NVIDIA’s AI-Powered Supply Chain Revolution | Tech News

0:00On September 11th, Nvidia is stepping up its game, using Palancy or Foundry and this thing called QOpt to automate how they move all their hardware around the world. They're tracking everything from when the silicon is made to when it's ready to power up a data center. This whole process is broken down into getting parts to the assembly line and then getting the final product ready to go. The complexity is wild, especially with their new NBL72 racks. Each one needs a ton of components. Think 18 compute trays, each packed with CPUs, GPUs and memory chips. All these parts come from thousands of different suppliers. And get this, the next system they're building, VR Rubin, is going to have a supply chain twice as big as the current one. It's a real balancing act. They can't build anything until parts arrive from different places and sometimes they're stuck waiting on just one missing piece which messes with. Their time of ownership metric. They're constantly tweaking their plans, looking two quarters ahead to figure out what parts they'll have, how fast they can build and when they can.

1:04Get it all to customers. To handle all this, they built a command center using Palancy or Foundry. It maps out all the factories, supplier promises and inventory. Then, Cool Opt, which is like a super fast optimization tool, takes that information and figures out the best way to distribute everything, trying to. Minimize that, time of ownership. It even flags bottlenecks like limited assembly space versus how much memory they actually have. But numbers only tell part of the story. They realize they needed to factor in things like weather or supplier calls. So, they trained an AI model called Nemaltron 35 Lightning on all these qualitative records, even stuff like email exchanges. This fine-tuned AI is now way better at making allocation decisions, hitting over 86% accuracy, which is a huge jump from the previous models. They're even planning to use this data for future learning, making the system smarter over time. Tech news, powered by AI. This is Tech News Today.