Binance's He Yi Puts $100 Million a Year Behind AI That Buys Bitcoin (BTC) on a Single Instruction

A $100 Million Annual Baseline

Binance co-founder He Yi laid out the exchange's all-in AI strategy during a Chinese-language AMA held on Tuesday, October 6, as part of Binance's smart product launch event. Her framing was financial parity: present the analytical methods the industry already uses, from on-chain analysis to investment research, in a structured form that a user with zero background can follow, so the gap between people who understand investing and people who do not keeps shrinking. The full session is available as a Chinese-language AMA replay on Binance Square.

She rejected the idea that financial AI is decided by models and coding skill alone. Security, proprietary data and sustained capital carry equal weight, and she put a figure on the capital part: in finance, she argued, a net investment of $100 million per year works as an entry bar. Anyone can assemble a functioning strategy with public code, she said, but clean financial data is scarce, and Binance's nine years of accumulated history give it a base few startups can match. In an industry where one security failure wipes out everything else, the platform's layered defenses, from cold wallet custody to live platform protections, are part of the AI stack itself.

The product vision reaches past the trading terminal. Wealth management is daily life and payments are routine, she said, and Binance has already widened its trading menu beyond crypto assets to gold and stock-linked products. She compared today's crypto interfaces to candle factories in the age of electric light, cited Apple's iPad and TikTok as products that erased usage barriers, and said financial software should do the same through customizable homepages and voice input. For users in high-inflation countries, holding stablecoins is itself a form of financial access, and blockchain plus AI should let someone invest as little as $5 in equities. Years of framing the mission as beating Wall Street have given way, in her view, to convergence with traditional finance.

From Chat to Executed Orders

The most concrete capability He Yi described is conditional execution. A user tells the AI to buy a fixed amount when the Bitcoin price reaches a level they set, and the system handles the order without anyone touching a trading screen; the function already works, she said, and the priority is making it smoother. Voice input and output are part of the same push, on the logic that typing is a learned skill the interface should not require. Advice comes next: the AI would read whether a user is a long-term holder or an active trader, whether they prefer dollar-cost averaging or grid methods, then propose structured strategies for the user to confirm with one tap.

Personalization will not scale with account size or VIP status, she said. Binance plans to sort users by interests and behavior, in the way TikTok tags interests, and she encouraged people to customize their homepages so the platform learns from real usage rather than a default template. On whether customer positions would be used to train models, she answered that Binance's data mining has not reached that depth yet and that an opt-out setting would be considered.

Trust was a running theme. She said Binance earns from trading fees and does not take the opposite side of customer positions, while warning against reading trade signals into her posts or CZ's: last cycle's winning tokens did not run on tweet-reading, and she admitted she chased FOMO when she began investing in 2008. Everyday payments, she added, should eventually be as effortless as the card services from Visa. She closed the session by confirming she is pregnant again, with her workload unchanged.

The $100 million figure is the strategic core of the session. He Yi is telling AI-native startups that model quality alone will not win financial AI, and pointing at capital, nine years of proprietary data and custody infrastructure as the moat instead. For retail users, the near-term change is practical: conversational order placement and preference-based strategy suggestions cut the friction that keeps beginners on the sidelines. Among the Best Crypto Exchanges, product usability is becoming a competitive field in its own right, and the two open questions from the AMA, when a data-training opt-out ships and how quickly conditional execution improves, are the ones worth tracking.