Airbus has switched on a new distributed supercomputer setup in Toulouse and Hamburg, giving the aircraft maker three times more computing power for virtual design and testing. The upgrade, delivered under a five-year, 100 million euro contract with Bull, is aimed at speeding up the work that decides how quickly a plane moves from concept to certified reality.
The Airbus supercomputer is not just about raw horsepower. Airbus is treating two national sites as one virtual cluster, a way to get around the usual ”bigger machine, bigger room” limitation that still slows plenty of industrial engineering projects.
Two sites, one virtual cluster
The new infrastructure is split between France and Germany, with the French side arriving in 2025 and the German side in spring 2026. Both parts were preassembled and tested at Bull’s plant in Angers before being shipped to Airbus facilities, which is the sort of practical move that saves pain later when thousands of components need to behave like a single machine.
Hardware-wise, the system is built around BullSequana XH3000 modular racks, with a mix of AMD and Nvidia CPUs and GPUs. Storage is based on IBM Spectrum Scale, while Nvidia InfiniBand NDR provides interconnects of up to 400 Gbit/s per port.
Airbus supercomputer specs and use cases
Airbus plans to use the setup for digital twins of passenger aircraft and helicopters, heavy-duty aerodynamic calculations, structural load analysis, and acoustic testing of fuselages. In plain English: more simulation, fewer expensive surprises once metal starts getting cut.
- Five-year contract value: 100 million euro
- Computing power: three times more than Airbus’s previous setup
- Sites: Toulouse, France, and Hamburg, Germany
- Interconnect speed: up to 400 Gbit/s per port
- Cooling: direct liquid cooling with waste heat reuse for administrative buildings
The company is also leaning into a useful side effect of all that heat. Its direct liquid cooling system will redirect waste heat from the servers to warm administrative buildings, a small but sensible example of industrial computing paying part of its own utility bill.
Why Airbus is betting on distributed computing
Airbus is not alone here. Big industrial players from automotive to aerospace have been chasing the same formula: more simulation capacity, tighter integration between design teams, and less dependence on physical prototypes. The difference is that Airbus is now trying to make two data centers behave like one, which is harder than it sounds and very much the point.
If the setup works as intended, Airbus will be able to iterate faster on aircraft concepts without waiting for compute queues to clear. The next question is whether competitors follow the same split-cluster model or keep throwing larger single-site supercomputers at the problem.