HPC and AI in Mexico: New Perspectives and Recent Advances

In this session, a speaker named Isidoro representing Mexico and the Abacus HPC laboratory presents new national perspectives on HPC and AI, building on preceding talks about Brazil and the SCALAC regional network. He describes a historic shift in Mexican federal policy: after years of limited investment, HPC and AI are now backed by constitutional provisions, a presidential mandate, and substantial funding. The plan centers on 10 main HPC nodes, mostly in central Mexico, to be upgraded and interconnected, collectively summing to about 300 petaflops in FP16. Mexico last appeared on the Top 500 list with the Abacus machine. A flagship new system, Cuatlicue, is announced with roughly 6,000 million pesos (about 345 million US dollars) of investment, targeting 314 petaflops and operation within 24 months, with a Mexican team currently preparing at the Barcelona Supercomputing Center. The national strategy rests on pillars of governance and ethics, a shared consortium model, connectivity and cybersecurity, processing and storage, and human resources, with connectivity backbones linking to the United States and, via CLARA and GEANT, to the rest of the Americas. The architecture spans a Tier-0 anchor (Cuatlicue), the national cluster of 10 nodes on a 100 GB backbone, and AI factory services combining public and private investment. Talent development is central, including a program targeting 25,000 graduates per year, aiming for sovereign national capability modeled partly on Brazil's approach.