Overview
- Officials said Coatlicue will use roughly 14,480–15,000 GPUs delivering 314 petaflops, housed in about 200 water‑cooled cabinets and equivalent to hundreds of thousands of PCs working in parallel.
- The installation is planned over 24 months covering design, civil works, integration, testing and commissioning after a collegiate site decision expected in January 2026.
- Use cases highlighted include climate and water modeling, energy planning, health research and crossing SAT and customs datasets to detect tax evasion and speed verifications, with data remaining in each agency’s systems.
- The nation‑owned system will anchor a National Supercomputing Cluster linking UNAM, IPN, Cinvestav and other centers, operate with a core team of about 80–100 people and recover some costs by offering services to industry and entrepreneurs.
- International cooperation includes technical support and training with the Barcelona Supercomputing Center and an Indian partner, and reporting indicates an interim Mexican compute center will operate from Barcelona while domestic facilities are built.