Barcelona Supercomputing Center Expands AI Factory Access via NVIDIA Inception
The Barcelona Supercomputing Center’s AI Factory Accelerator is now fully operational, connecting startups to MareNostrum 5 supercomputing resources and NVIDIA Inception tooling.
The Barcelona Supercomputing Center (BSC) has moved its AI Factory Accelerator Programme into full operation following an initial pilot phase. Designed to foster regional technological development, the initiative provides emerging companies with sovereign supercomputing power, technical guidance, and specialized training. In a recent expansion of the program, access to select BSC AI Factory services is now available directly through the NVIDIA Inception portal for startups.
Through this collaboration, startups gain combined access to public supercomputing infrastructure alongside NVIDIA Inception resources. Inception supports members by offering educational coursework through NVIDIA Deep Learning Institute open courses, business mentoring, connections to venture capital, and developer resources.
Sovereign Compute and the MareNostrum 5 Extension
The BSC AI Factory represents Spain’s first AI factory and stands among the earliest dedicated AI factories established in Europe. Since its initial rollout in April 2025, the facility has delivered more than 800 services to organizations across both the public and private sectors.
Compute resources for the accelerator are anchored by MareNostrum 5, Spain’s public supercomputing infrastructure designed for scientific research and technical innovation. Powered by NVIDIA accelerated computing, MareNostrum 5 is slated for an upcoming extension. This infrastructure upgrade will expand supercomputing availability through the BSC AI Factory as well as other AI Factories operating across Europe.
Pilot AI Champions and Domain-Specific Deployments
During the demo day concluding the pilot initiative, three NVIDIA Inception members were recognized as BSC AI Factory Accelerator Programme AI Champions for their deployments across specialized domains:
- Qatium: Serving over 250 water utilities across 16 countries, Qatium has entered early production with its Neural Water Intelligence technology. The company trained proprietary models on MareNostrum 5 using NVIDIA accelerated computing, the NVIDIA cuGraph library, the PhysicsNeMo framework, and NVIDIA Nemotron open models to help municipal operators improve water network decisions.
- OneCareAI: A BSC research spin-off, OneCareAI applied NVIDIA NIM microservices to detect hidden signs of atrial fibrillation within seemingly normal electrocardiograms (ECGs), aiming to help medical professionals identify stroke risks ahead of time.
- Avatia: The team created a voice-first, multi-agent platform tailored for pharmaceutical organizations. Designed as an agentic assistant for sales representatives and medical professionals, the platform curates clinical data on treatments. Avatia built the system using NVIDIA NIM microservices, the Nemotron 3 open model family, and the open-source NeMo Curator library.
Beyond these three champions, other startups in the BSC AI Factory cohort are deploying systems across several technical areas. These include Agrikola AI, Elna Health, Lokutor, Montevive AI, Oppy, Pack2Earth, and Up2Smart. Their projects encompass agricultural robotics designed to protect crops without chemicals, pre-biopsy breast cancer subtype detection, compostable biobased packaging materials to replace petroleum plastics, speech processing, and enterprise security.
What it means for developers
The expansion of the BSC AI Factory highlights how European infrastructure providers are working to lower the barrier between high-performance computing clusters and commercial software development. Rather than having to build infrastructure from scratch, engineering teams can combine open-source weights—such as the Nemotron 3 series—with modular deployment frameworks like NIM microservices to create specialized enterprise applications.
For engineering teams building AI-enabled software, infrastructure needs depend heavily on the phase of development. Training custom models for scientific or physical simulations often demands dedicated sovereign clusters like MareNostrum 5. At the application layer, however, developers working with large language models can prototype and run production workloads across various foundational architectures without managing dedicated hardware. Developers looking to build with leading models like Claude, GPT, Gemini, Grok, and DeepSeek can access them cheaply through one unified pay-per-token API at https://apixoai.online.
As public supercomputing initiatives and commercial accelerator programs continue to align, developers in Europe have a clearer path to move compute-intensive AI prototypes into production.
Source: Spain-Based Startups Advance Regional Innovation With Barcelona Supercomputing Center — NVIDIA blog. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key
