Together, Nvidia and Foxconn are developing "AI factories," a novel kind of data center that will use its supercomputing capacity to speed up the arrival of driverless vehicles, machine learning, and industrial robots.
Nvidia's GH200 Grace Hopper Superchip and AI Enterprise software would form the foundation for the factory. The plants' uses will extend beyond just producing components for autonomous vehicles.
Nvidia-Foxconn Partnership
Jensen Huang, founder and CEO of Nvidia, and Young Liu, chairman and CEO of Foxconn, made the partnership public on Tuesday, October 17, at Hon Hai Tech Day in Taiwan. The artificial intelligence (AI) plant will be built on top of an Nvidia GPU computing infrastructure to filter, enhance, and turn massive volumes of data into usable AI models and insights.
"We're building this entire end-to-end system where, on the one hand, you're building this advanced EV [electric vehicle] ... with an AI brain inside that allows it to interact with drivers and interact with passengers, as well as autonomously drive, complemented by an AI factory that develops a software for this car," stated Huang onstage at the event, as reported by TechCrunch.
He added that the vehicle will accumulate additional information via real-world use. The data will be sent to an AI factory, which will then use it to improve software and update the entire AI fleet.
The partnership between the AI factory and Foxconn expands upon their agreement to create platforms for autonomous vehicles, which was announced in January.
According to the terms of the deal, Foxconn will serve as a principal source for electronic control units (ECUs) that are equipped with Nvidia's Drive Orin system-on-a-chip (SoC), a supercomputing AI platform that enables autonomous driving functionalities.
Foxconn also said it will produce ECUs utilizing Nvidia's next-generation SoC, the Drive Thor. Production of these ECUs is expected to begin in 2025.
Foxconn, which has been slowly releasing off-the-shelf EV platforms for automakers to buy, said that the cars it contracts to make will be built with Nvidia's Drive Hyperion 9 platform. This platform includes Drive Thor and a set of sensors like cameras, radar, lidar, and ultrasonic that are needed for self-driving.
See Also : NVIDIA Boosts AI Performance Using TensorRT
Comparable to Tesla's Dojo
These AI factories are direct competitors to Tesla's Dojo supercomputer, manufacturing of which began in the summer under Elon Musk's leadership.
The neural nets that fuel, train, and develop Tesla's sophisticated driver assistance system, named full self-driving (FSD), will be trained by Dojo. Musk believes that the robust computing capabilities of Dojo are key to realizing his vision of a completely autonomous FSD.
While Tesla now makes use of a massive supercomputer powered by Nvidia GPUs, the new Dojo will be constructed from the ground up using Tesla's own unique processors.
