Summary
- Palantir has selected Amsterdam-headquartered Nebius as its preferred sovereign AI infrastructure partner.
- Nebius compute and inference endpoints are planned to sit inside Palantir's enterprise perimeter for eligible customers.
- The companies also intend to accelerate new capacity through modular data centre deployments at sites where power is available.
Nebius has been named Palantir’s preferred sovereign AI infrastructure partner under an agreement that includes plans to accelerate new compute capacity, including modular data centre deployments at locations where power is already available.
The partnership will combine Nebius’s AI-focused cloud and compute infrastructure with Palantir’s commercial software environment. Following an integration period, eligible Palantir customers are expected to gain access to Nebius compute and inference endpoints from within the Palantir enterprise perimeter.
The companies say the arrangement is intended to allow customers to retain control over their compute, data, and models while using open-model infrastructure. That is being presented as a sovereign option for organisations that want AI workloads to operate within a more tightly controlled infrastructure environment.
The physical capacity element makes the agreement more than a software integration. Palantir and Nebius said they will work together to bring new compute online faster and specifically identified modular data centre deployments at sites where electricity is already available as one possible route.
That wording reflects one of the harder constraints in the current European AI infrastructure market. Power-ready land and live grid capacity can be more valuable than a nominally attractive site that faces years of connection delays, and modular construction can shorten the physical deployment stage only after the electrical capacity, planning position, cooling strategy, and network connectivity are in place.
Nebius is headquartered in the Netherlands and has been expanding its position as a dedicated AI infrastructure provider. The Palantir agreement gives it access to commercial customers using a platform where data control, security, and deployment boundaries are central purchasing considerations.
The term “sovereign AI” is used increasingly broadly across the European technology market. In infrastructure terms, it can involve questions about data location, operational control, ownership of computing assets, legal jurisdiction, software dependencies, and the ability to move workloads between providers. The partnership announcement focuses primarily on customer control over compute, models, and data rather than setting out a single regulatory definition of sovereignty.
That leaves implementation detail important. Bringing Nebius inference endpoints inside Palantir’s enterprise perimeter can alter the way customers consume infrastructure, but the resilience and sovereignty of the resulting service will still depend on where workloads physically operate and how the underlying facilities, networks, and supply chain are governed.
The modular deployment plan could also change Nebius’s development model where suitable powered sites are available. Rather than waiting for full-scale greenfield campuses, smaller or repeatable modules can potentially be deployed against secured electrical capacity, although high-density AI infrastructure still requires substantial cooling, electrical distribution, backup systems, and network connectivity.
Power availability is therefore likely to remain the limiting factor. The ability to manufacture or deploy modular data centre equipment faster does not remove transmission and distribution constraints, and the companies have not yet named the locations, capacities, or commissioning dates associated with their planned joint expansion.
The partnership was disclosed through a regulatory filing by Nebius on 8 September. It establishes Palantir as a route to market for Nebius infrastructure and gives Palantir a preferred external platform for sovereign compute, but its significance for European capacity will depend on the projects that follow.
The next useful disclosures will be specific: where the modular facilities are planned, how many megawatts have been secured, what density and cooling configurations will be deployed, and whether the resulting capacity is dedicated to Palantir customers or shared with Nebius’s wider cloud platform.
