Bull has detailed a €31.7 million project to replace the supercomputing infrastructure behind Spain’s national weather service, with two clusters supporting weather, climate, environmental modelling and AI. The announcement is new, but the contract is not: Spanish procurement records show AEMET awarded it in May.
NOTE : The September 10 announcement is new, but the contract itself is not. Spain’s official procurement record shows that AEMET (the meteorological agency of Spain) awarded the project on May 26 for €31,685,680, with the formalisation notice published on June 26. Advanced Computing Spain, S.L. is listed as the legal awardee, while Bull is now presenting and delivering the project under its brand. The procurement record shows that one offer was received.
The four-year contract covers the supply, installation and interconnection of a supercomputing system distributed across two clusters, together with support and warranty services. The project is partly financed through Spain’s Recovery, Transformation and Resilience Plan using NextGenerationEU funding.
Weather forecasting needs more than raw computing power
Bull says the new infrastructure will operate across two separate locations in the Madrid area and include hardware, software and round-the-clock monitoring. The company claims it will increase performance by 10%, while allowing AEMET to optimise existing applications and introduce additional AI-based workloads. Taken in isolation, a 10% performance increase can appear modest for a new supercomputing project. The procurement documents show why peak performance is only part of the story.
The agency also explicitly said its existing infrastructure was approaching technical and operational limits. That was affecting not only deterministic and probabilistic weather models, but also chemical and environmental modelling, very-high-resolution forecasting, AI initiatives and large-scale data analysis.
A weather supercomputer is therefore not simply a machine used to produce a forecast faster. It is part of continuously operating public infrastructure that has to process increasingly large volumes of observational data and model output within strict time windows.
AEMET already runs a substantial Bull-based HPC system
AEMET has used high-performance computing for more than three decades. Its current Cirrus infrastructure reached its final configuration in 2023 and consists of two clusters, one for production and another for development. Together they contain 376 compute nodes, 752 processors and 48,128 CPU cores, with theoretical peak performance of 1.68 petaflops. AEMET also maintains separate high-performance storage systems, including a 3PB development filesystem.
The second cluster also provides operational resilience. AEMET says it regularly moves its production suite to the development system so that workloads can be transferred if the main production environment suffers an incident. The replacement project preserves the two-cluster principle while Bull says the new infrastructure will be located across two separate sites.
That helps explain why continuity, storage, networking and energy management were explicitly weighted in the tender alongside compute performance.
Bull’s European HPC story is getting broader
The AEMET announcement comes less than two weeks after Bull was selected for a much larger European project : LUMI-AI. The projects address very different needs. LUMI-AI is being built as the computational backbone of an AI Factory serving research, startups and industry. AEMET is buying mission-critical infrastructure for operational weather, climate and environmental modelling, while adding capacity for AI rather than making AI the system’s sole purpose.
That distinction matters for Bull, which became wholly owned by the French state on March 31 following its separation from Atos. France has presented the acquisition as part of its strategy to preserve sovereign capabilities in HPC, AI and quantum technologies.
Bull’s recent order book now spans AI factories, scientific computing, aerospace and national weather infrastructure. Its own 2026 announcements include Sweden’s Mimer AI Factory, a new supercomputing platform for Airbus, Finland’s Roihu system, LUMI-AI and now AEMET.
For AEMET, however, the relevant test will not be another contract announcement. It will be whether the new architecture can reliably produce more demanding forecasts, climate simulations and AI-assisted workloads while remaining available around the clock, precisely the combination of performance and resilience that Spain’s procurement documents were designed to buy.