Nvidia announces Drive PX 2, a computer for self-driving cars that uses deep neural network called Drivenet capable of recognizing 120M+ objects
Nvidia Announces New Drive PX 2 ‘Supercomputer In A Lunchbox’ For Self-Driving Cars — Following last year's Drive CX …
Context & Ripple Effects
Drive PX 2 is the second generation of a platform Nvidia has been building since CES 2015, when its Tegra X1 chip powered the original $10K Drive PX developer kit. The pitch has stayed consistent — put GPU-class deep learning in the car — but Drivenet's claim of recognizing 120M+ objects moves the conversation from hardware specs to trained perception.
The follow-through is what makes this announcement matter: within two years Jen-Hsun Huang had signed Toyota onto Drive PX supercomputers, and by late 2017 the roadmap reached Pegasus for Level 5 autonomy. Drive PX 2 is the hinge between Nvidia-as-gaming-chipmaker and Nvidia-as-autonomy-compute-supplier.
First-order effects
- Automakers evaluating self-driving programs get a production-intent compute option instead of research kits, cutting the path from experiment to vehicle integration.
- Drivenet's object-recognition benchmark gives Nvidia a concrete way to sell perception capability to carmakers rather than raw silicon.
Second-order effects
- Mobileye, Intel and Qualcomm-style suppliers of driver-assist silicon are pushed to answer with their own neural-network stacks, since OEMs now expect deep learning as part of the compute package.
- The $10K dev-kit pricing from the first Drive PX signals where margins sit: value migrates toward the software stack and training data around the box, not the lunchbox itself.
Third-order effects
- If the pattern holds, cars become one more market where Nvidia sells an integrated AI stack — chips plus networks plus middleware — the same structure it later formalized with Pegasus, Xavier and Drive AutoPilot, entrenching its position before regulators ever set autonomy rules.
The trend: Autonomous-driving value is consolidating around full-stack AI compute platforms supplied by chipmakers, with each generation of Nvidia's Drive line raising the performance floor OEMs must meet.