On August 17, Taiwan's president Lai Ching-te announced that the 2027 budget will set aside NT$235.7 billion — roughly $7.4 billion USD. Every resident gets NT$10,000, about $314.
> "The AI dividend should go to everyone." — Lai Ching-te, President of Taiwan
The budget still needs parliamentary approval, and the opposition holds the majority there. Local elections in November complicate things further.
Where the money comes from
Taiwan's economy is growing at its fastest pace in 39 years, as the whole world buys up AI chips. The main producer is the local company TSMC, and its tax payments feed the state budget. The government's logic is simple: if the country got rich off tech, everyone should get a cut — not just people working in semiconductors.
Paying out to 23 million residents will cost a third of what TSMC made in Q2 alone — its net profit came in around $22 billion. For scale: TSMC CEO C.C. Wei's 2025 compensation hit $76 million — a record for Taiwanese companies, and the equivalent of a one-time payout to 240,000 people.
Not just Taiwan
The idea of spreading windfall profits across the population isn't unique to the island. Elon Musk, for one, has long pushed the concept of "Universal High Income" — a theme that also shows up in his 2026 predictions. His argument: AI and robotics will usher in an age of abundance where physical labor stops being necessary, and the state covers people's basic needs.
US President Donald Trump also promised dividend checks for every American — though not funded by neural networks. In fall 2025 he announced $2,000-per-person checks, but named the funding source as record tariff revenue, not the AI boom (and we know how that ultimately turned out — spoiler: it didn't). Still, the mechanism itself — the state handing windfall profits straight to citizens — is becoming a global political trend.
The data problem
I looked at how AI profits get split elsewhere. Taiwan has a clean distribution mechanism: chips are countable, factories sit in one specific country, and corporate profit gets taxed transparently.
The other resource AI runs on — data — doesn't work that way. Neural networks trained on text and images people posted online for decades. None of those authors got a cent for it.
One of the rare exceptions came together on July 20. A court approved Anthropic's settlement with authors: the company will pay $1.5 billion for roughly 500,000 books — about $3,000 per book. But that compensation was for the illegal way the data was obtained — the books were downloaded from pirate sites. The court ruled that training models on other people's text is itself legal, and free.
Taiwan's move is an important precedent. For the first time, a government has written into its budget the idea that AI boom profits belong to all citizens. From here, the political fights will just be about the size of the payout — and $314 looks like nothing more than an opening bid.