The evolution:

• Alipay (2004) — Started as an escrow service for Alibaba's marketplace. Solved the trust problem between online buyers and sellers. This created the data pipeline: every transaction generated behavioral data.
• Yu'e Bao (2013) — Allowed users to invest idle Alipay balances in a money market fund. Became the world's largest money market fund within two years. Leveraged existing user base and payment data.
• MYbank (2015) — Online-only bank focused on small business lending. Loan decisions in 3 minutes. Zero human underwriters. Default rate below 1% (better than traditional banks with weeks-long review processes).
• Zhima Credit (Sesame Credit) — Credit scoring system using transaction data, payment history, social connections, and spending patterns. Extends beyond traditional credit bureau data.
• Insurance, Wealth Management, International Expansion — Each new product built on the same AI factory infrastructure.

Why it works — the AI factory in practice:

• Shared data pipeline — A loan application draws on transaction history from Alipay, spending patterns from partner merchants, payment behavior, and social graph data. No traditional bank has access to this breadth of real-time behavioral data.
• Shared algorithm infrastructure — The same ML platform that powers fraud detection also powers credit scoring, risk assessment, and product recommendations. Building one model improves all the others through shared learnings and infrastructure.
• Marginal cost approaching zero — Once the AI factory is built, serving a new customer or processing a new loan costs almost nothing. MYbank can profitably serve customers that traditional banks consider too small or too risky.
• Continuous improvement — Every transaction, every loan repayment, every default feeds back into the models. The system gets better every day without human intervention.

The key lesson:

Ant didn't build separate businesses. It built one AI factory that powers multiple financial products from the same data and intelligence layer. This is fundamentally different from a traditional financial conglomerate where each business unit has its own processes, data, and technology. The shared factory creates compounding advantages: more products generate more data, better data improves all products, better products attract more users, more users generate more data.