Taiwanese manufacturing large Foxconn plans to construct so-called “AI factories” with the assistance of Nvidia.
The collaboration, introduced throughout a gathering between Nvidia CEO Jensen Huang and Foxconn CEO Younger Liu this week, includes the large-scale deployment of the chipmaker’s GPU methods to energy a wide range of industrial purposes.
“A brand new kind of producing has emerged — the manufacturing of intelligence. And the datacenters that produce it are AI factories,” Huang stated per a press release. “Foxconn, the world’s largest producer, has the experience and scale to construct AI factories globally.”
If it wasn’t clear, the time period “AI manufacturing unit” is a metaphor for taking unstructured information from vehicles, robots, IoT sensors dotted round a metropolis, and different sources and turning it right into a helpful product, a lot in the identical method conventional factories may soak up uncooked supplies on one finish and come out an car on the different finish.
AI factories are “specifically constructed for processing, refining, and remodeling huge quantities of knowledge into precious AI fashions and tokens,” the chipmaker defined in a weblog publish. In different phrases, they’re datacenters brimming with very costly GPU methods tailor-made to assist the shoppers’ numerous enterprise verticals.
In Foxconn’s case, the iPhone maker plans to deploy a wide range of HGX, GH200 Tremendous Chips, and OVX reference designs alongside loads of Mellanox-derived networking equipment.
These methods will assist a wide range of services and products constructed atop Nvidia’s enterprise software program platforms. For example, Foxconn plans to construct a “Good EV” platform utilizing the chipmaker’s upcoming Drive Thor autonomous driving pc system, due out in 2025. It is value noting that Foxconn already builds digital management items for vehicles utilizing Nvidia’s Drive Orin platform.
Foxconn additionally has its eye on “sensible manufacturing” purposes that benefit from Nvidia’s Isaac autonomous robotics platform, and a “sensible metropolis” platform that includes the GPU-slinger’s Metropolis video analytics service. That stated, sensible cities discuss has been round for some time with little or no progress.
So, not solely does Nvidia have Foxconn hooked on what’s more likely to be a considerable variety of GPUs, however its rising software program enterprise as effectively.
One other use case Nvidia has talked up up to now is the usage of OVX methods developed to energy its Omniverse platform as a instrument to visualise machine studying and sensor information to create digital twins of precise manufacturing unit flooring or warehouses. The thought right here is these digital twins and can be utilized to simulate whether or not operational modifications will bear fruit earlier than they’re carried out in the true world – assume Excel however stepped up a notch.
“Simulating your complete robotics and automation pipeline from end-to-end supplies Foxconn with a path to operational effectivity positive factors and saving time and prices,” Nvidia claimed.
What occurred to the ‘Industrial Metaverse?’
Nvidia has spent the previous few years working to domesticate new markets for accelerated computing, past these usually related to hyperscale and cloud datacenters.
The thought of an “AI manufacturing unit” is not new, although the buzzword of alternative could have modified following the fervor round generative AI apps like ChatGPT, Secure Diffusion, and different giant language fashions.
In mid-2022, Nvidia was throwing round phrases like “industrial metaverse” amid a collaboration with German multinational Siemens to convey photorealistic digital twins to mainstream trade utilizing its Omniverse platform.
As we talked about earlier, Nvidia continues to be pushing this concept together with its extra graphics-optimized OVX methods. However for the reason that hype across the much-touted metaverse has seemingly fizzled, no less than in keeping with Gartner, Nvidia seems to have pivoted to AI factories as an alternative. ®
