Palantir Technologies and NVIDIA this week revealed they are cooperatively applying artificial intelligence (AI) to better manage the supply chains that NVIDIA relies on to build and deliver IT systems.

Announced at an AIPCon 11 event hosted by Palantir, the joint effort revolves around an instance of the NVIDIA Nemotron AI model that has been embedded into the Foundry data platform and an Artificial Intelligence Platform (AIP) developed by Palantir. Additionally, NVIDIA has incorporated an ontology developed by Palantir to make it easier to apply NVIDIA cuOpt, a decision optimization platform, across its supply chain.

Palantir also revealed plans to make instances of NVIDIA Nemotron AI models available to other organizations that rely on its platforms to optimize supply chains. Those Nemotron AI models, in addition to recommending actions, will be able to explain tradeoffs and flag emerging risks. Each outcome is also integrated with NVIDIA NeMo AutoModel and NeMo RL frameworks and a Palantir Autopilot tool to create a feedback loop that continuously trains the AI model.

Jeff Witmer, vice president of supply chain for NVIDIA, told conference attendees that the capabilities enabled by Palantir will also soon be extended to the company’s partners to further optimize the company’s supply chain for building and delivering IT infrastructure.

The overall goal is to reduce costs by minimizing the time of ownership of the components used to build a system, he added. “It’s the most important metric we have,” says Witmer.

NVIDIA has opted to deploy the software being used to optimize its supply chain in an on-premises IT environment to retain control over sensitive data. At the core of that platform is an NVIDIA reference architecture that has been extended to include a Palantir Sovereign AI Operating System Reference Architecture (SAIOS) running on IT infrastructure provided by Dell Technologies and Cisco.

It’s not clear to what degree organizations are now relying on AI to optimize supply chains, but as they become increasingly extended, they have in many instances become too complex for humans to manage on their own. Palantir is making a case for an approach that leverages the supply chain data organizations have to train custom AI models that can then be used to more easily decipher all the dependencies that may exist in a supply chain. Hopefully, that level of insight will sharply reduce the number of instances where a shortage of some critical components leads to unnecessary delays in shipments.

In the meantime, however, organizations will need to assess how their existing supply chain impacts their ability to compete. Every manufacturer at some point has lost a customer to a rival simply because they were unable to deliver a finished product when required. The number of times a manufacturer is likely to encounter that type of issue is likely to increase as global tensions continue to rise. Regardless of how challenging managing a supply chain is, the one thing that is certain is that end customers are not going to become any more forgiving about delivery deadlines.