A look at Tesla's first self-driving chip, manufactured by Samsung, which can be installed into existing Teslas, but the cost to do so is unknown
Advice on Pivoting to the Future and Avoiding Disruption Shreya Ganguly / Inc42 Media : Elon Musk Promises A Million “Self-Driving Robotaxis” By 2020 João Carrasqueira / Neowin : Tesla wants to launch a “Robotaxi” service by next year Reuters : Tesla says robotaxis coming to U.S. roads next year, slams rivals' use of Lidar Matt McFarland / CNN : Tesla doubles down on its radical approach to self-driving cars Alan Boyle / GeekWire : Tesla CEO Elon Musk unveils his Robotaxi concept for a self-driving rideshare fleet Olivia Rudgard / Telegraph : Elon Musk unveils plans to put one million Tesla robo-taxis on the road next year
Context & Ripple Effects
In 2019, Tesla's answer to autonomy was a chip, not a car: a Samsung-manufactured self-driving computer designed to slot into vehicles already on the road, announced alongside Elon Musk's pledge of a million robotaxis by 2020 and pointed criticism of rivals' reliance on lidar. The retrofit angle was the strategic point — Tesla's existing fleet could become the robotaxi supply without new vehicle production.
The arc since then has been slower than the promise: after teasing a purpose-built robotaxi and the sub-$30K Cybercab prototype, Tesla's Austin launch in mid-2025 started with roughly ten Model Ys and safety drivers, and by December still ran an estimated 30 monitored cars against Waymo's 200 driverless ones. The 2019 chip is the hardware foundation that bet rested on.
First-order effects
- Existing Tesla owners gain an upgrade path to full self-driving hardware without buying a new car, though the undisclosed retrofit price leaves whether that path is mass-market or premium unresolved.
- Rivals betting on lidar face a direct public challenge: Tesla is claiming camera-plus-custom-silicon can match sensor-heavy stacks at lower cost.
Second-order effects
- Samsung lands a marquee automotive-AI manufacturing win, strengthening its position as a custom-silicon foundry for carmakers rather than a commodity parts supplier.
- If retrofits work at scale, Tesla's installed base becomes a distributed robotaxi fleet — competing with purpose-built vehicles like the Cybercab on utilization rather than unit cost.
Third-order effects
- The gap between the 2020 million-robotaxi pledge and the 2025 reality of ~30 safety-driver cars versus Waymo's 200 driverless ones shows custom-chip announcements set expectations years ahead of deployment — a pattern regulators and buyers will increasingly discount.
- If the retrofit model holds, passenger cars become upgradable compute platforms whose autonomy capability arrives via silicon and software updates, shifting value from the dealership toward whoever controls the chip and the update pipeline.
The trend: Automakers are embedding custom AI silicon into consumer vehicles so that autonomy arrives through retrofits and software updates rather than new-car purchase cycles.