Public sector sets out sovereign multi-cloud AI platform push
A major framework will fund a secure, cloud-agnostic AI platform with GPUs and MLOps to speed approvals and unify operations, signalling a sovereignty-first shift.
A major framework will fund a secure, cloud-agnostic AI platform with GPUs and MLOps to speed approvals and unify operations, signalling a sovereignty-first shift.
Follow Tenderlake on LinkedIn for concise insights on public-sector tenders and emerging procurement signals.
A new federal framework aims to deliver a secure, cloud‑agnostic AI platform as a service across government, with an emphasis on sovereignty, open standards, and the ability to accelerate planning and approval processes. The AI Cloud Platform Services framework, published in October 2025, targets a maximum value of €249.6 million over up to four years and will appoint two operators.
The Federal Republic of Germany, represented by the Federal Ministry for Digital Affairs and State Modernization, is procuring a cloud‑agnostic AI platform as Platform‑as‑a‑Service (PaaS). The platform is positioned as a building block in a modular, cross‑cloud infrastructure and must:
Two initial applications are earmarked: KIPITZ and an AI application to accelerate planning and approval procedures. The framework allows additional applications to be added during the term, indicating a platform that is meant to scale as demand grows.
The estimated contract volume is €208 million (net) over a maximum four‑year term, with a maximum value of €249.6 million (net). Funds of €200 million (net) are secured, with a stated expectation that €195 million (net) can be ensured; financing of €49.5 million remains uncertain. The contracting authority stresses there is no entitlement to a specific volume of call‑offs, a point bidders are advised to factor into their calculations.
The framework is open to the federal administration, including ITZBund, with individual releases issued by the ministry for these entities. In addition, govdigital eG is authorised and entitled to call off for itself and its members, with releases issued by govdigital eG for those members. This design broadens the potential user base and creates a shared route to market for AI infrastructure across the federal landscape.
Participation is subject to sovereignty considerations. Economic operators from “non‑covered countries” (third countries without an agreement guaranteeing equal and mutual access to EU procurement) may be excluded case‑by‑case, with decisions based on the jurisprudence of the European Court of Justice, notably case C‑652/22 (Kolin). Sub‑contractors and consortium members from such countries are equally subject to review.
The move reflects a clear shift from commodity cloud towards AI‑ready, sovereign platforms. There are echoes of earlier cloud frameworks, but the AI specificity and multi‑cloud, open‑standards stance mark a step‑change.
In April 2020, the UK’s central purchasing body launched a broad Cloud Compute framework for IaaS and PaaS, designed as a self‑serve route to scalable compute from public cloud owners. That framework focused on “pure” compute rather than managed services. By contrast, the German AI platform requires MLOps pipelines, GPU environments and tight controls on identity, encryption and exit - features that speak to production AI, not only raw infrastructure.
Within Germany, cloud adoption has been accelerating. In May 2024, a federal notice sought IT services in the cloud sector for multiple ministries, including conception, development and maintenance of cloud architectures. In May 2023, a federal printing and security group set up a framework for cloud resources and managed enterprise Kubernetes - a building block for containerised workloads now echoed in the AI platform’s requirements.
Other European administrations are also codifying access to AI capability. Belgium signalled a centralised, vetted approach with its December 2024 prior information notice for a federal AI marketplace, aimed at streamlining access to pre‑vetted tools and services. In January 2024, Norway’s health sector initiated a framework for AI platforms in healthcare to bring clinical AI into routine use. And in January 2025, Copenhagen set out a cloud platform services agreement to support its data science and web teams - another sign of demand for standardised, scalable platforms in the public sector.
This latest German framework aligns with that European trajectory but places a strong premium on sovereign control, including case‑by‑case exclusion of providers from non‑covered countries and detailed exit processes. Those elements are particularly relevant as public bodies look to deploy AI in sensitive, regulated domains.
The specification reads like a checklist for production‑grade AI at scale:
Commercially, a two‑operator model creates competitive tension for call‑offs, while the “no entitlement” clause and funding caveats temper assumptions about guaranteed volume. The breadth of authorised users - from federal bodies to govdigital eG and its members - expands the addressable base but also raises the bar for consistent service management across multiple organisations.
Watch for the selection of the two operators and how quickly the initial applications - KIPITZ and the approvals accelerator - move into steady operation. The handling of the stated funding uncertainty, and how the case‑by‑case eligibility review for non‑covered countries shapes supplier participation, will set important precedents. Equally, the pace at which additional applications are onboarded, and the integration with existing centrally managed on‑premises systems, will indicate whether the platform’s multi‑cloud, open‑standards promises translate into practical gains across the federal estate.
Follow Tenderlake on LinkedIn for concise insights on public-sector tenders and emerging procurement signals.