Brazil is spending 2.3bn reais, about $444m, on two AI supercomputers. One will be built with Huawei, the other is expected to go to Nvidia, and that is the entire point.

“The strategy is not to depend on a single company, technology or country,” the Lula government said in setting out the plan, which is an unusually direct statement of what most countries do quietly.

The Chinese half is the larger one. Huawei Technologies and the speech-recognition company iFlytek will build a 1.3bn reais system in Rio de Janeiro, worth roughly $251m and intended for training large language models, with work scheduled to begin in July 2027.

The second machine goes to a tender expected to be won by Nvidia, according to unnamed officials and the science and technology minister, Luciana Santos. That project is worth about 1bn reais, or $193m, and is due to be operating by the end of 2027.

It is going to Rio Grande do Norte, in the northeast, chosen for its energy potential rather than its proximity to anything. The state has substantial wind and solar capacity, which, for a machine of this size, is the constraint that matters most, as operators everywhere have been discovering.

Officials say the Nvidia system would rank among the most powerful AI machines in the world, a claim that has not been substantiated with a specification and should be read as an ambition rather than a benchmark.

The money comes from the National Fund for Scientific and Technological Development, released in phases rather than as a lump sum, which gives Brasilia the option to slow either project without cancelling it.

Splitting the work is a policy choice with a cost attached. Two architectures mean two software stacks, two sets of trained engineers, and no ability to pool the machines, and the Brazilian research community will be working across an ecosystem divide that most national programmes try to avoid.

What it buys is insulation. American export controls have repeatedly changed which Nvidia parts can be sold where, and the Chinese domestic alternative has been advancing on custom silicon rather than GPUs, so a country committed to one supplier is exposed to policy it does not control.

Brazil has not said whether US export rules apply to the Nvidia hardware it intends to buy, or whether Washington has been consulted. Nor has it addressed whether the Huawei system falls foul of any American restriction on the company.

Those questions are not hypothetical for a middle-income country buying at this scale. Jensen Huang has argued publicly that Chinese models running on Chinese chips would be a bad outcome for the United States, and Brazil has now arranged for both stacks to run side by side under the same national programme.

The sums are modest by the standards of the buildout elsewhere. Britain put £225m into a single machine, Denmark bought an Nvidia system outright, and the EU has a €30bn gigafactory programme, against which Brazil’s $444m looks like what it is: a first move rather than a bid for parity.

It is also the largest sovereign AI compute commitment in Latin America. The region has been a destination for private data centre investment for years, without much of it belonging to the countries hosting it.

Sovereignty is the word the announcement leans on, and it is doing a lot of work. A machine built by Huawei and a machine built by Nvidia are both machines Brazil does not manufacture, and the sovereignty on offer is over the data and the models rather than over the silicon.

That is still worth something. Training a Portuguese-language model on Brazilian data inside Brazilian borders is a different proposition from renting capacity in Virginia, whatever the badge on the racks.

The research community is the intended beneficiary. Public funding of this kind is typically allocated to universities and state institutes rather than to companies, though Brasilia has not published the access rules for either machine.

Neither system exists yet. The Huawei project starts in 2027, and the Nvidia tender has not formally been awarded.

Get the TNW newsletter

Get the most important tech news in your inbox each week.