Binance founder Changpeng Zhao, or CZ, said at the Bitcoin Asia 2026 conference that AI and crypto may first come together through stablecoins before expanding into broader digital asset use.
According to CZ, once a system can accept stablecoins, adding Bitcoin becomes much easier. He said the same could then apply to BNB, Ethereum, Solana and other blockchains.
CZ also said some large AI companies may want to issue their own tokens, which would make crypto integration a practical step. He said he has spoken with several top AI companies about how token issuance could work, and that those discussions have produced many different ideas.
Token issuance discussions tied to data center funding
CZ described the sector as still being in a very early stage. He said building data centers requires large amounts of capital, putting the cost of 1 gigawatt of compute at about $30 billion to $50 billion. He added that many AI companies have high valuations but no longer have enough funding to build compute capacity at that scale.
He said some AI companies he has spoken with want to build hundreds of gigawatts of compute over the next few years, a plan he said would require trillions of dollars. One structure discussed in that context is a data center token, where token holders could later receive rewards from using the data center.
CZ said any company issuing a token would likely want to maximize its utility. That could include allowing the token to pay for subscription services and linking it to different language models and different services. He said there are many ways those models could be combined, and added that AI companies issuing and promoting their own utility tokens is not hard to imagine.
CZ points to Worldcoin
He said the market has already seen token launches from companies with very close ties to OpenAI, citing Worldcoin as an example. While he said the token may not have much utility today, he added that the people behind it are very close to AI companies.
From that, CZ said crypto will have a role in AI systems. When AI handles payments on behalf of users, he said, it will use cryptocurrency.
AI trading comes before AI payments, in his view
Still, CZ said AI payments are likely to come after AI trading. He used examples such as AI agents booking hotels or paying bills, but said that based on his conversations with some top AI companies, payments are not their priority today.
Instead, he said those companies are trying to make agents smarter so they can help users find the best deals. In the hotel example, once an AI finds the best option, most people can still book it themselves with a credit card.
CZ argued that AI may be a better fit for trading because trading requires very fast collection of large amounts of information, quick reactions to news, and chart analysis. He said AI could improve the efficiency of that process by roughly 10x, regardless of the trading approach.
He added that in that setup, users would not want AI to stop at analysis while leaving the final execution to a person pressing the button. For efficiency and timing reasons, he said, users would want AI to place the order directly.
CZ said AI is therefore more likely to help users trade first, and handle payments later.