To avoid relying on outside semiconductor suppliers, General Motors’ self-driving car division Cruise is developing its own chips. Due to high costs and limited supply at Cruise partner Nvidia, the GM subsidiary has been convinced that relying on third parties for necessary components like chips is neither practical nor reasonable.
Cruise disagreed with Nvidia’s pricing despite the lack of available chips. Carl Jenkins, the company’s head of hardware, told Reuters that Cruise wasn’t buying enough chips to be eligible for a discount because it is a small division of GM.
For the upcoming Origin model, a boxy self-driving electric vehicle (EV) without pedals or a steering wheel that is estimated to be available by 2023, Cruise started making its own chips rather than placing more orders for Nvidia chips.
Two years ago, Cruise began developing chips using then-commonly available ARM CPUs. Similar chips are utilized in portable gadgets like smartphones and tablets because they are energy-efficient. Cruise claims that Origin’s current ARM designs will increase range while reducing power consumption.
The Cruise has developed four internal chips: the Dune sensor data processor, the unnamed radar chip, the Horta computing chip, and the last, unnamed gadget that will be revealed later. The cruise will nevertheless need to work with an unnamed Asian semiconductor producer to develop its specialized CPUs. As a result, there will be some outside businesses present.
The specialized ARM chips will be available by 2025, but Cruise is already thinking about switching to the open-source (cheaper) RISC-V architecture, which would be more suited for its self-driving cars.
One of GM’s goals, according to Cruise, is to cut the cost of AVs and make them more affordable for more people by the middle of the decade. To do this, GM plans to develop a less expensive in-house chip and take control of its supply chain.
Carmakers don’t want to fall behind if there is another chip shortage, even though the current one is gradually getting better. Especially not automakers working on self-driving cars, which rely even more on microprocessors than regular cars do.
