Binance Founder CZ: Hot Money is Flowing Back from AI to Crypto
Binance founder Changpeng Zhao stated that some hot money is flowing back from the artificial intelligence sector to the crypto market, emphasizing that the currency industry will not disappear and that individuals and AI will still need funding in the future.
He described the capital return as a phase reallocation rather than a permanent exit of assets. Six months ago, he explained to English media that during the downturn of crypto in the first half of 2026, he listed geopolitical tensions, the four-year cycle, and capital shifting to AI together. At that time, Bitcoin fell about 50% from its peak of over $126,000 in October 2025, after starting the year near $89,000 and briefly touching $96,000 before dropping to around $60,000.
He has long viewed crypto and AI as sequential waves rather than zero-sum opponents, stating that the withdrawal of hot money from new industries is positive in the long run. His investment platform YZi Labs publicly allocates about 70% to crypto, 20% to AI, and 10% to biotechnology, with the AI side leaning more towards "shovel" assets like data centers and power, rather than directly betting on application layer model companies.
He repeatedly emphasized that AI agents require a programmable, cross-border, callable interface for settlement layers, noting that credit cards lack native APIs, making blockchain more suitable for machine-to-machine transfers. Binance has launched Agent OS, which integrates market data, payments, on-chain data, and trading capabilities into workflows, advocating that exchanges should not block AI orders but should open tools to all users.
Most of his net worth is concentrated in BNB, and his statements align with the platform's liquidity, on-chain payment narrative, and compliance expansion in the U.S. He previously discussed how tokenization could serve as a channel for countries and enterprises to attract external funding, viewing stablecoins as the first entry point for the combination of AI and crypto, with agency trading expected to precede agency payments.
In market mechanisms, this is a narrative-driven style rotation: short-term funds are withdrawing from overvalued AI growth stocks and computing power themes, seeking volatility and trading turnover; sellers are taking profits from AI-themed positions, while buyers include exchanges, market makers, and on-chain speculators. The event itself does not create new savings but changes risk preferences, benefiting trading volume and platform tokens, while putting pressure on application-layer AI projects that still rely on storytelling for financing and have yet to realize cash flow.
Source: Public Information
ABAB AI Insight
After making Binance the world's largest spot and futures trading venue, Changpeng Zhao pleaded guilty to anti-money laundering and sanctions compliance issues and spent several months in prison. The company paid approximately $4.3 billion in fines, and he was later pardoned while retaining the majority of his equity. Since then, he has not been involved in the daily operations of the exchange but continues to set industry pricing through public judgments: on one hand, promoting YZi Labs with a distribution of 70% crypto, 20% AI, and 10% biotechnology, while on the other hand, turning the educational project Giggle Academy into an asset export, repeatedly transferring BNB out and allowing it to be converted into fiat currency for expenses. This path follows the early strategy of "low fees to capture volume, rapid listing, and binding ecosystem with platform tokens": first generating traffic, then creating narratives, and finally sending the narratives back to their own settlement layer.
Money has not disappeared from AI; it has shifted from "training models" to "who pays for the models." The cost of a data center per gigawatt is estimated by him to be in the range of $30 billion to $50 billion. Leading model companies have high book valuations, but the cash available for infrastructure is insufficient to support hundreds of gigawatts of planning, leading to ideas like data center tokens, computing power sharing, and stablecoin settlements. The motivation is straightforward: AI companies lack financing tools that are divisible, pre-sellable, and cross-border, while crypto provides a channel for the securitization of future computing power usage rights; strategically, it aims to make BNB Chain and exchange APIs the default clearinghouse for the agency economy, rather than competing with Nvidia for chip narratives.
Similar structures have appeared twice before: in 2017, ICOs packaged "protocols that have not yet generated revenue" into tradable equity, and from 2020 to 2021, DeFi turned transaction fees and locked-up earnings into a money-printing machine that could attract hot money. This time, the benchmark is the capital expenditure cycle represented by Nvidia, OpenAI, and Anthropic, as well as institutions like BlackRock describing the integration of TradFi and on-chain clearing rather than replacement. The industry phase has shifted from simply expanding user numbers to competing for entry points and pricing power in machine trading—whoever controls agency ordering, stablecoin settlement, and compliance entry will collect tolls.
This is a transfer of pricing power. Hot money flows back and forth between AI and crypto, not because one side suddenly becomes more "correct," but because risk assets share the same batch of leveraged funds, and after the narrative climax, capital must seek the next venue for rapid turnover. The pricing power on the AI side still lies in chips, electricity, and closed-door financing for models, while the pricing power on the crypto side is in trading volume and clearing networks; when financing becomes difficult for model companies and computing power returns lengthen, short-term funds will reinterpret "machines also need money" as trading volume. The mechanism is: whoever can provide high turnover and programmable settlement simultaneously will act as the cash register in the rotation.
ABAB News · Law of Cognition
Hot money does not disappear; it just changes lanes for fees.
Machines cost money; the settlement layer has pricing power.
Narrative rotation earns tolls, not truths.