GM's Cruise built four self-driving car chips in-house, to be deployed by 2025, including for the car's main brains, for processing sensor data, and for radar
Context & Ripple Effects
Cruise began as a self-driving kit startup GM bought in 2016 to build driverless cars, and by 2018 Reuters sources described delays and software problems inside the unit despite its $5B in backing. The 2017 plan to mass-produce a self-driving Bolt EV-based vehicle came with no launch date, so owning the silicon was a way to pull cost and schedule control in-house.
Designing four chips — the vehicle's main compute, sensor-data processing, and radar — targets a 2025 deployment, making Cruise one of the few AV operators building its full compute stack rather than buying it. The later arc matters: GM ultimately pivoted Cruise's teams toward an eyes-off driver-assist system planned for 2028 after exiting robotaxis, so this silicon program is the asset that survived the pivot.
First-order effects
- Cruise gains control of its compute roadmap across the three costliest autonomy components — main brains, sensor processing, and radar — with deployment targeted for 2025 across its fleet.
- The move reduces Cruise's dependence on merchant chip vendors at exactly the moment its mass-production plans needed a locked-down bill of materials.
Second-order effects
- Merchant silicon suppliers keep courting the market anyway — Arm's Neoverse automotive chip designs bet that most automakers will not follow Cruise's in-house path, setting up a build-vs-buy split across the industry.
- When GM folded Cruise's teams into its own to work on autonomous and driver-safety technology, the in-house chip investment became the foundation for consumer driver-assist rather than a stranded robotaxi asset.
Third-order effects
- If the pattern holds, automakers split into two camps — those designing workload-specific silicon internally and those buying it from Arm-style merchants — with chip design cycles, not vehicle platforms, setting the pace of autonomy roadmaps.
- Autonomy silicon programs are proving more durable than the robotaxi business models that funded them, shifting the competitive moat from fleet operations to the integrated compute stack.
The trend: Automakers are internalizing AI chip design, and those silicon investments are outliving the robotaxi ambitions they were originally built to serve.