Stanley estimates. Musk has called Optimus an "infinite money glitch" and said sales could one day account for 80% of Tesla's value.

But getting the robots made is easier said than done.

Tesla first introduced the concept of its Optimus robots in 2021 and showed off a prototype the following year. While Musk thinks Tesla can make 1 million robots annually by the end of the decade, the company has been dealing with issues related to the design of the robots' hands and forearms, as well as sourcing parts.

"So with cars, you've got an existing supply chain," he said. "With computers you've got an existing supply chain. With a humanoid robot, there is no supply chain."

A prototype intended for volume production should be ready for a demonstration by March, Musk said. Analysts are excited for a closer look at Optimus, which was last officially displayed a year ago, even though it's not expected to generate revenue anytime soon.

"While it's a zero today, if we expect this will eventually be a general-purpose robot combining human-like dexterity with superhuman intelligence, it's quite easy to adopt a view that Optimus could sell millions per year for $30,000 each," said William Stein, an analyst at Truist Securities, in a note to investors.

-William Gavin

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

10-25-25 0830ET