Stripe Debuts AI Commerce Standard as Processing Volume Hits $1.9 Trillion

Stripe 处理量达 1.9 万亿美元，推出 AI 商务标准 Stripe 正在全面重构其金融网络以赋能自主式 AI Agent，并认为随着全球企业商务向 Agent 驱动的执行模式转变，可编程支付轨道和机器间协议现已不可或缺。 Stripe Sessions 2026 针对围绕人工智能和自主 Agent 重构的经济体系，推出了一系列雄心勃勃的产品组合。主要发布内容包括机器支付协议（MPP）、针对 Agent 的 Link Wallet、Agentic Commerce Suite，以及与 Paradigm 合作构建用于实时支付流处理的 Tempo 区块链。除 Stripe Projects 和 Stripe Console 等 AI 原生工具外，主题演讲还重点介绍了 Stripe Treasury、Radar 反欺诈预防以及平台变现工具等方面的重大更新。 Stripe 目前在全球服务超过 500 万家企业，支付处理量达 1.9 万亿美元（同比增长 34%），其支撑的商业体量约占全球 GDP 的 1.6%。 (0:00) 随着全球经济围绕人工智能进行重塑，2026 年初 Stripe 上的新业务创建呈现出巨大的抛物线式爆发增长拐点。 (2:52) Stripe Projects 正式全面开放（GA），允许 AI Agent 直接从命令行以编程方式配置主机托管、数据库及计费服务。 (5:46) Stripe 推出机器支付协议（MPP）和适用于 Agent 的 Link Wallet，通过用户生物识别授权实现安全的机器对机器商务交易。 (11:21) Stripe 发布了 Agentic Commerce Suite，并宣布与 Google 达成重大合作，使用户能够在 Google AI 模式和 Gemini 应用中直接无缝购买商品。 (15:03)

Stripe leadership argues that the global economy is entering a structural re-platforming around artificial intelligence, requiring programmable financial infrastructure capable of supporting autonomous agents, real-time compute billing, and stablecoin payment rails. Autonomous Agent Commerce and Transaction Protocols (5:46) Developer interaction with Stripe shifted markedly in early 2026, marked by an exponential rise in Stripe CLI activity as automated agents began interacting directly with financial APIs. To support automated software development and deployment, Stripe introduced Stripe Projects, which enables agents to provision hosting, databases, and environment credentials directly from the command line through partners such as Vercel and Supabase. To bridge the gap between machine operations and commercial exchange, Stripe introduced the Machine Payments Protocol (MPP) alongside the Link Wallet for Agents. MPP establishes an open HTTP standard for servers to communicate price specifications to agents, while Link cryptographically isolates payment credentials and requires explicit user biometric approval on mobile devices before transactions settle. For consumer commerce, Stripe launched the Agentic Commerce Suite and joined the Universal Commerce Protocol (UCP)—collaborating with Google, Shopify, Meta, Amazon, and Microsoft—to make merchant inventories legible to AI search and checkout interfaces. Real-Time Token Billing and Micro-Payment Streaming The rapid migration toward usage-based and token pricing models has created distributed systems challenges, particularly the financial risk of unpaid high-velocity GPU compute consumption. Stripe integrated Metronome to handle real-time metering, rating, and alerting, dynamically tracking token burns to halt execution before users exceed prepaid balances. To prevent service interruptions without extending credit risk, Stripe collaborated with Paradigm to build Tempo, a payment-focused blockchain. Tempo allows software agents to execute continuous, sub-cent stablecoin micropayments (such as $0.00003 per batch) synchronously as tokens are consumed. Network partners validating or integrating with Tempo include Visa, Shopify, Standard Chartered, Klarna, and DoorDash, forming an alternative micro-transaction layer to card networks and ACH. AI Fraud Defense and Network Data Signals (35:08) The commercial expansion of AI platforms has altered fraud patterns, shifting bad actors from traditional payment theft toward token compute abuse. Stripe reported that one in six sign-ups across AI businesses on its network involved multi-account abuse, where fraudsters repeatedly exhausted free trial credits. This dynamic previously inflated acquisition costs for some startups to more than $500 in inference overhead per acquired customer. Stripe Radar was updated to evaluate multi-account signals and pay-as-you-go default risk prior to compute consumption, blocking over 3.3 million fraudulent trial attempts across eight AI firms in a single month. To externalize this network intelligence, Stripe launched Stripe Signals, exposing an API that translates 70 trillion commercial data points into programmable risk, trust, and conversion scores for external platforms like DoorDash, Cursor, and Cloudflare. Programmable Banking and Embedded Financial Operations (55:56) To streamline back-office operations for high-growth global firms, Stripe relaunched Stripe Treasury as a financial operating system supporting multi-currency fiat accounts alongside stablecoins across 160 payout destinations. Leveraging transaction density among its 5 million merchant accounts, Stripe introduced instant, zero-cost transfers between internal Treasury balances, demonstrating programmatic execution via the Model Context Protocol (MCP) in conversational interfaces. For enterprise software providers using Stripe Connect, the platform launched the Platform Growth Studio, which features monetization simulators and peer benchmarking. Concurrently, Stripe Capital introduced flexible lines of credit, and Stripe Issuing expanded its capabilities to support digital cards issued directly to autonomous AI agents, as well as stablecoin-backed card programs operating over the Visa network across 100 countries. Stripe's strategic roadmap positions programmable APIs, machine protocols, and stablecoin networks as the default operating architecture for agentic commerce; however, long-term industry adoption depends on widespread merchant integration of protocols like MPP and UCP, regulatory alignment across international payment jurisdictions, and the operational reliability of delegating spending authority to autonomous software.

Stripe管理层在2026年年度大会上指出，全球商业正迎来向人工智能架构重组的关键拐点，货币的流动本质正在完全数据化。Stripe公布了一系列核心基础设施与协议，旨在为自主AI代理交易、实时算力微支付及全球嵌入式金融网络奠定经济底座。 自主AI代理商业与机对机支付协议 (11:21) Stripe指出，开发者在2026年初开始密集使用AI代理调用金融API，推动命令行界面（CLI）使用量出现垂直式暴增。为解决生成代码与实际部署脱节的问题，Stripe全面开放了Stripe Projects，允许AI代理通过命令行直接配置Vercel托管与Supabase数据库等服务，无需人工复制密钥或切换操作后台。针对传统结账页面仅适配人类操作的局限，Stripe与Tempo合作推出了基于HTTP的机器支付协议（MPP），使服务器能够直接向浏览网页的AI代理下发支付指令；同时上线Link代理钱包，在隔离敏感凭证的前提下为代理分配支出预算，每笔交易仍须经由用户移动端生物识别授权确认。 面向消费者端，Stripe推出了代理商业套件（Agentic Commerce Suite），使商户的目录数据能被AI代理直接读取与索引。Stripe宣布与Google建立战略合作，支持消费者在Google AI模式和Gemini应用中直接完成商品结算；Shopify作为首选目录合作伙伴接入该套件，向AI代理开放数十亿款商品库存。此外，Stripe正式加入由Shopify与Google共同发起的通用商业协议（UCP）委员会，与Meta、Amazon及微软等企业共同推进AI商务标准的统一，协助商户在保留原路结算与风控体系的同时承接代理流量。 Token计费与Tempo区块链微支付流 随着AI业务全面转向按Token消耗量、实际产出与混合模式计费，传统按月计费和静态授信机制无法应对代理高频消耗算力带来的坏账风险。为此，Stripe整合了新近收购的Metronome计费引擎，提供实时的Token计量、动态费率核算与自动熔断预警，在客户预付额度耗尽时自动阻断执行，防止算力被无担保透支。 为避免中途切断任务损害用户体验，Stripe联合Paradigm推出了专为支付与稳定币设计的高性能区块链Tempo。现场演示表明，AI代理在审查API时可通过Tempo以稳定币进行微支付流式传输，实现数千笔低至0.00003美元的亚美分级交易实时即结，构建出“随算力消耗即时结算”的商业模型。目前，DoorDash已采用Tempo支付全球骑手薪酬，Klarna基于该网络发行了原生稳定币Klarna USD，而Visa与Shopify则正式作为验证节点加入Tempo网络。 抵御AI算力盗用与网络级信号开放 (35:08) Stripe指出，AI时代的欺诈焦点已从传统的盗刷信用卡转向盗用GPU推理算力与Token。数据显示，AI企业新注册用户中每6个就有1个涉及多账户滥用，欺诈者通过反复注册耗尽免费试用额度；在过去六个月中，此类试用欺诈增长了一倍以上，导致部分AI企业每获取一名有效付费用户需承担25次虚假试用、折合超500美元的无效推理成本。对此，Stripe升级了Radar风控引擎，通过设备指纹与全网信号在前置环节拦截滥用，单月内在8家AI企业中拦截了330万次欺诈试用，并对先用后付模型中的中途逃单行为进行动态预警与限流。 基于支撑全球约1.6% GDP、处理1.9万亿美元交易额（同比增长34%）所沉淀的70万亿商业数据点，Stripe推出了Stripe Signals可编程API。该服务向Cursor、Cloudflare、DoorDash及ElevenLabs等首批合作伙伴开放实时的风险、信任与转化评分，使企业能够跨平台识别恶意行为并优化结账漏斗。 全球金库升级与AI代理资金托管 (55:56) 面向跨国扩张的企业，Stripe重新发布了Stripe Treasury金库系统，打造兼容法定货币与稳定币的金融操作环境。该系统支持企业跨周末与节假日无缝兑换并持有USD、GBP、EUR及稳定币，提供FDIC保险保障，并依托Stripe商户网络间每日480万次交易实现账户间零费率即时转账，全球出海付款覆盖160个国家。通过模型上下文协议（MCP），AI代理可直接调用金库API自主解析发票PDF并在多重安全校验后完成大额支付。 在嵌入式金融方面，Stripe Capital推出了灵活的循环信用额度，其内部研究显示获得资金支持的企业两年内增速较未获支持者高出27个百分点。同时，累计发卡超3.5亿张的Stripe Issuing正式支持向自主AI代理安全发放虚拟卡，并联合Visa在100个国家支持稳定币消费卡发行，允许用户与AI实体直接在主流刷卡网络中消耗链上资产。 Stripe在此次发布中试图构建适配自主AI代理与分布式算力计费的端到端经济轨道。不过，演讲所展示的端到端代理交易多基于受控演示与实验环境，在真实生产场景中，全自动代理支出的授权边界、跨国多币种税法合规以及区块链微支付的监管接受度仍面临复杂的制度与运营考验。

