European organisations developing AI hardware and infrastructure have been given access to €80m in new EU funding.
Investment is aimed at taking AI compute technologies from advanced chip designs through testing and towards deployment.
The Chips Joint Undertaking (Chips JU) has opened two linked funding opportunities covering different stages of the AI compute value chain.
One €10m call will fund European AI chip demonstrators, while the larger €70m programme will establish the infrastructure needed to test those technologies and procure pilot systems.
Both calls have a revised deadline of 27 October 2026 at 17:00 Brussels time, bringing chip designers, AI infrastructure operators and cloud providers into a shared development and evaluation process.
Connecting chip development with real-world testing
The initiative follows a relatively direct progression. Rather than supporting chip development and infrastructure procurement as separate activities, the two calls are designed to connect the stages.
Under the first opportunity, advanced European AI chips will be developed into working demonstrators. Those systems will subsequently be assessed through the common evaluation platform created through the second call.
This assessment will provide a basis for identifying technologies that can move beyond individual chip demonstrations towards rack-level systems and, ultimately, pilot deployment.
The first call, AI chip demonstrators for EU compute infrastructure, has an indicative EU contribution of €10m. Individual projects can seek up to €1m, although the programme allows applicants to request a different amount.
Funding available for advanced European chips
The funding targets AI chip designs that have already progressed significantly toward production.
Applicants must include an EU-established AI chip design company responsible for the chip and its intellectual property.
The technology must have been taped out at a commercial foundry, while engineering samples must be available or expected within the specified timeframe.
Projects must also have a PCB hosting the chip either in production or expected during the six-month development period.
The funding can then be used to complete the demonstrator, covering elements including the chip, PCB, cooling and interconnects. It also includes software integration and preparation for performance benchmarking.
Once completed, each demonstrator can undergo up to two months of assessment through the shared evaluation platform.
The strongest-performing technologies may receive the AI Compute Excellence Label. This will be a requirement for participation in the subsequent A2 rack-development stage, which is planned for a later point and will not open during 2026.
€70m programme targets infrastructure and procurement
The second opportunity represents the larger part of the funding package, with an indicative EU budget of €70m.
It focuses less on developing individual chips and more on creating the environment in which AI compute systems can be tested, compared, and eventually deployed.
The programme will establish common evaluation infrastructure and services for benchmarking the systems developed under the initiative. It will also support the procurement of pilot AI compute racks.
The application consortium must include at least one contracting authority or organisation operating an EU AI Factory or a comparable public AI/HPC facility.
It must also include at least two EU-established cloud service providers, with beneficiaries representing at least three different EU Member States or EEA countries.
From demonstrators to multi-rack systems
The procurement phase is intended to go beyond isolated hardware demonstrations.
Selected systems will be installed and optimised as scaled, multi-rack clusters, allowing distributed AI workloads to be tested across larger infrastructure.
The overall action is expected to last 42 months and will move through three connected stages: testing the initial A1 chip demonstrators, preparing and evaluating the later A2 rack systems, and procuring and installing selected systems.
This creates a broader development route for European AI compute, beginning with chip-level technology and extending through benchmarking, rack development and infrastructure deployment.
The programme is divided into two operational streams. Stream A covers the development of European AI chips and racks, while Stream B provides the evaluation and procurement framework.
For now, only the A1 demonstrator call and the Topic B evaluation and deployment call are open. The A2 rack-development opportunity is planned for a later stage and is not scheduled to launch in 2026.
