Elon Musk revealed on Tesla's earnings call that the company would be building its own AI chips for its cars, likely abandoning the Nvidia Drive platform the company has used up until now.
The new chips, known as "Hardware 3.0," will be compatible with all existing Tesla cars in the S, X or 3 ranges. "We made it easy to switch out the computer, and that's all that needs to be done. You take out one computer and plug in the next. All the connectors are compatible," the Tesla CEO said.
Tesla's cars have a semi-autonomous driving feature named Autopilot, which uses machine learning, Lidar and radar to determine where the cars needs to be on the road. (Image: Pixabay)
By using its hardware instead of relying on Nvidia's platform, Tesla can control its user experience and can have more input on the route it wants to take, rather than having that dictated by Nvidia.
This is not a new idea in the tech industry: Apple builds its chips for the iPhone and iPad, named the A-Series, which has got them to the top of the pile for performance over rivals Qualcomm and MediaTek. Not everyone has that luxury, but Tesla is choosing a similar route to Apple, recalling computer scientist Alan Kay's quote: "People who are really serious about software should make their own hardware."
For his part, Musk says the GPU element of the new chip can manage 2,000 frames per second, whereas the current Nvidia Drive platform can only muster 200, while building the hardware will cost Tesla as much as buying it in from Nvidia. Hardware 3.0 should start reaching cars sometime in 2019.