Industry & Logistics / AI news for Malaysia
NVIDIA puts Palantir AI into its own chip supply chain first
Two American companies are applying AI to the flow of parts behind every AI rack. Malaysia sits somewhere inside that flow, and the announcement says nothing about where.
In brief
- Palantir and NVIDIA announced on 10 September 2026 that they will run NVIDIA's open Nemotron models inside Palantir Foundry and its Artificial Intelligence Platform to model complex supply chains.[1]
- The first deployment is NVIDIA's own supply chain, which the release describes as millions of parts, thousands of suppliers and a global manufacturing network, with 1.3 million parts in each Vera Rubin rack.[1]
- No cost, contract value, timeline, measured improvement or customer other than NVIDIA is named, and the release does not identify a single supplier country.[1]
Palantir and NVIDIA put open models over NVIDIA's parts flow
The company that designs the chips behind the AI boom says it will point AI at the boring part: getting the parts to the factory. Palantir Technologies and NVIDIA announced on 10 September 2026 that they have built a joint AI stack for supply chains, and that the first place it runs is inside NVIDIA itself.[1]
The shape is straightforward. NVIDIA's Nemotron open models are brought into Palantir Foundry and its Artificial Intelligence Platform, grounded in the Palantir Ontology, to give what the companies call supply chain visibility while the customer keeps control of its own data.[1][3][4]
What the two companies actually announced
The release sets out an AI stack rather than a product launch. Organisations post-train NVIDIA Nemotron open models on their own data using Palantir Foundry and the Artificial Intelligence Platform, and NVIDIA cuOpt software inside that platform handles optimisation and scenario planning so teams can model constraints and assess trade-offs. NVIDIA's product page describes Nemotron as a family of open models under a permissive licence with weights published.[1][2]
Deployment is deliberately flexible. The release says the stack runs on NVIDIA reference architectures and on a jointly developed Palantir Sovereign AI Operating System Reference Architecture supported by Dell Technologies and Cisco, and that it can be installed on premises with Cisco and Dell or run in colocation and cloud with Rackspace and Nebius. NVIDIA's own deployment is the first, beginning with materials allocation decisions.[1]
Sovereign here means who owns the data, not where it sits
The word sovereign is doing specific work here. The release ties it to retaining control and ownership of proprietary data while customising open models with it, not to any national jurisdiction, data-residency rule or government programme. Alex Karp, Palantir's co-founder and chief executive, is quoted calling NVIDIA's supply chain arguably the most complex in the world and describing the stack as part of a sovereign AI revolution.[1]
That distinction matters for anyone in Malaysia who hears the same word from a vendor. On this evidence, sovereign describes an arrangement in which a company post-trains an open model on its own records and keeps the result. Jensen Huang's quote frames the goal as planning the journey from wafer to token, and the release says supply chain experts retain control of final decisions.[1]
The announcement carries no measured result
The only hard operational figure in the release is the 1.3 million parts in each NVIDIA Vera Rubin rack. There is no stated reduction in lead time, no accuracy figure, no cost, no contract value and no completion date. The release says the system helps teams identify constraints earlier, which is intent rather than a measured outcome.[1]
Nor is there a second customer. The companies say organisations across agriculture, manufacturing, pharmaceutical, retail, technology and government can use the stack, and that they plan to extend the learnings from NVIDIA's deployment, but no other named user appears. Everything beyond NVIDIA's own operation is still a proposal being showcased at a Palantir conference.[1]
Why Malaysia should care
NVIDIA describes thousands of suppliers and a global manufacturing network, and names not one country. A Malaysian electronics supplier cannot tell from this release whether it sits inside the graph being modelled. What it can expect is the second-order effect: customers running this kind of system start asking suppliers for cleaner, machine-readable capacity data, and treat a slow answer as a constraint.
Malaysian electronics and precision manufacturers
Your customers are building systems that read supplier commitments as data. Slow or unstructured answers become visible constraints.[1]
Practical move: Get committed dates, capacity and change notices into a structured feed before a customer asks for one.
Logistics and procurement teams in Malaysian groups
The release describes recommendation and scenario planning with people retaining final decisions, not automated ordering.[1]
Practical move: Judge any similar pitch on whether it keeps a named human accountable for the allocation decision.
Boards hearing the phrase sovereign AI
Here it means keeping ownership of your own data and model weights, not national jurisdiction or where data is stored.[1]
Practical move: Ask a vendor to write down which of the two meanings they are selling before signing.
What Malaysians can do now
- Read sovereign in this announcement as data ownership, not data residency. The release ties it to keeping control of proprietary data.
- Do not quote a benefit figure from this release. It publishes none, and describes only what the system is meant to help teams do.
- If a supplier pitch cites this stack, ask which named organisation outside NVIDIA is running it in production today.
What we still do not know
An architecture is not yet an outcome
- What the deployment costs, how long it takes, and what it has improved, because no figure or date is published.
- Which organisations other than NVIDIA are using the stack, since none is named.
- Where the data and models run in each deployment, and whether any Malaysian supplier or facility is inside NVIDIA's modelled network.
Sources
- 1.NVIDIA and Palantir Bring Sovereign Intelligence to Critical Supply Chains NVIDIA, 10 September 2026
- 2.NVIDIA Nemotron open models NVIDIA
- 3.Palantir Foundry Palantir Technologies
- 4.Palantir Artificial Intelligence Platform Palantir Technologies
- 5.NVIDIA Headquarters media assets NVIDIA Newsroom
- 6.Reunion con Alex Karp, CEO de Palantir Technologies, Davos, 21 de enero 2026 Presidencia de la Republica del Ecuador, 21 January 2026
