Nvidia and Foxconn are working collectively to construct so-called “AI factories,” a brand new class of knowledge facilities that promise to supply supercomputing powers to speed up the event of self-driving automobiles, autonomous machines and industrial robots.
Nvidia founder and CEO Jensen Huang and Foxconn chairman and CEO Younger Liu introduced the collaboration at Hon Hai Tech Day in Taiwan on Tuesday. The AI manufacturing unit is predicated off an Nvidia GPU computing infrastructure that can be constructed to course of, refine and rework huge quantities of knowledge into useful AI fashions and knowledge.
“We’re constructing this complete end-to-end system the place on the one hand, you’re constructing this superior EV automobile…with an AI mind inside that enables it to work together with drivers and work together with passengers, in addition to autonomously drive, complemented by an AI manufacturing unit that develops a software program for this automobile,” stated Huang onstage on the occasion. “This automobile will undergo life expertise and acquire extra knowledge. The info will go to the AI manufacturing unit, the place the AI manufacturing unit will enhance the software program and replace your entire AI fleet.”
The AI manufacturing unit tie-up builds off a partnership between Nvidia and Foxconn introduced in January to develop autonomous automobile platforms. That settlement concerned Foxconn turning into a main provider of digital management models (ECUs) for automakers, which can be constructed with Nvidia’s Drive Orin system-on-a-chip (SoC), a supercomputing AI platform that helps autonomous driving features. On Tuesday, Foxconn additionally dedicated to manufacturing ECUs with Drive Thor, Nvidia’s next-gen SoC, after manufacturing begins in 2025.
As a part of that partnership, Foxconn — which has been steadily unveiling off-the-shelf EV platforms for automakers to buy — stated the autos it makes as a contract producer can be constructed with Nvidia’s Drive Hyperion 9 platform, which incorporates not solely Drive Thor, but additionally a set of sensors like cameras, radar, lidar and ultrasonic which are mandatory for self-driving capabilities.
Foxconn is already contracted to construct EVs for Fisker, even because it will get sued by its erstwhile accomplice Lordstown Motors. The automaker will want scale so as to make its AI factories viable, particularly if it’s going to compete with Tesla.
As a result of these AI factories are primarily rivals to Tesla’s Dojo supercomputer, which the Elon Musk-owned automaker began manufacturing on over the summer season. Dojo will practice Tesla’s neural nets, that are used to energy, practice and enhance “full self-driving” (FSD), the automaker’s superior driver help system. Musk hopes FSD will truly be absolutely self driving sooner or later, which is the place the highly effective compute of Dojo is available in.
Tesla right now makes use of a big Nvidia GPU-based supercomputer, however the brand new Dojo can be custom-built utilizing chips designed by Tesla. The Foxconn-Nvidia AI factories can be primarily based on Nvidia’s GH200 Grace Hopper Superchip and AI Enterprise software program, based on the corporate.
The factories may have different purposes past self-driving automobiles.
On stage at Hon Hai’s tech occasion, Liu shared Foxconn’s purpose to transform itself “from a producing service firm to a platform options firm” by scaling the AI factories throughout varied industries. Foxconn is focusing on three platforms to start out: Good EVs, good cities and good manufacturing.
“This can be a manufacturing unit that takes knowledge enter and produces intelligence as an output,” stated Huang, as Liu nodded his assent. “Sooner or later, each business, each firm may have an AI manufacturing unit.”