UFMG Inaugurates the Soberano Supercomputer to Map Brazil With Artificial Intelligence
The Federal University of Minas Gerais (UFMG) has inaugurated a supercomputer called Soberano that will use artificial intelligence to map Brazilian territory, g1 reported on October 27. The machine is installed at the university’s Instituto de Geociências on the Pampulha campus in Belo Horizonte and represents about R$15 million in public investment.
Soberano runs on eight Nvidia H200 graphics processors, each with 141 GB of memory, backed by 400 terabytes of storage. The project also includes a Hyper-Cam Mini xLW hyperspectral camera, which records many more bands of light than an ordinary sensor and can help tell apart soil types, vegetation and minerals from the air. Together, the hardware is meant to process imagery and train models that turn raw data into usable maps much faster than manual methods.
The need is large. Only about 8 per cent of Brazil’s 8.5 million square kilometres is mapped at the 1:25,000 scale used for detailed planning, according to the project. Coordinator Sônia Carvalho Ribeiro said the new capacity is meant to help close that gap. Partners include the Military Institute of Engineering (IME) and the Army’s Geographic Service Directorate (DSG), and the system is expected to support agencies such as Inpe, Ibama and Conab.
What it means in practice
Better maps feed directly into monitoring. Planned uses include tracking deforestation and illegal mining, assessing flood risk and watching border areas, work that overlaps with Inpe’s own effort after it received R$147.6 million from the Amazon Fund for AI deforestation alerts. Conab, which forecasts harvests, could use hyperspectral data to estimate crop areas more precisely.
Soberano is a university-scale machine rather than a national flagship, but its name reflects a wider debate about Brazil owning its computing capacity. Larger plans are moving more slowly: the federal government has delayed its AI supercomputer tender until after the presidential runoff, while the LNCC is advancing a billion-real system in Macaíba. Smaller, targeted systems like the one at UFMG can deliver results sooner for specific public tasks.
What to watch next is which agencies sign data-sharing agreements with UFMG, how quickly the first AI-generated maps are validated by the Army’s cartographers, and whether the model can be repeated at other federal universities.