From ‘go local’ to ‘go global’. AI is expanding the ambition of startups

Brian Requarth sold the real estate marketplace Viva Real to OLX Brasil in 2020 for R$ 2.9 billion. Months later, he founded Latitud Ventures – a venture capital fund to be the “first check” for Latin American startups. Initially, almost all investments were focused on the countries where the startups were founded. But from late 2024 onwards, Brian noticed a significant change: increasingly, startups are born with a focus on the global market.

What explains this turn? “It was a ChatGPT moment,” Brian told Brazil Journal. Luisa Dalla Costa, another partner at Latitud, said that artificial intelligence “has blurred geographical barriers”; today it is possible to build global solutions with a small team from an office in São Paulo. “We started to see founders who are much more ambitious, wanting to dominate markets abroad,” Luisa said.

The transformation impacted Latitud's main investment thesis: instead of betting on businesses in Latin America, the fund now seeks to bet more on founders from the region. “We stopped seeing Latin America as a market, and now we think of the region as a base of talent,” Brian said. Latitud has already raised two funds, one of US$ 12 million and another of US$ 25 million, and today it has a stake in more than 100 startups.

In addition to the funds, the manager also offers a competitive fellowship, whose numbers prove the 'change in mindset' in the ambition of the new generation of Latin American startups. In 2021, less than 4% of the program's participants had businesses focused on the international market rather than their country of origin. This percentage jumped from 2024 onwards, and in the second half of last year it reached 46%.

In the world of venture capital for AI, (almost) all roads lead to San Francisco. Last year, the US concentrated 79% of funding. It was US$ 159 billion – and of this total, US$ 122 billion came from the Bay Area. The fellowship, previously essentially an opportunity for networking and meeting people who lead the technology sector in the US, has now evolved to be the first check for startups.

The selection funnel is cruel. There are only 8 selected per 'class', about 1% of the total registrations – a lower approval rate than Stanford or Harvard, where the percentages are between 3% and 4%. Among the winning startups of the most recent fellowships, most focused on creating 'horizontal' AI solutions, with application layers for specific purposes based on the frontier models of giants like OpenAI and Anthropic.

Two startups created by Brazilians (and supported by Latitud) are examples of this. One is Taste Labs, founded by Thais Castello Branco, which sells data for training models and aims to develop the taste and senses of LLMs. The other is BeConfident, created by Robson Amorim, which developed an AI tutor to improve English fluency and already has more than 3 million users worldwide.

“We see many opportunities to apply the models on the frontier of AI more efficiently, especially in the middle-market and in specific workflows,” Luisa said. “This is what we have seen happening.”