Elon Musk proposed using idle Tesla cars as a mobile inferencing fleet on the electric vehicle company’s Q3 2025 earnings call.

Seguing from an answer he was giving about expanding production capacity and his confidence in unsupervised autonomy in Tesla's driverless vehicles, Musk said that the company’s world simulator for reinforcement learning would see a 10x improvement, which “might be too much intelligence for a car.”

“I do wonder, how much intelligence should you have in a car? It might get bored,“ Musk said, before adding: “One of the things I thought of… if we've got all these cars that maybe are bored… we could actually have a giant distributed inference fleet. If they're not actively driving, just have a giant distributed inference fleet.”

He went on to explain that if you have tens or hundreds of millions of cars in a fleet that had 1kW of high-performance inferencing capability, “that’s 100GW of inference distributed with cooling and power conversion taken care of. That seems like a pretty significant asset.”

Musk did not expand on the idea any further, and the moderator of the call did not ask any follow-up questions.

For the third quarter of 2025, Tesla posted revenue of $28.1 billion, a 12 percent Year-over-Year (YoY) increase, with net income falling 37 percent YoY to $137bn. The company said the drop in profit partly resulted from the company increasing its operating expenses to fund artificial intelligence and other R&D projects.

Providing an update on the company’s $16.5bn chip deal with Samsung for the manufacturing of its forthcoming AI6 chip, to be used in its cars and robots, Musk clarified comments he’d previously made about the production of the current generation AI5 chip.

TSMC has originally been slated as the sole manufacturer of AI5, but Musk said on the earnings call that production would now be split between TSMC and Samsung, who will both make the chips in the US, at their respective fabs in Arizona and Texas – the latter of which Musk claimed has “slightly more advanced equipment.”

"By some metrics, the AI5 chip will be 40 times better than the AI4 chip. Not 40 percent, 40 times,” he said, adding that despite improvements Tesla is making to its home-grown chips, the company still intends to deploy them in combination with Nvidia GPUs in its data centers.