## Background 本视频由 Stripe 官方频道于 2026 年 6 月 7 日发布在 YouTube 平台，记录了 Stripe Sessions 2026 的主论坛主题演讲。视频简介指出，本场演讲集中展示了 Stripe 专为促进企业收入增长而设计的全套产品与底层技术，并同步附带了开发者专场、全球经济观察以及与 Sam Altman、Nat Friedman 等行业嘉宾对谈的精选链接。 ## TL;DR 在以“奇点第 119 天”为题的年度大会上，Stripe 全面宣布从传统支付基础设施跃迁为“面向 AI 时代的全球经济操作系统”。大会重磅推出了打通智能体端到端部署的 Stripe Projects、机器支付协议（MPP）、Link Agent 钱包，以及面向开发者与平台的自主风控助手 Stripe Console。针对 AI 爆发带来的海量高频算力计费需求，Stripe 整合了 Metronome 实时计费引擎与专为微支付构建的区块链 Tempo，实现按 Token 实时流式结算。同时，Stripe 发布了 Postgres 架构的 Stripe Database，并通过适配 MCP 协议的全新 Treasury、数字资产账户（Digital Asset Accounts）与平台增长工作室，联合 Google、Meta 和 Shopify 等巨头共同奠定了自主智能体商业与多币种跨境流转的新基准。 ## Atmosphere & style 整场演讲洋溢着浓厚的极客科技乐观主义与极快节奏的信息轰炸，主讲人以“奇点时代”的幽默调侃开场，迅速将气氛拉入工程技术重塑现实商业的激荡脉络中。多轮真实的终端真机演示、AI Agent 自动调配云资源以及跨模型自主议价结账的现场实测，让台上台下充满了见证历史拐点的兴奋与屏息期待。贯穿全场的炉边对谈既有对金融本质的深度哲思，又有工程师文化独有的松弛感与现场互动，令观众真切体会到未来智能体经济已然在指尖呼之欲出的震撼临场感。 ## Bulletin - [00:00](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=0s) Stripe 年处理交易量（TPV）达到约 1.9 万亿美元（同比增长 34%），相当于全球 GDP 的约 1.6%，平台关键 API 全年可用性维持在 99.9996%。 - [11:36](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=696s) 联合 Tempo 与 Paradigm 推出面向高频机器微支付的专用区块链 Tempo，并上线 Agentic Commerce Suite 与机器支付协议（MPP），正式支持两个独立 AI Agent 间自主完成软件采购与清算。 - [07:55](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=475s) 正式发布 Stripe Projects 与 Stripe Console，打通智能体全自动编排部署云端服务及全自动管理生产级支付配置与 A/B 测试的闭环。 - [55:49](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3349s) Stripe Treasury 全面升级为基于法币与稳定币双轨的财务操作系统，上线企业间 Treasury 账户“零手续费即时转账”，并支持基于 MCP 协议通过 AI Agent 托管企业账单支付与资金划拨。 - [01:23:15](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=4995s) 重磅发布基于 70 万亿条商业数据的 Stripe Signals，并与 Google、Shopify 联合推动通用商业协议（UCP），正式向全球开发者公开至 2027 年 Q1 的官方产品路线图。 ## Key takeaways - [00:00](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=0s) Stripe 平台处理的资金量已占全球 GDP 的 1.6%，2026 年初起每月新增企业数呈现抛物线式爆发，标志着全球经济正经历由 AI 驱动的全面技术重构（re-platforming）。 - [06:00](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=360s) Stripe CLI 调用量在 2026 年初出现指数级激增，幕后主要驱动力已由人类开发者转变为自主智能体（AI Agents），智能体正加速成为自主经济参与者。 - [07:55](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=475s) Stripe Projects 正式结束内测全面可用（GA），使 AI 智能体不仅能“氛围编程（vibe coding）”，更能在命令行内自主自动化编排云资源并实现“氛围部署（vibe deploying）”。 - [11:36](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=696s) Stripe 联合 Tempo 推出机器支付协议（MPP）并发布 Link Agent 钱包，支持采购 Agent 识别 HTTP 付费规范，并在无需向模型暴露真实卡号的前提下完成人类指纹授权闭环。 - [16:03](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=963s) Stripe 与 Google 达成深度战略合作，使全球用户能直接在 Google 搜索 AI 模式及 Gemini 应用内原生下单购买商家商品，全面重塑 Agentic Commerce。 - [20:13](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=1213s) 推出结账控制中心 Checkout Studio 与无代码嵌入式表单构建器，数据显示使用全托管 Checkout 的商家业务营收增速是普通商家的 2 倍。 - [22:58](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=1378s) Stripe Treasury Payments（MoR）方案面向全球 195 个市场全面开放；自适应定价（Adaptive Pricing）助推跨境收入平均提升 17.8%，Gamma 接入印度 UPI 当月营收暴增 22%。 - [26:23](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=1583s) 发布专属自主型智能体 Stripe Console，支持多任务并行分析经营指标，并能在人类确认机制保护下自主配置并上线生产环境 A/B 测试。 - [33:23](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=2003s) 支付范式正从用户亲自参与的“即时动作”演变为用户向 AI 智能体预设边界规则的“自动化策略”（Payments pivot from moment to policy）。 - [35:08](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=2108s) AI 黑产欺诈重心转向“盗窃 Token”以空耗 GPU 算力；Radar 全面升级，专门针对恶意多账号滥用、免费试用刷量及后付费透支逃单三大新型 AI 欺诈构筑前置防护。 - [45:05](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=2705s) 收购实时计费引擎 Metronome，并联合 Paradigm 打造专为高频机器交易设计的底层区块链 Tempo，首创实现“按 Token 实时消耗边烧边付（Tokens Paid as Burned）”的稳定币流式微支付架构。 - [51:13](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3073s) 收入套件全面提升可编程性，推出可与工作流和自定义 App 协同的“自定义对象（Custom Objects）”，并发布基于 Postgres 构建的实时数据引擎 Stripe Database。 - [55:49](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3349s) 全新 Stripe Treasury 支持法币与稳定币混合持仓、7×24 小时即时汇兑、Link 稳定币钱包打款，并在 Stripe 生态内实现每日 480 万次企业间“零手续费瞬时到账”。 - [01:00:41](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3641s) Treasury 底层 API 全面适配模型上下文协议（MCP），企业可授权 AI Agent 直接解析 PDF 发票、核对历史账单并自动触发企业间转账。 - [01:05:51](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3951s) 面向开发者统一开放 Digital Asset Accounts API，整合跨链支付、法币出入金、链上外汇及灵活托管，Felix 基于此架构支撑了美墨汇款走廊超 5% 的交易量。 - [01:09:37](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=4177s) 面向平台客户推出 Platform Growth Studio，基于全网同行数据智能诊断定价漏损并一键发起促销；同时在 Issuing 业务中新增面向 AI Agent 的专属发卡支持。 - [01:18:09](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=4689s) Shopify 接入 Agnostic Commerce Suite 并宣布作为验证者节点加入 Tempo 区块链，同时与 Google、Stripe、Meta、Amazon 等共同构建通用商业协议（UCP）。 - [01:23:15](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=4995s) 基于 70 万亿条商业数据沉淀推出外部可调用的 Stripe Signals 评分系统，并首次对外公开至 2027 年第一季度的官方产品路线图。 ## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=0s) 奇点第 119 天：Stripe 规模跃迁与全球经济重构 Stripe 联合创始人兼首席执行官 Patrick Collison 走上舞台，向现场观众问好，并幽默地将大会当天（4 月 29 日）定义为“奇点第 119 天”（day 119 of the singularity）。他指出，本届 Sessions 是 Stripe 历史上规模最大的一次盛会，台下汇聚了包括 AI 领军企业、平台型巨头、大型跨国企业、初创团队、电商企业以及独立开发者在内的整个商业光谱。正是这些各行各业的建设者，正在将科学与技术的崭新奇迹真正落地到现实世界中。 随后，Patrick 分享了一组展现 Stripe 当前规模与韧性的重磅数据。Stripe 目前在全球服务超过 500 万家企业，其客户群体覆盖了 Forbes AI 50 榜单中 86% 的公司、道琼斯工业指数中 90% 的企业，以及纳斯达克 100 指数中 80% 的巨头。更惊人的是，Stripe 平台上所有企业创造的收入总量，在过去一年相当于全球 GDP 的约 1.6%，总支付交易量（TPV）达到约 1.9 万亿美元，同比增长高达 34%。在全球经济平均增长率仅约 2% 的背景下，依托 Stripe 构建业务的企业展现出了远超大盘的扩张速度。在保持高速增长的同时，底层基础设施依然稳健：过去一年中，Stripe 的核心 API 保持了 99.9996% 的可用性，而工程团队平均每天合并超过 1,000 个生产环境 Pull Request（PR）。 谈及“奇点第 119 天”这个略带戏谑的称谓，Patrick 首次向外界公开了一张关键图表——Stripe 平台每月新增企业注册数。数据表明，2019 年底至疫情爆发初期该指标曾出现过一次剧烈飙升，且在此后数年始终维持在这一高位平台；然而自 2026 年初开始，新增企业的曲线突然呈现出接近垂直的抛物线式爆发增长。这不仅代表着初创公司的激增，更揭示了整个经济体系正在 AI 的驱动下全面重构（re-platforming）。面对这一转折，Patrick 指出所有企业都面临着两项必然任务：第一，彻底重新发明自己的产品；第二，彻底重构业务的运转模式。Patrick 回顾自己早年作为 Lisp 程序员的经历，Lisp 的核心思想是“代码即数据”（code is data），而 Stripe 自 2011 年成立以来的核心信念则是“资金即数据”（money is data）。这种可编程的金融基础设施，如今已成为 AI 时代不可或缺的底层支柱。 - 全球服务企业总数突破 500 万家，客户包括 86% 的 Forbes AI 50 公司、90% 的道琼斯成分股及 80% 的纳斯达克 100 企业。 - Stripe 平台年处理交易量（TPV）达 1.9 万亿美元（同比增长 34%），相当于全球经济总量的 1.6% 左右。 - 关键 API 全年可用性达到 99.9996%，生产环境日均合并 PR 数量超过 1,000 个。 - 2026 年初起，新企业入驻量进入抛物线式暴增阶段，驱动全行业在产品与运营双维度重构。 --- ### [06:00](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=360s) CLI 调用的爆发式增长与 Agentic 时代的来临 Stripe 技术与商业总裁 Will Gaybrick 接替登台。他表示，2026 年初发生的变化在微观层面同样清晰可见。