Look at Who Built It

Put a pin on a map wherever someone built a serious AI company. Then look at the map. It doesn’t look like a type. It looks like the world.

They built it in their own words

The most interesting AI companies outside the usual centers share one move. They refused to wait for someone else’s model to learn how they speak.

Tokyo. Sakana AI, founded by David Ha, Llion Jones and Ren Ito, builds models for Japanese language and culture. It was valued at $2.65 billion in November 2025, Japan’s most valuable startup unicorn.

Bengaluru. India’s government picked Sarvam AI from a shortlist of 67 companies to build the country’s first sovereign language model. In February it unveiled models with 30 and 105 billion parameters, and India’s sovereign models now cover 22 Indian languages.

Abu Dhabi. In January the Technology Innovation Institute released Falcon-H1 Arabic. TII says its 34 billion parameter model tops the Open Arabic LLM Leaderboard and beats systems more than twice its size.

Jakarta. Indosat and GoTo built Sahabat-AI, a 70 billion parameter model that works in Bahasa Indonesia, Javanese, Sundanese, Balinese and Batak, for a country of some 17,000 islands.

Johannesburg. Pelonomi Moiloa and Jade Abbott cofounded Lelapa AI to build models for African languages like Swahili, Yoruba, isiXhosa, Hausa and isiZulu, spoken by more than 350 million people.

They crossed an ocean to build it

Taiwan to America. Jensen Huang and Lisa Su, two of the biggest names in AI chips, were born in Taiwan and came to the United States as children. Huang cofounded Nvidia at 30. Su earned a PhD at MIT and has run AMD since 2014.

Ho Chi Minh City to Sydney. Brothers Dimitry and Aengus Tran were taught to code by their father in Vietnam, then moved to Australia. In 2018 they founded Harrison.ai, which reads X-rays and CT scans. Dimitry says half of Australia’s radiologists use it, in more than 20 countries.

London. Demis Hassabis, the son of a Greek Cypriot father who owned a toy shop and a Singaporean mother of Chinese descent, won the 2024 Nobel Prize in Chemistry for using AI to predict protein structures.

They built it where they were told they couldn’t

Paris. Arthur Mensch, Guillaume Lample and Timothée Lacroix founded Mistral in 2023 on a simple pitch: governments and companies should not depend on AI controlled from the US or China. In September it raised €3 billion at a valuation above €21 billion, the largest equity round in European tech history.

Hangzhou. When Liang Wenfeng started DeepSeek in 2023, venture firms were reluctant to fund it. His own hedge fund paid for it instead.

Tunisia. InstaDeep, cofounded by Karim Beguir, was acquired by BioNTech in 2023 for up to $680 million.

São Paulo. Enter, a legal AI company founded in 2023, raised $100 million at a $1.2 billion valuation and became Latin America’s first AI unicorn.

The odds

None of this came easy, and the numbers say so.

All women founding teams received 2.3 percent of global venture capital in 2024. Across 27 countries, women hold 13 percent of C suite AI jobs at AI companies. Black founders raised $942 million in the US in 2025, 0.32 percent of the venture total. African startups raised about $1.36 billion in the first half of 2026, and truly AI native companies got less than two percent of it.

Those are the odds. They are not the verdict.

Who beat them

Immigrants cofounded 25 of the 42 US based companies on the 2025 Forbes AI 50, drawn from 25 countries.

Forbes put four women led companies on its 2026 AI 50: Mira Murati’s Thinking Machines Lab, Fei-Fei Li’s World Labs, Minna Song’s EliseAI and Fireworks AI, run by Lin Qiao, who was born in China. Harper, an AI insurance brokerage cofounded by Dakotah Rice, raised a $47 million Series A that the company says is the largest ever publicly disclosed for a Black founder.

Four of fifty is a small number. They are on the list.

AI moved the ceiling

Here is the part nobody says plainly. AI lowered the cost of starting a company. It did not lower the cost of scaling one. Crunchbase’s head of research points to what holds founders back: access to networks, relationships and early introductions. Not talent.

So the ceiling moved. It used to sit at can you build it. Now it sits at who takes your call.

That one can be worked on. Build a network on purpose, ship in public and collect proof before anyone gives you permission. The tools do not check your passport. Stanford’s 2026 AI Index puts the UAE first for generative AI use, with 64 percent of people using it, ahead of Singapore at 61 percent. The United States, which invests more in AI than anyone, ranks 24th.

Bring all of it

Every founder here brought a country with them: a language, a family, a picture of what a good life looks like. Build as yourself. The sameness was never the asset.

Start where you have an advantage nobody can copy: your language, your market, your people. Get in the rooms where the introductions happen. Show the work.

When someone tells you what you can’t reach, ask for their sample. They are working from the people they have met. You are not in that sample yet.

The people on this page weren’t in anyone’s sample either. They built the sample.

Look at who built it. Then go bust the ceiling.

Andrew Quillen is the founder of AndMaverick, a global Enterprise AI Orchestration consultancy. To continue the conversation, visit andmaverick.com.