October 31, 2025 – During Tesla’s third – quarter 2025 earnings call as reported by Tom’s Hardware yesterday, CEO Elon Musk unveiled an innovative concept. He proposed that when Tesla vehicles are idle, they could form a massive distributed inference fleet to make full use of their powerful computing capabilities. Musk projected that in the future, the combined computing power of this advanced fleet could reach an astonishing 100 gigawatts.
Musk also shared Tesla’s production goals. He stated that the company is on track to produce 3 million vehicles annually within the next 24 months. He further emphasized that a crucial step in this production expansion will be the Cyber Cab, with production scheduled to start in the second quarter of next year.
In addition, Musk revealed that Tesla’s team is exploring “killer applications” for new models equipped with advanced processing power. He said, “If we tell consumers that this car is so smart that they can play on their phones and send messages throughout the whole journey, then anyone who can afford it will buy it – it’s that simple.”
Musk reiterated his belief in an “Autopilot safety revolution.” He firmly guaranteed, “I’m 100% confident that we can achieve fully autonomous driving without human supervision, and its safety will far surpass that of human drivers.”
To back up his confidence, Musk discussed the capabilities of Tesla’s AI4 computer (Hardware 4, HW4). He pointed out that the upcoming AI5 has shown a 40 – fold performance improvement over AI4 in tests. This significant performance boost implies that the safety of future autonomous driving systems will increase by 10 times.
Taking advantage of the topic of computing power, Musk also pondered the possibility that future systems might be “too smart” for cars. To address the issue of having “bored” cars, he came up with an interesting idea: using the computing power of idle vehicles to transform the Tesla fleet into a large – scale distributed computing network. “I’ve thought about it. If we have a large number of potentially bored cars… we can definitely create a massive distributed inference fleet,” Musk said.
According to the report, Musk optimistically estimated that this fleet could expand to 100 million vehicles. With each vehicle providing one kilowatt of computing power, “that would be 100 gigawatts of inference power, and at the same time, power and heat – dissipation issues would be solved,” Musk explained to financial analysts during the earnings call, adding that “this will be a very significant asset.”