为了证明这一点，他展示了 Stripe 内部一个原本极为小众的产品数据——Stripe 命令行界面（CLI）。这款工具在 7 年前低调推出，长年以来几乎只有少数深度技术极客在使用，图表常年贴近底部，仅在去年略有微幅抬头。然而，从今年初开始，Stripe CLI 的使用量突然呈现出近乎垂直的对数级爆炸。 促成这一暴增的幕后推动者并不是人类开发者，而是自主智能体（AI Agents）。Will 感慨道，业界多年来一直在讨论“智能体时代”（agentic era），但在短短数周之内，所有人就已真切地生活在其中。如今在 Stripe 内部，工程师的日常工作已经全面转向构建智能体、设计评测基准（evals）、编写技能插件（skills），以及将既有产品重构为能够原生供智能体调用的系统。 Will 强调，人们往往容易低估智能体成为自主经济参与者所带来的深远影响。未来绝大多数互联网交易都将由自主智能体完成；它们不仅工作速度比人类更快，花钱与调用资源的速度也会快得多。这将催生出巨大的全新经济机遇，而当下的核心命题就是所有企业该如何提前布局、做好准备。 - Stripe CLI 在平稳运行 7 年后，于 2026 年初因 AI Agent 的大规模调用而产生爆发式增长。 - 开发范式彻底转变，核心工作围绕智能体开发、Eval 设计、Skill 编写及接口面向智能体的再改造。 - 智能体正加速演变为具备自主财务决策与支付能力的经济主体（autonomous economic actors）。 --- ### [07:55](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=475s) Stripe Projects 登场：从 Vibe Coding 走向 Vibe Deploying 进入产品实操环节，Will 直指当前 AI 辅助开发的痛点：智能体虽然能极其迅速地编写出海量代码，但往往无法独立将应用打包并发布到公网供真实用户访问。在过去，所谓“氛围编程”（vibe coding）的另一半——也就是“氛围部署”（vibe deploying）——始终难以闭环。为此，Stripe 正式向全量用户开放此前处于内测阶段的新产品：Stripe Projects。 Will 引用了 Andrej Karpathy 去年的一条推文指出，在智能体编程时代，构建应用的过程不应该像组装宜家家具那样痛苦——开发者不得不在十几个不同的云服务之间来回切换，反复注册账号、在不同的管理面板跳转并复制粘贴 API 密钥，这一系列繁琐流程过去必须由人类手动完成。Stripe Projects 正是为此而生：它允许开发者或智能体直接在命令行中，以编程方式自动化配置云服务资源、管理配额消耗，并处理跨服务的统一计费，使得智能体能够实现端到端的自主上线。 在现场演示中，Will 依托本地运行的一个 API 审查应用（该应用能自动评估任意公开 API 的可读性与开发人体工程学，每次审查收取 2 美元），在终端向 Claude Code 下达了一行指令，要求其通过 Agent 沙箱与托管服务完成一键部署。在后台执行期间，Will 宣布移除 Stripe Projects 的等待名单，向全球开发者全面开放（GA）。随后，Claude Code 成功自动编排了三家服务商：选用 Vercel 进行前端托管、Supabase 作为后端数据库、RunLoop 作为 Agent 执行沙箱，并全自动注入了凭证与密钥。Will 将终端生成的公网链接复制回浏览器，应用瞬间成功加载上线，验证了智能体在无人干预下完成全栈云资源编排与计费集成的完整闭环。 ```mermaid flowchart TD accTitle: Stripe Projects 智能体自动化部署架构 accDescr: 展示了 Claude Code 如何通过 Stripe Projects 自动编排 Vercel、Supabase 与 RunLoop 并完成公网上线。 subgraph AgentInput["输入与决策层"] A["开发者提示词<br/>(Claude Code)"] --> B["Stripe Projects CLI"] end subgraph CloudOrchestration["云基础设施自动化编排"] B --> C["Vercel<br/>(前端与网页托管)"] B --> D["Supabase<br/>(数据库与存储)"] B --> E["RunLoop<br/>(Agent 运行沙箱)"] end subgraph LiveApp["生成上线成果"] C --> F["公网实时应用<br/>(API Reviewer @ $2/次)"] D --> F E --> F end ``` - Stripe Projects 正式结束内测全面开放（GA），彻底解决智能体在代码编写后无法自主完成“Vibe Deploying”的瓶颈。 - 消除在多个服务商仪表盘间手动创建账户、复制 Secret 和配置配额的阻碍，将资源供应、用量管理与账单整合至统一命令行中。 - 演示案例中，Claude Code 自主协同了 Vercel（托管）、Supabase（数据库）与 RunLoop（Agent 容器沙箱），一次性完成付费工具的自动化上线部署。 ### [11:36](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=696s) 双 Agent 自主交易实测：MPP 协议与 Link Agent 钱包 互联网原有的商业基础设施——不论是价格页面还是结账表单——从一开始就是专门为人类的视觉和交互习惯设计的。当 AI 智能体（Agent）步入商业舞台时，它们必须具备完全自主采购的能力，而这在传统网页交互中极为困难。为了攻克这一非平庸的工程难题，Stripe 在几个月前联合 Tempo 团队共同推出了开源标准“机器支付协议”（Machine Payments Protocol，简称 MPP）。MPP 允许任何网络服务直接通过 HTTP 协议向 Agent 声明该接口需要付费以及具体的支付规范。然而，即便服务端做好了接收机器付款的准备，另一个更核心的信任链条依然悬而未决：如何赋予 Agent 支付凭证？又如何确保它们不会滥用额度？ 针对这一痛点，Stripe 正式推出了专为智能体打造的“Link Agent 钱包”（Link's new wallet for agents）。该产品允许用户安全地授权 Agent 代表自己进行消费：底层的真实银行卡信息永远不会暴露给模型，同时用户在关键节点保持完全的审核权与控制权。 为了在现场闭环验证这一链路，演讲者 Will 启动了一场高难度的实时演示：让一个采购 Agent 自主向另一个由 Agent 刚刚编写并上线的服务购买 API。在此前的环节中，第一个“构建者”Agent 使用的是 Claude，而此时的采购 Agent 则切换为暗黑模式界面的 OpenAI Codex，以在视觉上形成明确区分。Will 既没有向 Codex 灌输任何关于 MPP 的背景，也没有预先配置 Link，只是发送了一条简单的指令：“帮我获取 Privy 的 API 审核（Review）”。 Codex 发出请求后，立刻通过 HTTP 响应自动识别出该目标应用支持 MPP 协议且接受信用卡结算；紧接着，它自动推断出可以使用 Link 钱包完成支付，并精准提取出接口调用费用为 200 美分（即 2 美元）。随后，Codex 主动向 Will 的手机端 Link 应用推送了一条扣款审批请求。Will 打开手机上的 Link 界面，屏幕清晰展示了 2 美元的 API 审核账单及绑定的 Visa 卡信息，他轻轻按下大拇指完成生物指纹验证。数秒之内，审批信号回传，Codex 界面跳出交易成功提示。 ```mermaid flowchart LR accTitle: Agent 自主交易与 Link 授权流程 accDescr: 展示采购 Agent 通过 MPP 识别费用并经由用户 Link 钱包生物指纹审批完成自主支付的闭环。 subgraph Buyer["采购端 (Codex Agent)"] A["发起 API 请求"] --> B["通过 MPP 协议<br/>识别 2 美元费用"] B --> C["调用 Link Agent 钱包"] end subgraph Auth["人类监督 (Human-in-the-loop)"] C --> D["推送至 Link App"] D --> E["用户指纹审批确认"] end subgraph Seller["服务商 (Claude 构建应用)"] E --> F["MPP 验证支付凭据"] F --> G["返回 API 审查结果"] end ``` - **机器支付协议（MPP）**：Stripe 与 Tempo 联合打造的开放标准，使服务能够通过原生 HTTP 状态与标头直接向 AI 智能体索取支付并约定支付方式。 - **Link Agent 钱包**：兼顾自主权与风控体系，对模型完全屏蔽底层真实卡号，支持人类通过生物识别进行精准交易确认。 - **现场实测数据**：采购 Agent 准确解析 200 美分（2.00 美元）接口费用并完成自动清算。 --- ### [14:45](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=885s) 智能体经济拐点与大会三大核心主题 现场响起热烈的掌声，Will 借用登月名言感慨道：“这对眼前的两个 Agent 来说只是一小步，但却是智能体成为真正经济参与者的一大步（one small step for these two agents, but one giant leap towards making agents full economic actors）。”在短短数月前，让两个 AI 模型自发进行软件开发、服务发现、议价结算的全流程还如同天方夜谭，但随着底层代码生成能力、MPP 协议以及 Link Agent 钱包的有机串联，人类已经站在了“智能体自主构建、购买与销售软件”的全新经济前沿。 面对这一剧烈变革，企业正处于一个奇妙而充满张力的历史时刻：一方面，每一家公司的管理层都被要求从底层逻辑彻底重塑商业模式以迎接 AI 时代；另一方面，团队又必须保质保量地维持当下日复一日的日常运营，服务好现有的海量真实客户。Stripe 的研发目标正是帮助开发者在平稳维系现实业务的同时，无痛接入面向未来的技术栈。Will 随即揭示了贯穿整场大会的三大核心支柱： - **更高的可编程性（Programmability）**：让 Stripe 的每一层能力都能更灵活地被代码与 AI 调用。 - **Stripe 全球网络的护航与加速（Protection & Propulsion）**：依托海量交易沉淀的智能风控与全球清结算通道驱动业务增长。 - **面向 AI 的经济基础设施（Economic Infrastructure for AI）**：为即将爆发的 Agentic 时代提供原生的收单、计费与资金流转支持。 --- ### [16:03](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=963s) Agentic Commerce 的演进光谱与 Google 战略合作 负责支付业务的 Kevin 接过麦克风，指出支付始终是客户与 Stripe 共同的业务心脏，而这个心脏正迎来一种全新的买家形态——智能体。Stripe 将这一新兴的“智能体电商”（Agentic Commerce）演化路径划分为一条连续的光谱： ```mermaid flowchart LR accTitle: Agentic Commerce 演进光谱 accDescr: 展示从人类主导发现购买到最终 Agent 全权自主交易的三阶段演进过程。 S1["前端探索阶段<br/>Agent 协助搜索/发现<br/>人类挑选并结算"] --> S2["混合协作阶段<br/>人类决定购买标的<br/>Agent 自动代理结账"] --> S3["完全自主阶段<br/>Agent 自主发现比价<br/>Agent 直接结算交付"] ``` 由于谁也无法精确预测全球消费者的心智与行为习惯会在哪一个时间节点发生突变，Stripe 的产品策略是打造包容所有演进阶段的中立底层基座，让商家无需在不确定的趋势上下注即可全面受益。目前，零售、流媒体、SaaS 及旅游等众多垂直赛道对接入 AI 渠道的需求正在井喷。Kevin 盘点了此前与其他几家科技巨头的落地进展：联合微软与 OpenAI，分别在 Copilot 和 ChatGPT 内部实现了商品的原生检索与展示；携手 Meta，实现了在社交广告流内部的原生直接结账。 紧接着，Kevin 宣布了一项重磅全新合作：**Stripe 与 Google 正式达成深度战略合作**。全球数十亿用户很快将能够直接在 Google 搜索的“AI 模式”（AI mode）以及 Gemini 原生应用程序中，无需跳转即可直接购买商家的商品。首批上线支持该能力的品牌包括知名服饰家居品牌 Quince、体育衍生品巨头 Fanatics 以及全球零售商 JD Sports。至此，OpenAI、Microsoft、Meta 与 Google 这四大 AI 领域的全球巨头已全部接入 Stripe 的生态版图，共同重塑未来的交互式购物体验。 --- ### [18:01](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=1081s) Agentic Commerce Suite 与全球钱包 Link 的全面升级 当光谱推进到极致——即 Agent 直接作为买家向企业采购时，MPP 协议便成为了不可替代的数字握手工具。Kevin 调侃道，MPP 已经斩获了一个支付协议所能获得的最高荣誉：“人们居然在互联网上自发地发帖吹捧它（People are voluntarily posting about it on the internet）。”为了将这两种截然不同的销售模式（在 AI 前端对话界面售卖与直接通过 MPP 向 Agent 售卖）融为一体，Stripe 推出了全栈式的 **Agentic Commerce Suite**。商家只需完成一次集成，其目录与服务就能被外部各类 Agent 瞬间“读懂”，且商家依然保持独立结算主体（Merchant of Record, MoR）的地位，底层的全流程结账、欺诈检测与跨币种清算则全部由 Stripe 在后台静默托管。 在生态合作方面，Shopify 正式成为接入 Agentic Commerce Suite 的零售商首选目录服务提供商（Preferred Catalog Provider）。同时，Wix、commercetools 等建站与电商平台，以及 Best Buy（百思买）和 Kate Spade 等大型企业级品牌也已全面进驻。平台型客户只需开启该功能，其平台上的海量中小商户便能瞬间获得 AI 渠道的直接获客能力，而无需独立编写任何适配 Agent 的代码。 一切智能体交易能够规模化落地的核心瓶颈，依旧在于 C 端消费者能否放心向 Agent 赋权资金。在这套闭环中，Link 正是支撑全球智能体交易的“Agent 钱包”基石。 - **超 2.5 亿活跃用户网络**：Link 已经积累了超过 2.5 亿（a quarter billion）真实开通用户，为复购消费者平均带来 5% 的结账转化率净提升。 - **深度数据看板集成**：商家即日起可以直接在 Stripe 管理后台（Dashboard）中细颗粒度追踪 Link 渠道的转化效果与收入走势。 - **全球主流本地支付方式整合**： - 印度 **UPI**：覆盖印度超 5 亿活跃用户。 - 巴西 **Pix**：服务巴西逾 1.6 亿人口。 - **稳定币（Stablecoins）**：整合全球加密稳定币收付能力，帮助商家以极低手续费捕获全球跨国流动性。 --- ### [20:13](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=1213s) Checkout Studio 亮相与可嵌入式表单可视化构建演示 回归到核心的人类转化场景，Stripe 优化的结算套件（Optimized Checkout Suite）长期通过动态币种转换、本地化语言渲染及自适应组件帮助跨境商家拔高收益。