Brazil Splits $444M AI Supercomputer Bet Between US and China
Brazil announced 2.3 billion reais in AI infrastructure — an Nvidia-expected national supercomputer in Rio Grande do Norte and a Huawei-iFlytek LLM cluster in Rio — hedging between both superpowers.
On Thursday, Brazil’s government unveiled roughly 2.3 billion reais ($444.2 million) in artificial intelligence investments — and in doing so, drew one of the clearest maps yet of how middle powers intend to survive the US-China technology war. Rather than picking a side, President Luiz Inácio Lula da Silva’s administration deliberately split the money between the two rival ecosystems.
Two machines, two superpowers
Just over half of the total — 1.3 billion reais — will fund a supercomputing infrastructure project in Rio de Janeiro, developed in partnership with China’s Huawei Technologies and iFlytek. According to the government, this cluster will be used primarily to develop large language models for both general-purpose and sector-specific applications.
Separately, about 1 billion reais will go through a tender for a national AI supercomputer expected to rank among the world’s ten most powerful AI processing machines. It will be installed in the northeastern state of Rio Grande do Norte, chosen specifically for its energy potential — a nod to the reality that compute-hungry AI is, above all, a power problem. Lula personally attended the announcement ceremony at the Augusto Severo Technology Park in Macaíba.
Government officials, speaking on condition of anonymity, told Reuters they expect US chipmaker Nvidia to win the tender. Science and Technology Minister Luciana Santos had already told Folha de S. Paulo the previous week that she anticipated Nvidia as the supplier.
So the architecture of the announcement is hard to miss: a Chinese-built LLM factory in Rio, an American-built national flagship machine in the northeast, announced the same day by the same president.
“Neither Chinese nor American is better than us”
Lula, facing re-election later this year, framed the investments as a matter of national protagonism. At the ceremony he claimed the United States controls “65%” of global AI while China holds “15 or 20%” — and that Brazil intends to force its way into that market.
“Neither Chinese nor American is better than us. We’re going to enter this market,” the president said, according to Poder360. He tied the project to retaining Brazilian talent and academic capacity, arguing the country must stop exporting the researchers it trains.
The administration’s official statement put the doctrine bluntly: “The strategy is not to depend on a single company, technology or country.” The investments, it added, are aimed at strengthening national sovereignty over data.
The dependence problem being solved
The numbers behind the push explain the urgency. According to the Brazilian government, roughly 60% of the AI processing used in Brazil today is performed abroad, meaning domestic research, industry, and public services rent their compute from foreign clouds under foreign rules. The stated goal is to transform Brazil from a user into a developer of large-scale AI technology.
The initiative is part of the PBIA (Plano Brasileiro de Inteligência Artificial), the national AI plan commissioned by Lula in 2024 with a total envelope of R$23 billion through 2028 — of which the government says 64% has already been executed. The supercomputer track builds on earlier milestones, including the fourfold upgrade of the Santos Dumont supercomputer at LNCC, which was pitched as the first concrete step of the PBIA.
Funding for the new tranche flows through the National Fund for Scientific and Technological Development (FNDCT) via phased disbursements. The timeline is ambitious: the Nvidia-expected supercomputer is slated to begin operating by the end of 2027, while the cooperation agreement with Huawei and iFlytek is scheduled to start in July 2027.
Beyond the megamachines
The announcement contained three additional pillars that received less attention but may matter more in the long run:
- A RISC-V semiconductor partnership with Spain, built on the open-source instruction set architecture — a bet that Brazil can carve out silicon sovereignty outside the x86 (US) and Arm (UK/Japan-owned, US-designed) duopoly.
- A Brazilian cloud-computing service to be built through public-private partnerships, attacking the data-sovereignty problem directly.
- A national center for algorithmic transparency and trustworthy AI, to be operated by the Federal University of Minas Gerais (UFMG).
Together these sketch a full-stack sovereignty strategy: chips (RISC-V), compute (supercomputers), platforms (national cloud), and governance (transparency center).
Why the balancing act is the story
Brazil’s position gives it unusual leverage — and unusual exposure. China is Brazil’s largest trading partner, particularly for soybeans, iron ore, and oil. The United States remains the biggest source of foreign direct investment in Latin America’s largest economy, even as it has lost trade market share and recently imposed additional tariffs on Brazilian goods.
Splitting AI infrastructure procurement between Huawei/iFlytek and (expected) Nvidia is a textbook hedge. It buys access to Chinese LLM engineering and hardware at Chinese prices, keeps Washington’s flagship chipmaker embedded in the national stack, and gives Brasília leverage over both: each vendor knows the alternative is already in the room.
The political calendar sharpens the stakes. Lula is running for re-election against right-wing Senator Flávio Bolsonaro, who has pledged closer ties with US President Donald Trump. A supercomputer deal structured so that both superpowers win makes the AI program harder to dismantle — and harder to campaign against as capitulation to either Beijing or Washington.
The skeptical read
There are real questions about execution. A “top-10 globally” supercomputer delivered by end-2027 on a R$1 billion (~$193M) budget would be remarkably cheap by frontier standards — US and Chinese flagship AI clusters now run to tens of billions of dollars. Officials may be qualifying “most powerful AI processing machines” carefully, or counting phased upgrades toward the claim.
The Huawei-iFlytek track also carries geopolitical risk: US export controls could complicate maintenance, spare parts, or future upgrades for any system perceived as Chinese-backed infrastructure in the Western hemisphere, and a Trump administration fond of tariffs has additional levers beyond export rules.
And sovereignty ambitions collide with market reality — Brazilian enterprises increasingly adopt Microsoft’s AI stack, and the pull of US frontier models among Brazilian developers is strong. Two supercomputers do not, by themselves, make a model ecosystem.
What to watch
- Whether Nvidia formally wins the Rio Grande do Norte tender, and at what final price.
- The specifications released for the machine — petaflop counts will show how literally to take the “top 10” claim.
- Any US reaction to deepening Huawei-iFlytek involvement in Brazilian strategic infrastructure.
- Post-election continuity: whether a Bolsonaro victory would unwind the Chinese pillar of the plan.
Either way, Brazil has just given the world one of the clearest demonstrations yet of the emerging playbook for countries that refuse to choose: buy from both, own the data, build the cloud, and open-source the chips.
Sources
- [1] https://www.reuters.com/world/americas/brazil-launches-ai-supercomputer-push-splits-projects-between-chinese-us-firms-2026-08-20/
- [2] https://www.aljazeera.com/economy/2026/8/21/brazil-launches-ai-supercomputer-push-while-balancing-us-and-chinese-tech
- [3] https://www.poder360.com.br/poder-governo/lula-anuncia-supercomputador-nem-chines-nem-americano-e-melhor/
- [4] https://atos.net/en/2025/press-release_2025_07_09/santos-dumont-lncc-supercomputer-receives-fourfold-upgrade-as-the-first-step-in-the-brazilian-artificial-intelligence-plan