The European Space Agency’s first supercomputer, Space HPC, is now operating at the agency’s ESRIN center in Frascati, near Rome. Inaugurated on March 12, 2025, in partnership with Hewlett Packard Enterprise, the system was built to process the growing volume of data generated by Earth-observation missions and other space projects.
Michel Leonard, Space HPC’s project manager, describes its purpose in straightforward terms: providing enough computing power to process large datasets and perform calculations on them. ESRIN is the ESA center responsible for Earth observation, whose satellites continuously produce data used by scientific teams and by a range of applications.
Access is not limited to ESA staff. The agency can also make the system available to companies, start-ups and laboratories from its member states, with each request assessed individually.
Space HPC combines about 34,000 AMD and Intel processor cores with 108 Nvidia H100 graphics processing units. Together, these components provide a peak performance of 5 petaflops, equivalent to five million billion operations per second. The system also includes 3.6 petabytes of SSD storage.
Its architecture is divided into several computing sections. One handles conventional physical simulations using processors, while another uses graphics processors for artificial intelligence, fluid dynamics and the processing of large datasets. A third section is dedicated to quantum-computing emulation: using 64 graphics processors, Space HPC can simulate a 39-qubit quantum computer, allowing researchers to test algorithms intended for future quantum systems.
The supercomputer uses direct liquid cooling for its components. According to the ESA, its power usage effectiveness, or PUE, is below 1.09, meaning that cooling and other data-center equipment require less than 9% additional electricity beyond the power consumed by the servers. The heat recovered from the system is also used to warm the ESRIN campus.
Space HPC is considerably smaller than Europe’s largest supercomputers. Jupiter, inaugurated in Jülich, Germany, in September 2025, reaches 1,000 petaflops. The ESA presents its own machine not as a competitor to national and European supercomputers, but as a system intended to complement them.
One of its main applications is space-weather forecasting. Solar eruptions can send clouds of charged particles toward Earth, potentially affecting satellites, electrical grids and aviation. Observatories such as SOHO, operated by the ESA and NASA at the L1 Lagrange point 1.5 million kilometers from Earth toward the Sun, can detect these events before they arrive. Models then use the measurements to estimate the particles’ trajectory and arrival time.
On January 18, 2026, SOHO detected a coronal mass ejection that reached Earth after 25 hours at nearly 1,700 kilometers per second. The ESA’s space-weather office said initial models had estimated its speed at 1,400 kilometers per second, and the resulting geomagnetic storm reached the highest levels of the warning scales.
According to Jorge Amaya, who coordinates space-weather modeling at the ESA, Space HPC could reduce the time needed for these forecasts from several hours to a few minutes. That faster processing would help researchers analyze solar activity and model its possible effects more quickly.
Sources: Numerama.