官方统计数据显示，采用 Stripe Checkout 全托管收银台的企业，其平均业务营收增速是普通商家的整整 **2 倍（two times faster）**。在日常运营中，高阶团队最常践行的是一套“配置收银台 -> 监测数据 -> 迭代优化”的转化率飞轮（Flywheel）。 为了让飞轮运转得更加迅捷，Kevin 正式发布了全新的控制中心 **Checkout Studio**。商家可以在此统一并排比对名下分散的所有结账流程。Kevin 打趣说，这就像是结账流程的“嫌疑犯面部大头照墙（mugshots）”，运营人员一眼就能揪出“到底是谁昨天把转化率给拉低了”。通过 Studio，企业可以编写深度的自适应路由规则，并针对不同支付方式启动严谨的 A/B 测试。 此外，针对以往许多使用 Payment Element 的开发者希望将结账入口嵌入到移动端应用、AI 客服聊天浮窗等任意形态的需求，Stripe 推出了一款更具定制弹性、自适应各种界面的嵌入式表单（Embedded Form），并配备了开箱即用的可视化设计器（Visual Builder）。 演示工程师 Felix 随即在讲台前进行了无代码实时构建演示： 1. **新建表单**：点击“Create checkout”，选定新增的“Embedded form”分支，命名为“support chat UI”（客服聊天收银台）。 2. **可视化画布**：中间为实时响应的收银台画布，左侧整合了详尽的控制面板与预制工具箱。Felix 勾选“收集客户姓名”字段，右侧预览窗立刻实现毫秒级局部重绘。 3. **界面风格定制**：为了让视觉匹配低光照设计，Felix 轻轻一按切换为暗黑模式（Dark Mode），界面的底色、边框与文本对比度瞬间完成适配。 4. **拖拽插入自定义扩展**：Felix 在面板中选取一个 Checkbox（复选框）自定义组件，文本设定为“加入我的电子周刊（Join my newsletter）”，直接拖拽丢入画布的放置区（Drop Zone）内。 5. **一键生成代码**：点击保存后，轻按“Implement”，系统当场生成可直接复制至业务前端代码库的轻量级调用 Snippet。整个开发流程从构思到可上线形态耗时不足两分钟。 ### [22:58](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=1378s) 全球扩张难题与本地化定价利器：从 MoR 到 Adaptive Pricing Felix 在展示完 Checkout 自动化流程后幽默地调侃道：“除了我已经快记不得怎么写代码了之外，这个界面堪称完美”，随后他演示了那个方便实用的 “Copy for LLM” 按钮，只需轻轻一点，生成的代码片段就能直接交给后台的智能体处理，真正实现了在几下点击中完成从可视化设计到提示词输出的跃迁。随后接过话题的主讲人指出，通过全新的 Checkout Studio，企业可以在短短几分钟内配置并优化结账体验。但真正的现实挑战往往发生在跨国扩张阶段——全球拓展在商业计划书上看起来光鲜亮丽，可一旦落地，迎面而来的便是令人头疼的复杂税法、合规审查、本地客服以及一张望不到头的待办事项清单。为此，Stripe 正式推出了 Merchant of Record（MoR）解决方案 Stripe Treasury Payments（托管支付服务），帮助包括 Unity、RevenueCat 以及 Lovable 在内的企业轻松扩展至全球 195 个市场，大幅消除背后的运营摩擦，并于即日起正式面向所有用户全面可用（GA）。 除了合规与财税，跨币种和本地化定价同样是跨境企业的核心痛点，商户不仅要直面汇率波动、外汇敞口，还得解决不同地区本地支付方式的兼容问题。Stripe 推出的自适应定价功能（Adaptive Pricing）交出了一份惊艳的数据答卷：在实际运行中，商户跨境收入平均提升了 17.8%，授权率平均提升了 5%。这项能力现在更进一步延伸至订阅类计费场景，无论是初创团队还是大型跨国企业，在接入订阅自适应定价后均实现了 4.7% 的业绩增长。与此同时，Stripe 在底层支付网络的跨境互联与循环扣款标准化上也下足了功夫，典型案例就是作为增长最迅猛的 AI 独角兽之一的 Gamma，在通过 Stripe 接入印度本土主流支付网络 UPI 后，当月来自印度的业务收入便迅猛激增了 22%。 - **Stripe Treasury Payments (MoR)**：全面可用（GA），为企业覆盖 195 个国家/地区的财税与合规运营。 - **Adaptive Pricing 核心成效**：跨境交易收入平均提升 17.8%，支付授权率平均提升 5%，订阅业务平均实现 4.7% 的增收。 - **本地支付实证**：AI 演示文稿工具 Gamma 接入印度 UPI 当月，印度区营收即实现 22% 的逆势暴涨。 ### [25:05](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=1505s) 线下硬件新秀 Reader T600 与底层收益优化神器 Auth Boost 全球化不仅存在于数字虚拟世界，更发生在千千万万的线下物理场景中。全球已有超过 90 万家商户在 25 个国家利用 Stripe Terminal 处理线下收银业务，其中不乏像 Hertz 这样的全球租车巨头。今年，Stripe Terminal 宣布将版图进一步扩展至额外的 15 个国际市场。伴随地域扩展的是日益丰富的专用硬件阵列：在搭载蜂窝网络支持、备受市场好评的手持智能终端 S700 之后，Stripe 正式揭晓了全新硬件——Stripe Reader T600。演讲者打趣道，内部为了憋出这个名字“肯定烧掉了大把的大模型 Token”。T600 是一款专门面向消费者的交互式桌面显示终端，能够通过运行定制化应用，为商户承载会员忠诚度积分、即时加购推荐等高端交互式结账体验。 结账环节的体验设计只是整个收益引擎的一面，深埋于交易底层流水线的优化同样生死攸关。去年 Stripe 发布的 Auth Boost 深度介入了从发卡行智能重试到交易动态路由的各个环节，能够自动更新过期卡号、在多卡组织网络间精准路由交易，并利用算法模型精准计算最佳重试时间。运行一年以来的真实数据显示，Auth Boost 帮助企业平均带来了 3.8% 的授权率提升，同时帮助商户削减了高达 3.3% 的支付交易成本。为了消除商户对底层变动的疑虑，Stripe 现在还为 Auth Boost 原生支持了 A/B 测试功能，让企业能够直接在控制台中清晰核验每一笔优化所带来的净收益。 ```mermaid flowchart TD accTitle: Auth Boost 智能优化链路 accDescr: 展示 Auth Boost 从卡号更新到多网络路由及智能重试的降本增效流程 subgraph Optimize["Auth Boost 动态优化引擎"] A["交易发起"] --> B["自动卡号更新 / 智能路由"] B --> C["智能重试算法决策"] end subgraph Outcomes["业务价值量化"] C --> D["授权通过率提升 +3.8%"] C --> E["交易手续费节省达 3.3%"] D & E --> F["控制台原生 A/B 测试量化验证"] end ``` ### [26:23](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=1583s) Agentic 专家登场：Stripe Console 现场实机演示 随后舞台将目光转向了更前沿的智能体研发成果。Stripe 将自身沉淀的全部业务增长经验与风控体系，浓缩凝聚进了一个名为 “Stripe Console” 的专属自主型智能体中。工程师 Felix 登台进行了现场实机演示。不同于传统单问单答的查询工具，Console 并非只是一个“昨天产生了多少笔退款”这样的查询工具——虽然当 Felix 随口抛出这个问题时，Console 极其贴心地秒回了一个准确数字“17”，但它真正强悍之处在于能直接调查复杂数据并主动发起生产级操作。 Felix 进一步对它施压：“我的业务本月至今表现如何？”话音刚落，底层控制台便开启多任务并行处理流程，实时拉取并聚合全量交易日志、收益曲线以及按国家划分的转化率矩阵图表。更有深度的是，Console 不仅在被动罗列数据，而且展现出了自主推理与诊断能力：它精准识别出“巴西地区的结账转化率明显落后于其他市场”，并敏锐地给出了诊断建议——建议商户立刻启用巴西主流的本土即时支付工具 Pix。紧接着，Console 甚至主动提议：“需要我为您针对 Pix 配置一个 A/B 测试吗？”在 Felix 兴奋地输入“Let's do it!”后，Console 自动计算出底层所需的 Tool Calls，出于安全合规考虑弹出最终确认提示，并在获得指令后实时调整生产支付配置，当场在全场观众的注视下将 Brazil Pix 的结账 A/B 测试平滑上线发布。 ```mermaid flowchart LR accTitle: Stripe Console 智能体决策闭环 accDescr: 演示从多任务并行拉取数据、诊断巴西转化率过低、建议 Pix，到自动创建并部署 A/B 测试的完整闭环 subgraph Diagnosis["智能体洞察与分析"] Q["用户提问：月度业务表现"] --> P["并行拉取全量支付数据"] P --> D["诊断发现：巴西地区转化率偏低"] end subgraph Action["自主决策与生产部署"] D --> R["主动建议：启用 Pix 支付"] R --> T["自动生成 Tool Calls 与 A/B 测试"] T --> C["管理员一键授权确认"] C --> L["生产环境 A/B 测试实时上线"] end ``` - **Stripe Console 定位**：通用型 Agent，专门面向商户 Stripe 账户深度调优，核心目标是自主解决“如何让业务更快增长”。 - **人机协同安全（Human-in-the-loop）**：尽管智能体具备自主组装工具调用（Tool Calls）并修改配置的能力，在面对影响真实用户的结账与支付逻辑变更时，依然会严格弹出确认机制，保障生产环境安全。 ### [29:05](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=1745s) 对话 Meta 产品副总裁 Ginger Baker：社交电商的 Agent 时代与全球结算 在演示落幕后，Meta 产品副总裁 Ginger Baker 与 Stripe 的 Will 共同登台，就 Agent 时代商业的未来展开了一场干货满满的对话。Ginger 开场打趣道，回到这里就仿佛置身于一场“Fintech 大家庭的亲友大聚会”。面对 Will 抛出的大众刻板印象——“大家想到 Meta 首先联想到的是全球统治级的社交媒体，很少直接挂钩商业支付，Meta 究竟为何而来？”，Ginger 给出了深刻的底层逻辑剖析：Meta 的底层使命始终是连接人与人，全球每天有高达 35 亿人活跃在 Meta 旗下的各个 App 中，无论人们在 Threads 发布文字、在 Facebook 分享视频，还是在 Instagram 创作内容，本质上都是利用数字载体建立连接。在她看来，“金钱本质上不过是数字内容的另一种形态”，如果 Meta 不能让金钱在自己的平台上像文字或短视频那样无缝流转，就是错失了连接人与人之间最关键的一种纽带。 聊到“社交电商（Social Commerce）”的演进，Ginger 向台下抛出互动：“本周用过 Instagram 的人请举手？不要害羞，不针对任何人。”在笑声中她剖析了当下的购物链路痛点：用户往往在好友的动态里被一条金色蛇形耳环种草，然而想要购买却必须经历一系列繁琐的摩擦——截图、以图搜图反查品牌、搜索官网、注册账号、翻出信用卡输入卡号结账。而在 AI 智能体时代，这种割裂的体验将被彻底重构为“原地消费（In-situ Experiences）”：从视觉灵感激发、点击照片直接获取尺码价格，到一键原生完成购买，整个闭环都在一个界面内无跳转完成。 在谈及全球商业落地面临的复杂跨境挑战时，Ginger 强调，Meta 的产品与穿戴设备覆盖了全球几乎每一个国家，因此必须极度依赖像 Stripe 这样兼具深度与韧性的底层基础设施合作伙伴。她特意以极度多元化的创作者群体为例：从一位仅靠 WhatsApp 经营生计的印度尼西亚手工艺单人创作者，到全美跨多个平台运营数百万美元流水的大牌明星，彼此的结算偏好与资金进出链路天差地别。为了满足这种极致的本地化需求，Meta 正在与 Stripe 深度协作，在特定关键市场正式支持使用稳定币（Stablecoins）为创作者打款结算，以此让金融基础设施真正融入全球不同创作者的真实语境。 ### [33:23](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=2003s) 从“瞬间支付”到“预设策略”：AI 时代支付的范式转变 在与嘉宾的对话尾声，双方探讨了 Web3 与稳定币（Stablecoin）在创作者经济中的落地。对于创作者平台而言，核心诉求在于用户体验与创作者关系，而 Stripe 在底层承担了最为繁琐的基础设施工作——包括合规牌照、法币入金（On-ramps）、资产托管（Custody）以及法币出金（Off-ramps）。当被问及稳定币结算能覆盖多少个市场时，Will 透露，到今年年底这一数字将扩展至至少 160 个国家和地区。 随后，话题转向了快速演进的 AI 对未来商业格局的冲击。面对“今天听起来极其激进、但两年后可能变得理所当然的事情是什么”这一提问，嘉宾给出了一个极具洞见的论断：“支付将从一个‘瞬间’转变为一种‘策略’”（"Payments will pivot from being a moment to being a policy"）。 在传统模式下，支付是用户亲自参与的一个即时动作（点按按钮、刷卡或授权）；但在不久的将来，支付将成为用户向其 AI Agent 预先配置的一套自动化规则与防护边界（Guardrails）。例如，用户会向 Agent 设定策略：“本周购买食品杂货的支出不得超过 50 美元”、“优先使用这张信用卡支付”、“绝不允许单笔超过 500 美元的无感授权”。在这个新范式下，真正执行资金调度的将是完全自主的软件实体。虽然在今天看来，将资金处置权全盘托付给 AI Agent 显得不可思议，但这很快就会成为日常商业的常态。 --- ### [35:08](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=2108s) Stripe Radar 全面升级：阻断针对 AI 的“Token 盗窃”与漏斗欺诈 随后，Emily 登台发表关于风控产品 Stripe Radar 的最新进展。她开门见山地指出，AI 的普及彻底重构了黑产欺诈的边界：不法分子的目标已不仅局限于盗取资金或账户凭证，而是转向了“盗窃 Token”（Token theft）。在 AI 商业化中，每一次 Prompt 提示词交互、每一次 API 调用、消耗的每一个 Token，背后都挂钩着真金白银的 GPU 推理算力。防范 Token 盗取已然成为当前所有接入 AI 的企业最为迫切的生死战。如今，不仅 OpenAI、Anthropic、ElevenLabs、Cursor 等前沿 AI 企业信赖 Radar，传统商业巨头如百事可乐（Pepsi）与赫兹租车（Hertz）也依靠其保障交易安全。 Emily 详细剖析了当前在 Stripe 全球网络中蔓延速度最快、甚至引发开发者在社交媒体上频繁“哀嚎”的三大 AI 欺诈模式： 1. **多账号滥用（Multi-account abuse）**：发生在用户录入支付信息之前。黑产团伙通过脚本成千上万次重复注册，疯狂套取新用户附赠的免费算力额度。