11 年，110 亿美金，然后呢？｜对话 Airwallex 吴恺：AI 时代，下一站 1000 亿

Summary

This Shizilukou Crossing episode interviews Wu Kai / 吴恺 of Airwallex about the company’s path from a 2015 Melbourne cross-border payment pain point to a source-scoped $11 billion valuation and a next-stage AI finance strategy. The source frames Airwallex as a Global Financial Network that moved beyond money transfer into accounts, FX, cards, acquiring, yield, credit, APIs, and finance workflow software. Its AI-era thesis is Intelligent Finance: products such as Kai (Airwallex), Airwallex Agent OS, T0 Finance, and ARID may turn financial operations, agent payments, and finance SaaS functions into a larger intelligent entry point.

Key Claims

• Airwallex began in 2015 in Melbourne after its founding team experienced slow, expensive cross-border payments while importing coffee beans for a cafe.
• Wu Kai / 吴恺 argues that the company’s important early choice was building a global network from day one rather than a single-region network for Southeast Asia, the Middle East, or Latin America.
• The source says Airwallex spent its first three to four years on product, bank partners, local licenses, and clearing-network access before revenue became meaningful around 2018-2019.
• The episode describes Airwallex as supporting real-time payment into more than 90 countries, making local banking and clearing access central to its Global Financial Network position.
• Wu argues that money transfer alone is a weak business because it is too low-level; Airwallex reached a stronger product-market fit only after adding account, FX, payout, card, acquiring, yield, and credit products.
• Traditional banks are presented as weaker on global reach and product experience, while Airwallex emphasizes web, mobile, API, ERP integration, and programmable finance workflows.
• The source says AI companies create more dynamic billing problems than classic subscriptions because model type, performance, usage tier, time-of-day pricing, and customer usage data can change quickly.
• Kai (Airwallex) is described as an embedded natural-language assistant for policy analysis, liquidity optimization, expense controls, workflow generation, and error analysis.
• Airwallex Agent OS exposes Airwallex capabilities through command-line and API/MCP-style interfaces so customer agents can call financial actions directly.
• T0 Finance is presented as an AI-native finance platform that can do bookkeeping, reconciliation, reports, policy work, cash-flow forecasting, and revenue/expense forecasts for young companies.
• ARID starts from one-click checkout and is framed as a possible consumer wallet or agent-to-agent payment layer as Agentic Commerce matures.
• Wu argues that financial SaaS is vulnerable to AI consolidation because narrow tools each hold only partial data, while better automation needs broader transaction, workflow, and policy context.
• The source says Airwallex’s AI-era acquisition logic is mainly data and talent: proprietary workflow data can train agents, and strong product founders may become more valuable as code generation lowers implementation cost.
• The episode presents source-scoped valuation support around $1.3 billion ARR, roughly 90% growth, about 70% gross margin, and an 8x PS multiple.
• For globalization, Wu emphasizes local general managers and market-size-based entry order, eventually using GDP ranking to prioritize new markets.

Key Quotes

“只做底层资金转移不够” — Wu Kai’s reason for expanding beyond payment rails.

“从第 0 天开始就能拥有一个 mini CFO” — Wu’s explanation of the T0 product idea.

“数据和人才” — the source’s summary of Airwallex’s AI-era acquisition logic.

Connections

• Airwallex and Wu Kai / 吴恺 — company and guest anchoring the episode.
• Kai (Airwallex), Airwallex Agent OS, T0 Finance, and ARID — Airwallex products used to explain its AI finance strategy.
• Global Financial Network, Money Movement Infrastructure, Payment Clearing Network / 支付清算网络, and API Product Design — infrastructure layer behind the company’s original cross-border payment wedge.
• Intelligent Finance, Financial AI Agents, Business-Led AI Transformation, Agent-Facing Interfaces, and AI Native SaaS Threat — AI-era finance and software-shape thread.
• Agent Payment Infrastructure / 智能体支付基础设施, Agentic Commerce, Agent Spend Controls / 智能体消费控制, and Model Context Protocol — agent payment and agent-callable product surface context.
• Stripe, Shopify, DeepSeek, ChatGPT, OpenAI, Visa, and Mastercard — comparison points and market actors mentioned or adjacent to the source’s claims.

Contradictions

• No direct contradiction found. The source reinforces Money Movement Infrastructure and Agent Payment Infrastructure / 智能体支付基础设施 by showing an enterprise-finance version of payment rails becoming agent-callable.
• Tension with SaaS Trust Moat and AI Native SaaS Threat remains compatible: Wu argues that narrow finance SaaS tools may be consolidated by broader agents, while existing SaaS sources argue that trust, compliance, workflow depth, proprietary data, and systems of record defend stronger incumbents. Airwallex’s thesis fits the latter if its own regulated data and transaction layer become the defensible system of record.