CZ: AI Integration May Prioritize Stablecoin Transactions
Binance founder Changpeng Zhao stated at Bitcoin Asia 2026 in Hong Kong that the integration of artificial intelligence with cryptocurrency may start with stablecoins; once stablecoins are accepted, it will be much easier to add Bitcoin, BNB, Ethereum, Solana, and other chains. Major AI companies may also issue their own tokens. He has discussed issuance methods with several top AI companies.
The discussion began with the cost of computing power. Building 1 gigawatt of computing power requires about $30 billion to $50 billion; some companies want to build hundreds of gigawatts in the coming years, with expenses in the trillions, although valuations are high, cash is insufficient. One idea is to issue data center tokens, allowing holders to gain rights to use computing power or rewards in the future. If tokens are issued, efforts will be made to enhance utility: using tokens for subscriptions, then linking different language models and services. Worldcoin, which has a close relationship with OpenAI, has been cited as a precedent, although its current utility is still limited.
Payments are not a priority for these companies today. He mentioned that they are making agents smarter, helping users find better trades; after booking hotels and paying bills, most people can still use their credit cards. Trading is more suitable for AI: it must quickly gather information and react to news and charts, potentially increasing efficiency tenfold; analyzing and then placing orders manually would miss opportunities, so agents are preferred to place orders directly. His judgment is that AI will first help people trade, then help them pay. In the future, billions of agents will automatically buy, sell, and negotiate, using cryptocurrency, likely starting with stablecoins.
He still primarily positions Bitcoin as a savings tool, while stablecoins or other coins will handle machine-to-machine commerce and transactions. Long-term Bitcoin holders are reluctant to convert to stablecoins pegged to fiat currency for payments, adding an extra step; as payment volumes increase, some scenarios may revert to using native cryptocurrencies. During the Hong Kong conference, capital and attention continue to flow heavily into AI infrastructure, with organizers stating that the industry feels "smaller," but Zhao translates the computing power gap directly into a demand for token issuance.
In market mechanisms, this turns electricity bills into a token entry point. Buyers are AI companies lacking funds to expand data centers and speculators needing future computing power; sellers are traditional payment channels that can only use credit cards. Funding expectations shift from equity financing, which cannot meet gigawatt expenses, to a pre-sale structure of "holding tokens for computing power," then entering trading agents through stablecoins. Beneficiaries are exchanges that can provide both stablecoin settlement and token listing channels; those under pressure are wallet projects that treat AI payments as near-term narratives and brokers that must first prove agents will place orders themselves. The event-driven factor is the reported price of computing power at $30 billion to $50 billion per gigawatt.
Source: Public Information
ABAB AI Insight
Zhao frames the electricity shortage in AI as a scenario where cryptocurrency can be monetized. The $30 billion to $50 billion per gigawatt is the denominator for the story of turning Nvidia orders and land electricity into tokens. AI companies have high valuations but low cash, and equity financing cannot keep up with factory construction speed, leading to the emergence of "data center tokens": collect tokens first, then provide machine time. This is not a payment revolution, but a project financing rebranding. Stablecoins are placed at the first level because agents need a clear settlement dollar leg, not a store of value faith.
The capital path has two steps. The first step is to enable agents to place orders, increasing trading frequency and fees; the second step allows agents to pay for hotels and bills themselves. Worldcoin is mentioned because the issuing team is close to the model company, not because the iris has become a universal currency. Bitcoin remains locked in the savings layer to avoid competing with stablecoins for payment entry. The position of exchanges is: whoever first allows AI companies to receive stablecoins, launch computing power tokens, and enable agents to place orders will collect three tolls.
Analogies include power grid capacity contracts pre-selling future electricity and game companies binding points to props. ICOs once financed protocols without products; data center tokens are pre-securitizing cabinets and electricity costs. PayPal once facilitated Bitcoin payments but stopped at the reality that "people can still use credit cards." The current phase is a transformation: AI companies shift from selling subscriptions to selling future computing power, while cryptocurrency shifts from selling coins to serving as a clearing layer for agents.
Structural judgment belongs to the financing reconstruction when technological substitution is not yet complete. Agents are not yet allowed to make payments on a large scale, but they may be permitted to trade themselves; payment lags because credit cards are sufficient for humans but not for millisecond-level orders. The mechanism is: the cash gap in computing power expansion forces token issuance, and tokens must be tied to utility, with utility initially falling on trading frequency rather than retail payments. When electricity is more expensive than wallets, tokens sell future machine time, not future coffee.
ABAB News · Cognitive Laws
- First let machines place orders, then let machines pay.
- The electricity cost of one gigawatt can force token issuance more than ten payment visions.
- Stablecoins are the pocket money for agents, while Bitcoin remains the savings that agents are reluctant to spend.