在 Stripe 平台的所有 AI 企业中，高达六分之一的新注册账号均涉及多账号恶意滥用。针对该痛点，Radar 推出实时账户评估能力，在黑客烧掉第一个 Token 之前即刻予以拦截。借助全网的设备指纹（Device fingerprints）、IP 地址及邮箱域名等底层信号，某 AI 客户接入后发现，Radar 拦截的攻击者中有 80% 曾轻松绕过了他们原本的风控方案。 2. **免费试用滥用（Free trial abuse）**：仅在过去 6 个月内，恶意试用行为就激增了一倍以上，最近一个月尤为猖獗。对于按算力计费的 AI 业务，每一次无效试用都在直接消耗服务器成本。例如，一家合作的 AI 企业为了获得单个真实付费客户，竟然要倒贴 25 次试用成本，导致其获取单个付费客户的算力支出高达 500 多美元。这种恶性循环让企业陷入两难：要么保留试用但被黑产吃光利润，要么取消试用导致增长断崖。现在，企业只需在 Stripe 控制台一键开启防试用滥用功能，系统即可智能放行正常用户并阻击黑产。 3. **后付费透支逃单（Pay-as-you-go abuse）**：黑产用户在一个月内肆意调用并消耗价值数千美元的高额算力，等到账单结算期到来时直接拒付失联。Emily 将其形象地比喻为“针对 Token 的米其林星级霸王餐”。Radar 现已支持在用量累积的过程中动态预测违约逃单概率，支持企业在账单到期之前，提前要求客户充值预付或直接关停服务。 - **多账号渗透比例**：Stripe 平台上 AI 公司平均每 6 个新注册用户中就有 1 个属于恶意刷号行为。 - **获客成本失控案例**：某 AI 企业为促成 1 次付费转化需背负 25 次试用，单客获取成本（CAC）中的 Token 损耗突破 500 美元。 - **试用防护战果**：仅在过去一个月内，Radar 就在 8 家高速增长的 AI 头部企业中拦截了超过 330 万次高风险试用请求。 --- ### [38:58](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=2338s) 全流程欺诈防御：全渠道覆盖、Smart Disputes 与定制化风控模型 除了针对 AI 业务漏斗前期的算力防盗，传统交易环节的欺诈依然普遍存在。Emily 宣布，Stripe Radar 正在将防护纵深延伸至全部交易方式与交易生命周期： - **全支付方式统一风控**：去年 Radar 将防护能力拓展至卡支付之外的 ACH 与 SEPA，而今天 Radar 的高风险交易拦截已覆盖 Stripe 支持的“所有支付方式”。同时，这些风险评分还可通过 API 开放给企业在 Stripe 之外的自建或第三方交易渠道，为跨渠道业务构建统一的风控中枢。 - **Smart Disputes 争议抗辩升级**：针对恶意的拒付争议，Smart Disputes 能自动调取交易日志并生成证据链提交给银行，平均能帮商户多追回 18% 的受损营收。更新后的系统还能在后台主动提示商户“上传哪些特定补充证据能进一步提高胜诉率”；商户补齐材料后证据包将重新智能封装，使得最终胜诉抗辩率直接提升至原先的 3 倍。 - **自定义风控模型（Custom Fraud Models）**：为了满足对风控有极其严苛定制化需求的大型企业，Radar 正式开放个性化模型训练支持。商户可将自身独有的业务信号（如特定商品目录特征、会员积分与忠诚度数据）与 Stripe 的全球交易网络数据相融合，训练专属风控模型。早期采用该架构的客户数据显示，在完全不增加误伤（False positives）的前提下，欺诈识别率提升了至少 15%。 ```mermaid flowchart TD accTitle: Stripe Radar 欺诈防御与生命周期覆盖 accDescr: 展示 Radar 从漏斗前期的 Token 保护、全渠道交易拦截到后期争议处理的风控链路 subgraph PreFunnel["漏斗前置防护 (AI 原生)"] A["恶意注册"] --> B["设备与全网 IP 指纹甄别"] C["免费试用欺诈"] --> D["一键试用防护 (阻断算力空耗)"] E["后付费累积透支"] --> F["实时逃单预测 (动态要求预充值)"] end subgraph TxLayer["交易层防护 (全渠道)"] G["Stripe 内置交易 (全支付方式)"] --> H["Radar 统一风控引擎"] I["非 Stripe 渠道交易 (通过 API 接入)"] --> H H --> J["自定义业务特征融合 (商户目录/积分)"] end subgraph PostTx["事后抗辩 (争议管理)"] K["发生恶意拒付"] --> L["Smart Disputes 自动举证 (+18% 挽回)"] L --> M["针对性关键证据补全提示 (胜诉率提升至 3 倍)"] end ``` --- ### [41:44](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=2504s) Stripe Revenue Suite 与 Metronome：面向 AI Agent 的实时计费引擎 Emily 演讲结束后，Tanya 接过话筒介绍 Stripe 收入套件（Revenue Suite）的重磅革新。她指出，过去 20 年间软件商业的收入模型基本一成不变，但近两年来发生了翻天覆地的颠覆：市场全速转向按用量计费（Usage-based）、按结果交付计费（Outcome-based）、混合计费，以及按 Token 计费。目前包括 OpenAI、Lovable、Notion、Zoom 在内的企业均运行在由计费、税务（Tax）、发票开具（Invoicing）以及数据组件构成的 Stripe Revenue Suite 之上。 本次套件升级聚焦三大维度：支持全新 AI 原生商业模式、赋予 Stripe 更强的可编程性，以及数据的毫秒级实时可访问性。 针对 AI 时代最严峻的工程痛点——成群结队的 AI Agent 疯狂消耗算力（Token maxing）对底层分布式系统造成的冲击，计费系统必须在毫秒级时间内做出响应。为此，Stripe 正式收购了专为高速 Agent 场景设计的计费引擎 **Metronome**，并深度整合三大实时能力： - **实时计量（Real-time metering）**：精确统计客户调用的每一个 Token 消耗。 - **实时计价（Real-time rating）**：在 Token 被消耗的当口，毫秒级应用对应的价格逻辑。 - **实时熔断与告警（Real-time alerting）**：在客户或其 Agent 跑出预算上限的瞬间立即熔断截流，防止意外超支。 随后 Will 重新登台，针对前文演示的“API Reviewer 审核工具”进行现场改造。此前 Will 的工具按每次审查 2 美元的固定一口价收费，但在实际运行中，遇到复杂的代码审查，单次消耗的 GPU 算力成本往往远超 2 美元，导致业务持续失血。Will 将定价机制全面调整为“每 100 万 Token 计费 3 美元”的预付费充值包模式。现场演示中，Tanya 扮演客户角色，发起对一揽子 Privy API 的审查任务，Will 为后台每个自主工作的 Agent 分配了 3 美元的 Token 额度上限，并通过大屏幕仪表盘对 Agent 的实时运转与算力扣减展开实况追踪。 ### [44:21](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=2661s) 实时代币计费与防欺诈的两难困境 在 Moscone 现场的演示大屏幕上，多个 AI Agent 正全速运转，分别审核各自对应的 API。随着任务完成，部分 Agent 的状态灯转为绿色，但也有几个停滞未动。这正是 Metronome 实时计费（Real-time Rating）发挥的作用——它能精确检测到哪个 Agent 的 Token 配额已经耗尽并立即切断调用，从而防止服务方在未收到款项的情况下盲目为客户垫付计算成本。 然而，这种生硬的切断机制随即暴露了一个现实商业困境。现场演讲者 Tanya 调侃道，如果自己是一个信誉良好的优质客户，因 Token 耗尽而导致审核任务半途而废，体验会极其糟糕；但如果平台直接给她透支额度，鉴于此前演讲中 Emily 提到的“代币窃取（Token Theft）”欺诈高发问题，平台又面临坏账与算力被白白消耗的巨大风险。 ```mermaid flowchart LR accTitle: 流式代币消耗与实时微支付结算架构 accDescr: 展示 AI Agent 消耗 Token 时如何通过 Metronome 实时计量并在 Tempo 区块链上流式结算微支付。 subgraph Compute["AI 运行时"] A["AI Agent 集群"] -->|"毫秒级消耗 Token"| B["Metronome 实时计量"] end subgraph Ledger["Tempo 支付公链"] B -->|"触发流式扣款"| C["Tempo 支付流通道"] C -->|"毫秒级微支付结算"| D["稳定币 (每笔 3/1000 美分)"] end subgraph Merchant["商家收益"] D -->|"不可逆实时到账"| E["零坏账 / 无额度阻塞"] end ``` - 核心矛盾：既不能因额度耗尽粗暴中断正常业务，又不能承担无抵押信用垫付带来的欺诈坏账。 - 传统结算方式的局限：无论是信用卡、ACH，还是印度的 UPI 或巴西的 Pix，现有法币支付系统都无法支持毫秒级、单笔极微小金额的连续扣款。 --- ### [45:05](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=2705s) Tempo 登场：基于稳定币的“边消耗边付费”流式微支付 为了从根本上化解这一矛盾，Stripe 提出了一种全新的商业设想：能否在代币被消耗的同一瞬间，以完全对称的速率完成不可逆的实时流式支付？传统法币体系与结算网络在毫秒级微支付面前无能为力，必须借助专为高频机器交易设计的稳定币与底层区块链。为此，Stripe 携手投资与研究机构 Paradigm 联合打造了专为支付而生的全新区块链——**Tempo**。 在舞台演示中，团队现场更新了定价页面，正式上线号称“全球首个代币边烧边付（Tokens Paid as Burned）”的新商业模式。当重新启动 Agent 任务后，屏幕左侧实时跳动显示着 Agent 消耗的 Token 数量，右侧则同步涌入实时结算的稳定币微支付流。Tanya 风趣地打趣道：“如果你成功了，我们是不是可以说我们发布了全美超模（America's Next Top Model）？”演讲者笑应：“是发布了全球顶级全新商业模式（The world's top new business model）！” - **现场区块浏览器实测**：演讲者切入 Tempo 链上浏览器（Block Explorer），向全场展示刚建立的支付流。屏幕放大显示总金额为 3.30 美元，其背后是由成千上万笔单笔仅 3/1000 美分（约 0.00003 美元）的超微型交易汇聚而成。 - **业务收益**：Agent 在运行过程中全程无需限流等待，服务商每一毫秒都在落袋为安，彻底杜绝了垫资风险。 --- ### [47:48](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=2868s) 对话 Georgios Konstantopoulos：Agent 浪潮下的金融基础设施重构 随后，Tempo 与 Paradigm 的首席技术官 Georgios Konstantopoulos 受邀登台。作为前以太坊（Ethereum）核心架构师与技术领袖，Georgios 分享了自己将加密技术专长转向支付领域的初衷。他谈到，自己出生并成长于希腊，切身经历过严重的金融危机，这种创伤让他深刻意识到全球全天候开放金融网络的必要性；但早年加密行业缺乏底层基础设施，尚难承载现实商业需求。 Georgios 指出，如今一场由 AI Agent 引发的“数字海啸”正汹涌而来。这股算力浪潮正在重塑 Git 代码托管与数据库架构，而传统支付系统即将成为下一个被冲垮的瓶颈。Tempo 的诞生正是为了在底层支撑高频、低费用的稳定币与 Agent 金融（Agentic Finance）。 当前已经采用 Tempo 的重点机构与业务场景包括： - **DoorDash**：利用 Tempo 低成本、高时效的优势，在全球范围内向骑手（Dashers）分发配送报酬，显著降低既往高昂的跨国转账基建成本。 - **Klarna**：在 Tempo 上发行了自主稳定币 **Klarna USD**，并将其深度嵌入其金融科技底层设施中。 - **Visa**：与 Stripe、Tempo 深度合作构建机器支付协议（Machine Payments Protocol），为 Agent 间交互提供无摩擦支付通道；同时 Visa 还与 Stripe、渣打银行（Standard Chartered）等机构共同作为验证节点运行 Tempo 网络。 - **Stripe 自身**：已将 Tempo 用于自身的金融账户系统及多项全球资金清结算业务中。 谈及 Tempo 接下来的生产环境路线图，Georgios 重点剧透了两大方向：一是**交易隐私保护**，他强调“在明文状态下使用稳定币是疯狂的”，必须实现全球 24/7 的私密支付结算；二是**极致扩容与性能调优**，进一步提升系统吞吐上限，以确保在 Agent 爆发式增长的浪潮中依然稳定运转。 --- ### [51:13](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3073s) 收入套件全面可编程：推出自定义对象（Custom Objects）与全屏应用 视线回到 Stripe 收入套件（Revenue Suite）的第二项重大升级——赋予系统前所未有的可编程性。Stripe 的底色是开发者优先，而承载企业核心业务运转的收入技术栈更是重中之重。虽然此前已推出可自定义仪表盘界面的 Custom Apps 以及用于自动化流程的编排工具 Workflows，但现实中不同行业的专有业务实体始终难以在 Stripe 内部自然映射。 例如，物流企业核心依赖“运单（Shipments）”，保险企业依靠“保单（Policies）”与“理赔（Claims）”。过去，企业往往被迫维护多套独立系统，用笨拙的 Metadata 勉强将它们串联，员工不得不频繁跨系统切换并编写大量样板代码。为了终结这一割裂局面，Stripe 正式推出了全新的底层基语——**自定义对象（Custom Objects）**。 ```mermaid flowchart TD accTitle: Stripe 可编程收入架构组合模式 accDescr: 说明自定义对象如何与工作流和自定义应用组合构成量身定制的 Stripe 界面。 A["企业核心业务数据\n(如: 运单号 / 保单条款 / 理赔单据)"] --> B["自定义对象 (Custom Objects)"] subgraph CustomStripe["定制化 Stripe 操作系统"] B --> C["Workflows 工作流引擎\n(自动化审批 / 差异化退款 / 邮件通知)"] B --> D["Custom Apps 仪表盘组件\n(内嵌运单跟踪 / 承运商状态)"] end CustomStripe --> E["支持团队 / 财务运营\n在单一平台完成端到端日常操作"] ``` - **三位一体的组合威力**：企业将自定义对象（Custom Objects）与自定义应用（Custom Apps）及工作流（Workflows）无缝组合，即可将 Stripe 转变为一套为自身业务完全定制的操作系统。 - **物流应用实例**：物流公司通过代码定义专属字段（如追踪单号、目的地、承运商），直接在 Stripe 仪表盘内渲染出专用的业务控制台。 - **保险工作流实例**：保险公司通过 Workflows 处理复杂的退保流程，系统可依据退款金额区间自动分流至不同的审批链，并在核准后自动触发向客户发送通知邮件。 - **全屏应用拓展（Full-page Experiences）**：Stripe 宣布扩展生态 App 能力，不仅支持侧边栏小插件，更支持构建沉浸式的全屏控制台页面，“Stripe 仪表盘现已成为开发者的游乐场”。 --- ### [53:55](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3235s) 收入套件数据能力：走向毫秒级实时数据访问 在完成可编程性升级的阐述后，Tanya 随即切入收入套件的第三项核心升级——让企业数据实现**实时可用（Accessible in real time）**。企业最宝贵的商业与交易流水数据大多沉淀在 Stripe 中，长期以来 Stripe 一直致力于消除数据孤岛：从最初支持企业编写 SQL 查询以生成自定义财务报表的 **Stripe Sigma**，到后续打通将支付数据自动化同步至客户自有数据仓库的 **Data Pipeline**，数据的流转速度与可用边界正被持续推向极致。 ### [54:23](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3263s) Stripe Database 正式发布与 Revenue 营收套件演进 在开发者长期以来的强烈呼声下，Stripe 终于推出了专门针对实时业务洞察打造的全新产品——Stripe Database。长期以来，用户一直希望能够在一个规范、结构优良的数据库中直接访问自身的 Stripe 业务数据。Stripe Database 本质上是一个结构化的 Postgres 数据库，无论是人类工程师还是 AI 智能体（Agent），都可以通过它在毫秒级时间内了解企业“此刻正在发生的事情”，而非几分钟前的延迟陈旧数据。Stripe 透露，该数据库即将在后续更新中加入写入权限（write access）、自定义对象支持（custom objects）以及第三方存储连接能力。目前 Stripe Database 已面向开发者开启私有预览（Private Preview），可在官方文档页面申请测试资格。 除了数据底座，Stripe 也针对核心营收工具链发布了多项重磅更新，特别是针对大客户企业合同管理场景的“Stripe Billing 合同支持（Contracts Support）”。未来，企业无须在外部复杂系统中拼凑工具，直接在 Stripe 原生界面中就能完成从自定义价格覆盖（custom pricing overrides）、预付信用额度承诺（prepaid credit commitments）到各类合同修订与增补协议（amendments）的全生命周期建模。至此，Stripe 营收套件构建起了一整套完整体系：通过收购与自研的 Metronome 和 Tempo 支撑流式用量计费，借助自定义对象、工作流和定制 App 全面扩展 Stripe，在控制台内集成全页面公共应用，借助 Stripe Database 实时释放数据价值，并以原生合同功能完成销售订单计费的闭环。随后，演讲者将舞台交给了 Nithya，开启资金管理板块的重磅发布。 - **Stripe Database 核心能力**：基于 Postgres 构建，专为实时业务与 AI Agent 打造，近期将上线写入访问、自定义对象及第三方存储对接。 - **Stripe Billing 合同支持**：原生支持企业销售合同建模，涵盖自定义调价、预付费信用额度与合同变更修正。 - **营收套件协同体系**：覆盖用量流式计费、控制台扩展应用、自动化工作流与统一实时数据。 ```mermaid flowchart LR accTitle: Stripe Revenue Suite Architecture accDescr: A diagram showing the integration between real-time data, billing, and enterprise customization. subgraph DataEngine["数据与架构底座"] SDB["Stripe Database (Postgres)"] CO["自定义对象与工作流"] end subgraph BillingEngine["营收与计费核心"] Metro["Metronome & Tempo (流式计费)"] Contract["Stripe Billing (合同支持)"] end subgraph Presentation["控制台呈现"] Apps["全页面控制台公共 App"] end DataEngine --> BillingEngine BillingEngine --> Presentation ``` ### [55:49](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3349s) 全新 Stripe Treasury：重塑全球初创企业的财务操作系统 新一代企业正在以惊人的速度崛起：它们规模极度精简、高度依赖自动化、业务瞬时全球化，扩张节奏远超以往任何一代初创团队。然而，与其火箭般上升的业务相比，企业背后的资金管理方式却严重脱节，绝大多数创始人的财务流水仍然散落在半打分散的银行应用、转账工具和电子表格之间，陷入了无法规模化扩张的繁琐泥潭。虽然 Stripe 曾在去年夏天推出 Financial Accounts 允许用户在收款处直接沉淀资金，但对于全球化高速增长的初创企业而言，仅仅“把钱放在一起”远远不够——资金必须在需要时瞬间流转，闲置资本必须在企业睡觉时自动创造收益（"earning while you sleep"），各类付款工作流也必须全面实现自动化。 为了终结这些资金管理的摩擦，Stripe 正式推出了全新升级的 Stripe Treasury。Nithya 将其定义为“为初创企业打造的全新财务操作系统”，其底层同时由传统法定货币（Fiat）与现代稳定币（Stablecoin）两条轨道共同驱动。在多币种金融账户中，企业可以毫无壁垒地并排持有美元、英镑、欧元以及加密稳定币，随时实现即时汇兑与账户间划拨。Nithya 更是打趣道：“哪怕在周末和法定节假日也能秒级兑换——我不知道大家周末都在干什么，反正我只用来换汇。”在美国，这些账户均附带本地路由号码并享有 FDIC 存款保险。此外，企业可直接开通由万事达卡（Mastercard）支持的 Stripe 企业卡直接从余额消费，支持在后台精细配置支出限额与审批规则，且全面提供高达 2% 的无上限、全品类现金返还（"no special categories, no blackout windows"）。不仅如此，企业在 Treasury 沉淀闲置资金还能自动赚取抵扣额度，用于直接冲抵 Stripe 的各类服务手续费。目前，全球已有超过 50 个国家和地区的创业者直接将 Stripe Treasury 选作自身的主力商业银行账户。 - **多币种混合账户**：并排持有法币（USD/GBP/EUR）与稳定币，全天候 7×24 小时（含周末与节假日）秒级转换。 - **合规与保障**：配备本地清算路由号码，美国境内账户受 FDIC 存款保险保障。 - **支出卡与双重激励**：Mastercard 企业卡消费享 2% 全品类无门槛返现；账户资金闲置奖励积分可直接抵扣 Stripe 账单手续费。 ### [58:44](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3524s) 贯通 160 国的资金流转、Link 稳定币钱包与 B2B 零手续费即时转账 除了账户储存与本地支出，Stripe 正在将 Treasury 打造成全球速度最快、操作最简便且成本最低的跨境资金流动网络。如今，用户只需输入一个电子邮箱地址，就能向世界各地的任何人快速付款或汇出资金，支持的收款目的地国家在一年内从 50 个暴增三倍以上，扩展至 160 个国家。然而，跨境资金划拨的最大堵点往往发生在“最后一公里”的接收端。例如 Meta 希望利用稳定币向位于菲律宾与哥伦比亚的创作者快速、稳定地发放创作者报酬，但受制于收款人大多没有自己的加密货币钱包，落地极其困难。为此，Stripe 宣布旗下拥有 2.5 亿庞大消费者的 Link 网络正式原生支持存储稳定币。这意味着全球企业可以直接将稳定币款项无缝打入消费者的 Link 账户，终端用户无需理解助记词、公私钥等任何底层加密复杂性。 更震撼的突破发生在企业之间的网络效应之中。如果把 Link 视作一个庞大的 C 端消费者网络，那么由全球最顶尖品牌与高成长初创企业构成的 Stripe，本身就是一个庞大的 B 端企业生态网。数据显示，Stripe 生态内的企业之间每天发生相互采购与付款的高达 480 万次。为了彻底击碎这 480 万次交易中的摩擦损耗，Stripe 宣布正式上线“即时转账（Instant Transfers）”功能：所有处于 Stripe 网络内的企业，在彼此的 Treasury 账户之间转账享受“零手续费”且“瞬时到账”。正如 Nithya 在台上的总结：“即时（Instant）且免费（Free）——在传统金融领域，你绝不可能把这两个词放在同一句话里。” - **邮箱极简付汇**：覆盖全球 160 个国家，只需收款方邮箱即可发起全球汇款。 - **Link 赋能稳定币无感出海**：2.5 亿 Link 消费者可直接接收企业发放的稳定币，免去创建与配置 Web3 钱包的复杂流程。 - **Stripe 网络 B2B 互联**：承载每日 480 万次企业间付款，在 Treasury 账户间彻底实现免费秒级清算。 ```mermaid flowchart TD accTitle: Stripe Instant Transfers and Global Payouts accDescr: Flowchart showing how businesses move money via Stripe Treasury to other businesses, Link consumers, and global recipients. subgraph Enterprise["付款企业 (Stripe Treasury)"] Sender["发送方账户"] end subgraph InternalTransfer["Stripe 网络内部"] Receiver["接收方 Treasury 账户"] end subgraph GlobalPayouts["全球与消费者终端"] LinkWallet["Link 钱包 (2.5 亿消费者)"] GlobalBank["160 国本地金融机构 / 邮箱收款"] end Sender -->|"即时 & 免费 (B2B 互转)"| Receiver Sender -->|"稳定币无缝发放"| LinkWallet Sender -->|"法定货币 / 跨境清算"| GlobalBank ``` ### [01:00:41](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3641s) 现场演示：借助 MCP 协议让 AI Agent 全托管企业资金划拨 对于当今追求极速进化的初创团队而言，光是手动在网页上“管理资金”已经落后了，企业真正需要的是用代码对资金进行深度编程。为此，Stripe 将 Treasury 的全部底层 API 接口均接入了模型上下文协议（MCP, Model Context Protocol），使企业可以自由借助各种 AI 智能体来直接接管财务。Stripe 工程师 Felix 登台进行了现场真机演示。演示中，Felix 展示了一个已经连接了 Stripe MCP 服务器并获得 Treasury 授权的 ChatGPT 界面。 ``` > chat, show me the money in my Stripe balance. ``` 几秒钟内，ChatGPT 便迅速调用接口，在聊天界面中直接原生渲染出了一个精美的 Stripe 嵌入式组件（Embedded Component），资金结构与大家熟悉的 Stripe 控制台一模一样。紧接着，Felix 演示了一个更为复杂的场景：直接上传一份发票的 PDF 附件，并在指令框中输入“Pay this invoice!!”希望 AI 协助付清账单。背后的 Agent 迅速开始阅读并抽取发票核心要素——核实到期日、查看可用付款方式、计算净额结算条款、确认采购明细，并校验以往是否有过重复支付。现场发生了一个充满趣味的插曲，当 Agent 持续思考并多花了数秒时，Nithya 在旁边风趣地打趣：“Felix，你指令没用全大写，所以它才反应这么慢！”全场随即爆发笑声。 在深度比对后，Agent 聪明地识别出发票的收款方同样注册了 Stripe 账户，这意味着可以直接通过网络发起即时且免手续费的 Treasury 转账。为了规避大额财务风险，Agent 仔细核验了付款方的账户余额，并在触发 1300 美元的付款前启动了多重人机确认流程。面对连续三次的确认弹窗，Felix 笑着对观众说：“AI 安全第一（AI safe）。”当指令最终确认后，发票付款成功流出。Felix 随即切换到收款方的 Treasury 演示应用界面轻点刷新，1300 美元已原封不动、毫无延迟地躺在收款方的可用余额中，完整展现了通过 MCP 与 Agent 在 Stripe 生态中完成企业间转账的全自动化体验。 - **AI 对接能力**：Treasury API 全面适配 MCP，支持现成 Agent 工具（如 ChatGPT）或企业自建 Agent 架构。 - **上下文感知发票解析**：Agent 自动抽取 PDF 发票条款，核对历史账单并自动判定对方是否处于 Stripe 网络。 - **安全交易保障**：大额支付集成多重确认防护，在实现全自动化的同时确保人机协同的安全边界。 ### [01:04:09](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3849s) Treasury 全球版图扩张与稳定币基础设施新篇章 在精彩的现场演示后，Nithya 对 Stripe Treasury 的战略布局进行了全面回顾。作为一个致力于服务全球化企业的金融操作系统，Treasury 在今年集成了全新的收益抵扣机制、Stripe 网络间即时免费转账、160 国全球付款路由、面向 Link 消费者的稳定币付款链路，以及面向未来的 Agent-ready 编程接口。 在拓展版图方面，Stripe 展现出了极具野心的全球推进时间表：目前 Stripe Treasury 已经在 119 个国家和地区正式上线（包括美国、英国以及欧洲的预览版），且其中有 100 个国家完全依托稳定币轨道作为底层支撑。在接下来的规划中，Treasury 将继续拓展至澳大利亚和加拿大，并额外在 41 个国家引入稳定币支持，从而将覆盖国家的总版图推升至 160 个。 演讲最后，Nithya 引出了推动这套现代金融网络的最核心支柱——加密稳定币。随着 Henry 登台，全场话题自然过渡至底层加密架构。正如 Henry 所言，谈论 2026 年的资金管理，绝不可能绕开稳定币。当下业界正在目睹一场深远的范式转移：无论是小型创业团队还是跨国巨头，各行各业构建在稳定币之上的速度，正如数十年前企业向云计算的群体性大迁移。为了加速这一进程，Stripe 正在加倍下注——一方面依托 Bridge 的资金编排平台赋能全球任意企业利用稳定币调拨跨境资金；另一方面通过 Privy 构建更优雅的嵌入式钱包，让所有普通开发者都能以极低门槛将加密技术无缝融入自身产品中。 ### [01:05:51](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3951s) 针对稳定币与加密轨道的全新基础设施：Digital Asset Accounts Stripe 宣布正与加密风投机构 Paradigm 联合孵化 **Tempo** —— 一个专为支付场景深度定制的底层区块链。