In early October 2026, Techerati reported that the Chips Joint Undertaking opened two funding calls totaling €80 million to help European-designed AI chips move from prototypes toward testing and deployment in European compute infrastructure.
The first call, “AI chip demonstrators for EU compute infrastructure,” has an indicative EU budget of €10 million. Its purpose is to turn sufficiently mature European designs into demonstrators that combine silicon, board, cooling, interconnects and software needed to run and measure the chip.
Chips JU estimates an EU contribution of up to €1 million per project, though applicants may request a different amount. Projects must include at least one EU-established AI chip design company responsible for the design and associated IP. System integrators and equipment makers can join where needed.
The programme targets relatively mature technology: the chip must already have been taped out at a commercial foundry. Engineering samples must be available or expected within the project window, and a board capable of hosting the chip must be in production or expected during the six-month development period. Once completed, demonstrators undergo two months of performance testing on a common evaluation platform.
Top-performing technologies can receive an “AI Compute Excellence” label, required to access a later programme (Topic A2) that develops selected technology into rack-level systems. Topic A2 is not open now and is not scheduled to launch in 2026.
The second call, “AI compute evaluation and deployment platform for EU infrastructure,” has an indicative budget of €70 million. Rather than funding more individual chip designs, it establishes a common way to test AI compute technologies and a route for selected systems to be purchased for pilot deployment.
Applicants must form a buyers’ group involving organisations from at least three EU Member States or EEA countries. The group must include at least one organisation operating an EU AI Factory or comparable public AI/HPC facility and at least two EU-established cloud service providers. The programme is expected to run for 42 months.
Initially, the infrastructure will benchmark demonstrators from the €10 million programme. Later it is intended to test complete rack systems from future Topic A2. The final phase moves from evaluation to deployment: selected systems would be purchased, installed and optimised in multi-rack clusters under distributed AI workloads.
Both calls close on 27 October 2026 at 5 p.m. Brussels time. The path Chips JU envisages is: European chip design → working demonstrator → common benchmarking → rack development → rack-level testing → procurement → pilot deployment.
For European chip companies, that offers a route beyond proving a processor boots: comparison under common conditions and, if they excel, a path toward complete AI compute systems. For AI Factories, public HPC and clouds, the second call is a chance to shape evaluation and buying criteria.
The €80 million does not by itself establish European independence in AI compute. It does address a practical bottleneck: giving home-grown silicon somewhere to prove itself and, if successful, a clearer route toward infrastructure scale.
Neural Pulse links this funding to Lagarde’s dependency warning: without chips and evaluation platforms of its own, model sovereignty stays half-built. The 27 October close is the first filter for who enters that pipeline toward 2027.
