The Brazilian government announced on Thursday (August 20) in the “Official Federal Gazette” (DOU) that it will invest approximately 23 billion reais (about 4.4 billion US dollars) to strengthen the country’s artificial intelligence infrastructure and related ecosystem construction.
This funding will be allocated to a supercomputer project expected to adopt American technology, as well as a supercomputing infrastructure project located in Rio de Janeiro.
The funds will come from Brazil’s National Fund for Scientific and Technological Development (FNDCT), and will be disbursed in stages. These projects are part of the “Brazil AI Plan” (PBIA), and the Brazilian government plans to invest 23 billion reais (about 44.2 billion US dollars) between 2024 and 2028 to develop artificial intelligence infrastructure, talent training, public services, and related technologies.
The Brazilian government aims to reduce dependence on overseas data centers and foreign technology infrastructure by expanding the country’s AI computing capabilities.
Approximately 10 billion reais will be used through open bidding to purchase and construct a new artificial intelligence supercomputer.
According to the Brazilian government’s plan, this system’s AI computing capacity can reach up to about 7.2 ExaFLOPS, equivalent to approximately 7200 petaflops of FP16 precision. This positions it to potentially become one of the world’s most powerful AI computing supercomputers.
The supercomputer project will be installed at the Augusto Severo Technology Park in Macaíba, Northeast Brazil, and will be operated by the National Laboratory for Scientific Computing (LNCC).
President Lula of Brazil attended the announcement event in Rio Grande do Norte on August 20. The decision to place this large AI infrastructure in the northeast is also linked to the region’s rapid development of wind energy and other renewable energy industries.
An unnamed Brazilian government official told Reuters that the American chip company Nvidia is expected to be a major supplier for the supercomputer.
The Brazilian government aims to have this system operational by the end of 2027, making it available for use by businesses, universities, research institutions, and government agencies to assist in training AI models and conducting related research and development.
On another front, the Brazilian government will invest approximately 13 billion reais to build another artificial intelligence supercomputing infrastructure in Rio de Janeiro in the southeast of the country.
According to the Brazilian news agency Agência Brasil, this project will be coordinated by the National Education and Research Network (RNP) of Brazil, in collaboration with a Chinese company. The project, which is set to commence operation in July 2027, plans to allocate 1.276 billion reais of public funds over five years, focusing on establishing supercomputing infrastructure for developing large language models in general and specific industrial areas.
The project aims to enhance Brazil’s ability to develop and deploy AI systems independently through technology transfer, research collaboration, training professionals (expecting to train 5,000 Brazilians in five years), and building autonomous software capabilities.
Furthermore, the project will develop national-level large language models (LLMs) and artificial intelligence models in Portuguese to create solutions more tailored to the Brazilian language, data, and needs, reducing the long-term reliance on English language data and foreign technical platforms imposed by large language models.
In July of this year, the Brazilian government signed documents to establish the World Artificial Intelligence Cooperation Organization (WAICO), hoping to establish an AI governance framework through open-source models so that various enterprises or organizations can grasp this technology.
In addition to the two large supercomputing projects, the Brazilian government also announced a series of AI complementary measures, including establishing the “Brazil Cloud” service, setting up a National Center for Algorithm Transparency and Trustworthy AI, and promoting semiconductor and open-source chip technology research and development.
This demonstrates that Brazil’s AI layout this time is not simply about acquiring two facilities, but rather an attempt to build a complete ecosystem covering computing power, cloud services, artificial intelligence models, talent cultivation, algorithm governance, and chip development.
Luciana Santos, Minister of Science, Technology, and Innovation (MCTI), emphasized, “Currently, about 60% of Brazil’s AI computing is done abroad. We will reduce external dependence, and in the future, we will be able to handle computing independently.”