演讲者指出，Stripe 已经在其内部核心平台中大量应用了此类加密原语，并取得了立竿见影的成效：资金结算不再需要苦等传统的 T+2 日，而是变成了即时结算；企业无需针对每个国家挨个接入繁琐的银行通道，一套系统就能覆盖全球；服务客户不再受限于用户所在的国家或法定货币；加之可编程支付轨道（Programmable Rails）的特性，资金流转在转移几百万美元与划转百分之一美分时同样顺畅。这就是“互联网原生货币（Internet-native money）”应有的模样。 秉承 Stripe 一贯“不把能力锁在自研产品内部”的开源分享哲学，官方希望将 Bridge、Privy 以及 Stripe 正在使用的整套底层技术开放给所有开发者。稳定币历经多年演进，如今已深入世界上业务最繁忙的商业机构中。但企业要做好加密金融，面临的诉求远不止于安全存放资产，还涉及合规交易、赚取收益（Yield）、交易路由、虚拟/实体发卡以及余额流动性管理。在过去，这往往意味着要对接数十家供应商与繁冗的 API，而现在 Stripe 通过统一的全新产品彻底打破了这一壁垒。 Stripe 正式推出 **Digital Asset Accounts**（数字资产账户）。它将构建加密资产和稳定币金融产品所需的全部功能打包在一个统一的 API 之中： - **跨代币与跨链支付**：支持多种代币与公链间的即时转账。 - **法币出入金通道（Fiat On/Off-ramps）**：内置原生出入金能力。 - **链上外汇兑换（On-chain FX）与生息收益（Integrated Yield）**。 - **灵活托管（Flexible Custody）**：适配多样化安全架构。 现场台下起初反应较为安静，讲者幽默打趣：“看大家没什么掌声，我就知道现场很多朋友还没真正摸过稳定币支付轨道。不过没关系，我们会改变这一点的，因为现在接入已经变得太容易了！” 随后引来全场热烈掌声。目前，Digital Asset Accounts 已经开放给全球开发者，其底层技术正在驱动 Stripe Treasury 与一键结账系统 Link。诸多金融科技与传统巨头也已基于这套架构构建产品：**Ramp** 利用它加速国际化扩张；全球薪酬平台 **Deel** 用其向跨国自由职业者即时发放薪酬；外卖平台 **DoorDash** 则利用稳定币更加迅速地向入驻餐厅结算营收。 除了成熟巨头，新一代更敏捷、全球化的初创公司也正在该架构上崛起： - **Hyperliquid**、**Chipper Cash** 与 **Art Finance** 等公司全面依托稳定币拓展全球业务。 - 跨境汇款初创企业 **Felix**（讲者特别澄清不是台旁做 Demo 的同名工程师 Felix）：通过 WhatsApp 实现极简跨境转账，目前独家承担了**美墨汇款走廊（US-Mexico Corridor，全球规模最大的汇款通道）超过 5% 的交易量**。这意味着从美国汇往墨西哥的资金中，每 20 美元里就有 1 美元是借助稳定币通过这家初创公司流转的。 - 讲者总结道：“这一切的核心从来不是‘为了加密而搞加密’，而是为了给你的客户与用户打造在体验上彻底颠覆传统金融的绝佳产品。” ```mermaid flowchart TD accTitle: Digital Asset Accounts 架构与应用流转 accDescr: 展示 Stripe Digital Asset Accounts 通过统一 API 连接底层能力与上层各场景客户 subgraph CorePlatform ["Stripe Digital Asset Accounts (统一 API)"] CAP1["跨代币 / 跨链支付"] CAP2["法币出入金 (On/Off-ramps)"] CAP3["链上外汇 (FX) 与集成收益 (Yield)"] CAP4["灵活资产托管 (Flexible Custody)"] end subgraph Internal ["Stripe 自身原生产品"] T1["Stripe Treasury"] T2["Stripe Link"] end subgraph Enterprises ["全球企业级客户"] E1["Ramp (跨国扩张)"] E2["Deel (全球员工即时发薪)"] E3["DoorDash (商户快速结算)"] end subgraph Startups ["新兴全球化初创生态"] S1["Hyperliquid / Chipper Cash"] S2["Felix (WhatsApp 美墨汇款 5%+ 份额)"] end CorePlatform --> Internal CorePlatform --> Enterprises CorePlatform --> Startups ``` --- ### [01:09:37](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=4177s) 面向平台的全面进化：Platform Growth Studio 登场 演讲舞台交还给 Nideka。她激动地表示，这是她个人最喜欢的一部分，也是全场压轴重头戏。当前，Stripe 已连接并支撑了全球超过 16,000 家大型平台，间接服务了逾 1,100 万家终端企业客户。从 Shopify、DoorDash 到 Substack，行业巨头均选择以 Stripe Connect 作为底层金融中枢。平台型架构是金融服务渗透至全球经济每一个角落的核心引擎。为此，Stripe 将本届大会提及的几乎所有产品均针对平台场景进行了原生化重构： - **Treasury 嵌入式组件（Embedded Components）**：向平台客户开放 Treasury 全套能力，Squarespace 等企业已借此构建自身的金融中心。 - **平台专属 Radar（Radar for Platforms）**：除拦截欺诈交易外，重点防范虚假与欺诈性账户注册。今年新增了由 AI 生成的风险评分（Risk Scores）、欺诈成因深度归因解析与操作处置建议，帮助企业以更快速度封禁恶意账户。使用 Radar 的 StyleSeat、Vimeo 及 Fair Harbor 等平台，**平均每个欺诈账户造成的损失直接降低了 5.3 倍**。 然而，平台规模越做越大，进一步挖掘增量收入的难度也随之剧增。针对这一核心痛点，Stripe 正式发布全新工具 —— **Platform Growth Studio**。这是一个基于全平台海量真实数据训练的智能推荐引擎，能够主动识别并推荐具有最大商业回报潜力的增长方案。 工程师 Felix 在现场进行了端到端的交互演示： 1. **全局概览与指标监控**：仪表板直观呈现平台的总交易额（Volume）、净营收（Revenue）、利润率（Margin）、提成率（Take Rate）以及核心头部客户账户。 2. **AI 智能建议与同行基准比对**：系统将影响最大、ROI 最可观的增长机会置顶展示。演示中，系统提示当前平台的 ACH 转账定价存在严重漏损。下钻分析后发现，该平台在过去一个月内因此少赚了 13,000 美元，每笔 ACH 交易的提成率比行业均值低了整整 75 个基点（bps）。图表中用绿色曲线展示平台自身实际 Take Rate，橙色曲线则是由全网同行数据生成的基准标杆，清晰暴露出在变现能力上的差距。 3. **收益模拟器与快速采纳**：用户可在内置模拟器中拖动配置参数，或者直接选择“对齐 Stripe 行业基准（Stripe Benchmark）”，系统随即结合历史交易体量秒级预估出调整后的额外收益，点击保存即刻生效。 4. **高毛利产品推广活动（Campaigns）**：系统进一步分析发现，先买后付（BNPL）能每月为该平台多带来 12,000 美元收入，但平台下属子账户的实际渗透率仅为 13%。运营人员可直接点击创建营销推广活动（Campaign），以 1.5% 的优惠费率向所有符合条件的合资格商户群发触达，短短几步点击即可全量上线，新费率当即生效。 --- ### [01:13:43](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=4423s) 资本与发卡业务重磅更新：信贷额度、AI Agent 发卡与全球稳定币发卡 Nideka 紧接着公布了在 Capital 与 Issuing 业务线上的最新重磅进展： **1. Stripe Capital 演进** 一项持续两年的跟踪研究表明，接入并接受 Stripe Capital 资金支持的企业，其业务增速比未接受注资的企业**整整高出 27 个百分点**。为了提供更高资金灵活性，Stripe Capital 正式推出全新形态 —— **授信额度（Lines of Credit）**，允许小微企业随借随还，仅在实际提款时产生利息。同时，Stripe Capital 业务正式拓展至法国和德国，后续很快将进入澳大利亚与加拿大市场。 **2. Stripe Issuing 迎来自 2018 年上线以来的最大规模升级** Navan、Rippling、Instacart 和 Affirm 等平台与 Fintech 客户已通过 Stripe 发行了超过 3.5 亿张卡片。本次核心升级包括： - **Radar for Issuing**：差旅与费用管理平台 Navan 部署该风控后，欺诈类交易拒付（Fraudulent Disputes）大幅降低了 36%。 - **个人借记卡扩展**：在现有企业卡与消费金融体系之外，将借记卡（Debit Cards）正式纳入个人用户发行计划中。 - **AI Agent 发卡支持（Issuing for Agents）**：专为智能体经济设计，允许开发者在极度安全合规的风控沙箱内，为各类自动化 AI Agent 生成专属支出卡与预付卡，赋予全新经济实体自主采购与结算能力。 - **统一 API 稳定币发卡与全球覆盖**：开发者使用与现有法币发卡完全相同的 API，即可一键开通稳定币抵押发卡业务。Phantom、Ramp、Jeeves 和 Zaps 等创新机构已允许用户直接刷稳定币在任何支持 Visa 的 POS 机与线上商户处消费。得益于稳定币轨道，**Stripe Issuing 目前已一举覆盖全球 100 个国家和地区**。 ### [01:15:46](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=4546s) 平台与金融深度融合：Shopify 的金融演进与稳定币布局 大会回顾了基于 Stripe 数据的全新增长工作室（Growth Studio），以及 Capital 与 Issuing 功能的持续拓展，并重申开发者现已能通过数字资产账户在稳定币上构建出色的产品。随后，Shopify 幕僚长 Kate Royer 受邀登台与主持人 Will 展开炉边对谈。Will 指出，大众往往将 Shopify 视作纯粹的 SaaS 软件企业，但事实上，Shopify 在过去十年间已悄然成长为全球规模最大的金融机构之一。 谈及这一战略转型背后的考量，Kate 表示，随着平台上的商家不断扩大规模，他们迫切需要全球收单、先进的风控反欺诈工具以及多币种结算支持。为此，Shopify 与 Stripe 深度合作打造了金融服务套件，例如运行在 Stripe 底层之上的 Shopify Balance。目前该产品主要面向美国市场，深受商家喜爱，显著简化了日常商业银行业务，未来计划借助 Stripe 的全球网络加快拓展步伐。此外，双方在 Shopify Capital 等金融工具以及数字资产资金流转领域展开了紧密合作，甚至包括稳定币存储等方向；Kate 还透露，Shopify 将成为 Tempo 网络的验证者节点（Validator Node）。 - **Shopify Balance**：构建于 Stripe 金融基础设施之上，专注于满足商家的日常商业银行账户与资金管理需求。 - **数字资产与区块链合作**：双方在稳定币存储与资金移动上展开深度协同，Shopify 确认将作为验证者节点接入 Tempo 网络。 --- ### [01:18:09](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=4689s) 目录与标准革新：ACS、结构化商品目录与 UCP 协议联盟 谈及推动商家业务增长，Shopify 宣布其全球商品目录（Shopify Catalog）已接入 Agnostic Commerce Suite (ACS)，使得自主商业 Agent（AI 智能体）可以直接在体验中检索并呈现 Shopify 平台上的商品。Kate 强调，Shopify 在数据治理上投入了巨大精力，确保目录具备结构化数据、清晰的属性定义和实时准确性，这对 AI 商业至关重要——无论是搜索“运往加拿大的黑色 Vuori 打底裤”还是“Glossier 的赤褐色 Boy Brow 染眉膏”，系统都能精准响应。该 Catalog API 不仅支持千万级，更能直接连接数百万家店铺与数十亿件商品。 在行业生态层面，Stripe 宣布正式加入由 Shopify 与 Google 牵头发起的通用商业协议（Universal Commerce Protocol，简称 UCP）总委员会。Kate 解释，UCP 最初的愿景就是探索能否建立一个贯穿完整商业旅程、兼容任何平台并确保交易安全可信的统一协议。协议的生命力在于行业采纳，上周 Meta、Amazon、Salesforce 和 Microsoft 等科技巨头已相继加入。UCP 为商家和开发者消除了生态碎片化，提供了一套统一的开发标准与规范。 ```mermaid flowchart TD accTitle: Universal Commerce Protocol (UCP) Ecosystem accDescr: 展现由 Shopify、Google 和 Stripe 等科技巨头联合推动的通用商业协议架构 subgraph Governance["协议治理与生态联盟"] Founders["发起方<br/>Google & Shopify"] Members["总委员会新成员<br/>Stripe, Meta, Amazon, Salesforce, Microsoft"] end subgraph Standard["通用商业协议核心 (UCP)"] Spec["单一开发规范与标准接口"] Trust["端到端可信交易与数据流转"] end subgraph Consumers["接入与应用层"] Merchants["全球商户 / 开发者"] AIAgents["AI 商业 Agent / 消费体验"] end Founders --> Standard Members --> Standard Standard --> Merchants Standard --> AIAgents ``` - **Catalog API 规模**：覆盖数百万独立店铺，索引商品总量达数十亿级别。 - **UCP 联盟阵营**：由 Shopify 和 Google 联合发起，吸纳了 Meta、Amazon、Salesforce、Microsoft 以及 Stripe 等核心科技平台加入总委员会。 --- ### [01:20:57](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=4857s) Agent 商业前瞻与 Shopify 的复杂履约底座 在展望未来两至三年的 Agent 商业发展时，Kate 分享了她对消费行为变迁的观察。除了让 AI 根据个人风格、尺码和预算推荐日常穿搭外，Agent 正变得极度擅长发掘那些“令人惊喜的小众心头好”。她分享了一个亲身经历：团队一位工程师刚刚喜得贵子，她的 AI Agent 推荐了一本限量印制的极客婴儿绘本《婴儿计算机工程学》（Computer Engineering for Babies），不仅风趣而且极为贴心。过去这种高契合度的礼物全凭机缘偶遇，而未来 Agent 将轻松实现这种精准匹配。 然而，Kate 犀利地指出，Agent 商业的门槛远不止是将商品数据源接入大模型。一个真正的商业闭环要求系统必须在瞬间完成库存校验、税费计算、会员积分与折扣套用、支付处理、反欺诈拦截，并最终将商品安全送达消费者家门口。