UFMG Unveils 'Soberano' Supercomputer to Map Brazil With AI

Overview

• The university inaugurated the Soberano system and a Hyper‑Cam Mini xLW hyperspectral camera, placing them into operation this week to process satellite, drone and radar data with AI.
• Soberano is built with eight Nvidia H200 GPUs, additional processors and about 400 TB of storage, and was designed, installed and configured by Brazilian teams with public funding of R$15 million.
• The platform can run thousands of AI models in parallel to produce maps up to six times finer than current public products and can generate 1:25,000 scale cartography used for planning and water‑basin management.
• Researchers showed immediate applications such as detecting hundreds of dirt roads near BR‑163 and revealing finer detail of the São Francisco channel, and they expect thematic maps for six states by the end of 2027.
• Officials plan to make Soberano available to federal and state bodies including Inpe, Ibama, Conab and the Geological Service to speed detection of illegal deforestation and mining, help mineral prospecting and give farmers quicker proof of legal origin.