The machine, built in Germany’s Juelich Supercomputing Centre, has more than 24,000 Nvidia chips and cost €500 million ($580 million) to develop and run. Half of that comes from the EU, the rest from Berlin. The system will be shared across research projects ranging from climate modelling to neuroscience and renewable energy, as well as companies looking to train their own AI models.

Exascale means the kit can handle one quintillion calculations a second. The US already has three of these monsters, tucked inside the Department of Energy. China is thought to have several.

Juelich centre boss Thomas Lippert said: “Jupiter is a leap forward in the performance of computing in Europe,” adding that it is 20 times more powerful than anything else in Germany.

The whole thing takes up 3,600 square metres of floor space with rack after rack of processors waiting to be fed enormous workloads.

On the downside, Jupiter will chew through around 11 megawatts of power, the same as thousands of homes or a small factory. Its operators claim it is the most energy-efficient among the fastest systems, thanks to water cooling and the cunning reuse of waste heat to warm nearby buildings.