Palantir Taps Nebius for Sovereign AI Stack

Palantir Technologies and Nebius have joined forces to deliver a sovereign AI stack that keeps chips, storage, networking and models inside a single controlled perimeter, a physical constraint answer before it is a branding one.

Key Takeaways

Palantir and Nebius have partnered to deliver a sovereign AI stack keeping chips, storage, networking and models inside one controlled perimeter
Nebius GPUs and inference tools are embedded inside Palantir’s security perimeter for government and enterprise clients
Nebius emerged as an independent entity after Yandex sold off its core businesses following the invasion of Ukraine
Neither company disclosed financial terms or a deployment timeline in the joint release

Inside A Sovereign AI Stack

The arrangement, laid out in a joint release Tuesday, gives Palantir customers access to Nebius graphics processing units and inference infrastructure without those workloads leaving Palantir’s controlled environment. The sovereign AI stack means the full chain of infrastructure, chips, storage, networking and the models running on top of them, stays under the direct legal control of one customer or government rather than a shared public cloud.

Palantir Technologies (PLTR) built its business supplying software to intelligence agencies and the Pentagon before expanding into commercial AI through its Foundry and AIP platforms.

It has spent two years marketing itself as the home for workloads that cannot touch public clouds.

Governments and regulated industries have pushed for dedicated setups because AI systems built on outside infrastructure can expose sensitive data to foreign jurisdiction or unexpected shutdown, a capacity and control problem before it is a branding one. Nebius GPUs and inference tools are now embedded inside Palantir’s security perimeter so government and enterprise clients can run AI workloads without touching outside clouds.

From Yandex Spinoff To Nvidia-Backed Compute Supplier

Nebius (NBIS) supplies the compute layer under the new arrangement, running inference and training workloads on hardware placed inside Palantir’s security boundary.

The company emerged as an independent entity after Yandex, the Russian internet group, sold off its core businesses following the invasion of Ukraine, leaving its cloud and AI infrastructure unit to rebrand under new ownership based in Amsterdam.

The Next Buyers In Line

Neither company disclosed financial terms or a deployment timeline in the release. The structure could pressure rival AI infrastructure providers built around pay-per-token pricing, since it offers government and enterprise buyers a fixed, dedicated setup instead of metered billing, and dedicated capacity is ultimately what constrained buyers are paying to secure.