Barcelona Supercomputing Center has moved its AI Factory Accelerator Programme from pilot to full operation, opening sovereign compute, technical support and training to startups working across industries. Since its launch in April 2025, the BSC AI Factory has delivered more than 800 services to private and public-sector clients. Access to some of those services is now available through the NVIDIA Inception portal. For regional companies, this turns a national supercomputing resource into a practical channel for product development.
How the accelerator and Inception access work
Three startups from the pilot were named AI Champions at the demo day. Qatium operates a water-network platform used by more than 250 utilities in 16 countries and is now in early production with Neural Water Intelligence, with proprietary models trained on MareNostrum 5. OneCareAI, a research spin-off from BSC, built technology to detect signs of hidden atrial fibrillation in seemingly normal ECGs. Avatia built a voice-first, multi-agent system that gives pharmaceutical sales teams agentic coworkers and supplies doctors with curated information about medicines.
The technical stack described in the source centers on NVIDIA accelerated computing and software libraries. Qatium uses the NVIDIA cuGraph library, the NVIDIA PhysicsNeMo framework and NVIDIA Nemotron open models. OneCareAI and Avatia use NVIDIA NIM microservices, while Avatia also uses the Nemotron 3 open model family and the open source NVIDIA NeMo Curator library. MareNostrum 5, the public infrastructure behind the factory, runs on NVIDIA accelerated computing, and an upcoming extension will provide supercomputing access through the BSC AI Factory and other AI factories in Europe.
The programme sits inside a wider European factory network. BSC AI Factory is described as among the first AI factories in Europe and the first in Spain. NVIDIA Inception adds training through NVIDIA Deep Learning Institute open courses, business mentoring, venture capital connections and other resources for member startups. Other participants named in the cohort include Agrikola AI, Elna Health, Lokutor, Montevive AI, Oppy, Pack2Earth and Up2Smart. Their fields cover precision agriculture, medical imaging, speech services, enterprise security, sales intelligence and sustainable packaging.
What this means for companies adopting AI
For small and midsize firms, the main change is lower barriers to high-performance computing and specialist guidance. Instead of building their own clusters or hiring large infrastructure teams, startups can apply for compute time, training and technical support tied to a public supercomputer. That structure fits product teams in health, agriculture, manufacturing and services that need to train or test models on larger datasets. Large companies gain a different benefit: a pipeline of local vendors already tested on shared infrastructure.
The limits are practical and still need direct checking. The source does not disclose selection criteria, capacity quotas, pricing terms, or waiting times for MareNostrum 5 resources. Availability through the Inception portal covers only some BSC AI Factory services, so applicants should clarify which services, support hours and training formats are included. The news by itself does not prove commercial readiness of any single product, and buyers should ask for deployment references, data requirements and integration effort.
A concrete marker to follow is the extension of MareNostrum 5 and the next intake of the accelerator. The timing of added capacity, the number of companies accepted, and follow-on contracts or deployments by Qatium, OneCareAI and Avatia will show whether shared sovereign compute converts into sustained products. If those signals strengthen, regional supercomputing will count as a regular supply option for business AI projects.
