"Sovereign AI" has become the buzzword of choice across the global tech sector, at least outside the U.S. and China. Officials and companies worldwide are increasingly wary of tying their economies too closely to one country's AI systems.
What sovereignty means shifts by geography. In Europe, it centers on data, keeping personal information within borders. In the Middle East and much of Asia, it means building local industry. For smaller economies, it means autonomy — insurance against being cut off by a foreign supplier.
"AI has become such an essential need, and so you don't want to be tethered to anybody else who can turn it off," said Pak-Sun Ting, founder of Hong Kong-based Votee AI.
Votee's flagship product is an AI model built for Cantonese, the dialect spoken in Hong Kong and Guangdong province. "The whole AI revolution is in English and Mandarin," Ting said. "Cantonese is used in education, health care, and police communications. If those don't get covered, then AI is essentially useless."
Votee is not alone. Indosat, one of Indonesia's largest telecoms companies, is building Sahabat AI, a model focused on Bahasa and other Indonesian languages. South Korean companies are competing in the government-sponsored "AI Squid Game" to build the best homegrown model. And Humain, an AI company backed by Saudi Arabia's Public Investment Fund, just debuted an Arabic-language model built by Chinese developer MiniMax.
None of these are small markets. Korean and Cantonese each have around 80 million speakers. More than 200 million people speak Bahasa Indonesia. Yet these languages lack the massive text libraries that fuel English- and Mandarin-language models.
The capital math is the real story. Ting said Votee trained its model for roughly $250,000 — real money, but a rounding error next to the tens of billions of dollars poured into frontier models by Anthropic and OpenAI. Governments, the likeliest customers for sovereign models, don't need the most powerful systems on the market. Restrained ambitions keep costs down. A growing supply of free, open-source Chinese models gives smaller players another way to build without writing a nine-figure check.
The numbers come first, and here they tell a simple story: sovereignty doesn't require a state-run mega-project. It requires capital discipline and the freedom to shop among suppliers — American, Chinese, or otherwise — rather than being locked into one.
That is the real lesson for Washington and Beijing alike. The nations building Cantonese, Bahasa, and Arabic AI models aren't rejecting American or Chinese technology; they're refusing exclusive dependence on either. Capital rewards clear rules and open markets, not gatekeepers. For a $250,000 outlay, a small firm in Hong Kong just proved that point better than a government white paper ever could.