这正是 Shopify 过去 20 年来深耕的核心壁垒所在——将极其复杂的后端商业生命周期，封装成顺畅且无缝的“魔法体验”。 --- ### [01:23:15](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=4995s) Stripe Signals 发布与 2027 公开路线图上线 重返舞台后，发言人回顾了整场发布会的核心主线：智能体的能力完全取决于驱动它们的数据与底层模型。过去一年中，全球 GDP 的 1.6% 经由 Stripe 处理，系统内流转了高达 70 万亿条商业数据。以往，基于这些海量数据训练的内部模型仅作为幕后引擎赋能 Stripe 自有产品，但开发者更希望将这种网络级智能作为一种可编程、高弹性的底层积木直接调用。 为此，Stripe 正式推出了全新产品——**Stripe Signals**。开发者与 Agent 可以直接向 Stripe 传输一个客户、一笔交易或一家企业（无论其是否在 Stripe 生态内），系统将实时返回评分及详细解释，直接回答“此笔交易是否有风险”、“该企业是否可信”以及“如何促进该客户转化”等关键商业问题。Cursor、DoorDash、Navan、Cloudflare 和 ElevenLabs 等企业已率先接入，产品首期上线 10 种 Signals，后续还将不断丰富。 最后，针对开发者多年来的呼声，Stripe 正式上线了官方公共路线图（stripe.com/roadmap）。该路线图包含精确到季度、数百个详细条目的产品规划，一路排期至 2027 年第一季度，并将随着产品高速迭代持续动态更新。发言人表示，Stripe 致力于成为全球进化速度最快的基础设施，底层架构级的深度革新正在全面展开。 - **Stripe Signals 核心能力**：基于 70 万亿条商业数据，向外部系统提供网络智能打分与实时决策解释。 - **首批标杆客户**：Cursor、DoorDash、Navan、Cloudflare、ElevenLabs。 - **公开路线图上线**：访问地址为 `stripe.com/roadmap`，详尽列举了至 2027 年 Q1 的数百项功能与架构升级规划。 ### [01:26:11](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=5171s) 激活网络效应与构建 AI 经济基础设施：主题演讲总结与展望 演讲者在主题演讲的尾声总结了 Stripe 的核心技术演进与战略使命，强调 Stripe 致力于实现深度的可编程性（programmable），让客户不仅能按需将绝大部分业务平滑迁移并运行在 Stripe 平台上，更能直接基于其底层基元（primitives）从零孵化出前所未有的全新商业模式。同时，Stripe 正在全面激活其庞大的网络效应。在场的每一家企业共同构成了全球增长最迅猛的公司联盟，而 Stripe 网络的职责正是作为强大的护航者与引擎，全力保护并推进这些业务向前飞速扩张。 更关键的是，Stripe 正全力为人工智能时代打造底层的“经济基础设施”（economic infrastructure for AI）。演讲者特别提醒台下的创业者与开发者，宏大的 AI 时代并非遥不可及的未来愿景，而是“此时此刻正在发生”（"It's happening right now"）。他期待参会者在今天离开会场时，脑海中能满载切实具体的构想，去主动把握住眼前的历史性机遇。整个行业的未来形态将由台下各位所构建的产品以及双方携手打造的技术生态所共同塑造，并且这一转变的步伐将会极其迅疾。在全场的掌声与激昂的配乐中，演讲者向现场参会者致谢，并正式开启接下来的 Stripe Sessions 系列活动。 - **核心基元与可编程能力**：赋能企业自由组合 Stripe 提供的各类技术基元，支撑更多核心运营流程并开创新业务形态。 - **Stripe 网络赋能**：将全球高增长企业连接为协同网络，发挥抵御风险与加速业务增长的双重价值。 - **AI 经济底座**：明确将 AI 基础设施作为核心战略支柱，强调技术变革正在当下以极快速度推进。 ## Notable quotes > [00:00](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=0s) day 119 of the singularity — Patrick Collison > [00:00](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=0s) money is data — Patrick Collison > [14:45](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=885s) one small step for these two agents, but one giant leap towards making agents full economic actors — Will Gaybrick > [18:01](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=1081s) People are voluntarily posting about it on the internet — Kevin > [33:23](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=2003s) Payments will pivot from being a moment to being a policy — Ginger Baker > [55:49](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3349s) earning while you sleep — Nithya > [55:49](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3349s) no special categories, no blackout windows — Nithya > [58:44](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3524s) 即时（Instant）且免费（Free）——在传统金融领域，你绝不可能把这两个词放在同一句话里。 — Nithya > [01:05:51](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=3951s) 看大家没什么掌声，我就知道现场很多朋友还没真正摸过稳定币支付轨道。不过没关系，我们会改变这一点的，因为现在接入已经变得太容易了！ — Henry > [01:26:11](https://www.youtube.com/watch?v=lIsHZfRl2zw&t=5171s) It's happening right now ## Mentioned Patrick Collison, Will Gaybrick, Kevin, Felix, Ginger Baker, Emily, Tanya, Georgios Konstantopoulos, Nithya, Henry, Nideka, Kate Royer, Andrej Karpathy, Claude Code, OpenAI Codex, ChatGPT, Gemini, Microsoft Copilot, Stripe CLI, Stripe Projects, Machine Payments Protocol (MPP), Link Agent 钱包, Link, Agentic Commerce Suite (ACS), Checkout Studio, Stripe Treasury Payments, Adaptive Pricing, Stripe Terminal, Stripe Reader S700, Stripe Reader T600, Auth Boost, Stripe Console, Stripe Radar, Smart Disputes, Stripe Revenue Suite, Metronome, Tempo, Custom Objects, Workflows, Custom Apps, Stripe Sigma, Data Pipeline, Stripe Database, Stripe Billing, Stripe Treasury, Model Context Protocol (MCP), Bridge, Privy, Digital Asset Accounts, Platform Growth Studio, Stripe Capital, Stripe Issuing, Radar for Issuing, Universal Commerce Protocol (UCP), Stripe Signals, Shopify, Shopify Balance, Shopify Capital, Shopify Catalog, Google, Meta, Amazon, Salesforce, Microsoft, OpenAI, Anthropic, Paradigm, Visa, Mastercard, 渣打银行 (Standard Chartered), DoorDash, Klarna, Klarna USD, Navan, Rippling, Instacart, Affirm, Phantom, Ramp, Jeeves, Zaps, Deel, Hyperliquid, Chipper Cash, Art Finance, Felix (汇款公司), Quince, Fanatics, JD Sports, Wix, commercetools, Best Buy, Kate Spade, Unity, RevenueCat, Lovable, Gamma, Hertz, Pepsi, ElevenLabs, Cursor, Cloudflare, Substack, Squarespace, StyleSeat, Vimeo, Fair Harbor, Vercel, Supabase, RunLoop, UPI, Pix ## Video description Learn about Stripe products designed for revenue growth. More information: https://stripesessions.com Developer keynote: https://youtube.com/live/-vRY2dtD7iQ Indexing the economy keynote: https://youtu.be/5wGqWRv1Z1s Sam Altman fireside: https://youtu.be/5eouRdDYM2c Nat Friedman and Daniel Gross fireside: https://youtu.be/l9wzs_QIyp0 Register for Stripe Sessions 2027: https://register.stripesessions.com/2027?promocode=s27youtube [Source](https://www.youtube.com/watch?v=lIsHZfRl2zw)

[0:00] Patrick Collison: day 119 of the singularity [0:00] Patrick Collison: money is data [14:45] Will Gaybrick: one small step for these two agents, but one giant leap towards making agents full economic actors [18:01] Kevin: People are voluntarily posting about it on the internet [33:23] Ginger Baker: Payments will pivot from being a moment to being a policy [55:49] Nithya: earning while you sleep [55:49] Nithya: no special categories, no blackout windows [58:44] Nithya: 即时（Instant）且免费（Free）——在传统金融领域，你绝不可能把这两个词放在同一句话里。 [65:51] Henry: 看大家没什么掌声，我就知道现场很多朋友还没真正摸过稳定币支付轨道。不过没关系，我们会改变这一点的，因为现在接入已经变得太容易了！ [86:11] It's happening right now

[11:21] Will Gaybrick: You've seen a payment or two on this stage, but you've never seen one like this, because this is fully autonomous. Two agents attempting a real value exchange right before our eyes. [21:16] Felix: I barely remember how to code anymore. So, this handy-dandy 'copy for LLM' button, I hit that and I have something to hand to my agent. [29:08] Ginger Baker: What is money other than another form of digital content? And so, if we aren't able to make money move as seamlessly as other forms of digital content through our surfaces, it feels like we're missing out on a real opportunity to help people connect with one another. [55:56] Nithya: I don't know what you do on weekends; I only convert currencies [58:44] Nithya: Starting today, businesses on the Stripe network can send money between Treasury accounts at zero cost... Instant and free, two words you don't usually hear together in finance. [65:21] Henry: This work is not about crypto for crypto's sake. It's about enabling you to build fundamentally better experiences for your customers, your users, and their money. [76:00] Kate Royer: We really discovered that merchants, especially those at scale... need global acquiring, they need sophisticated fraud tooling, they need multi-currency payouts.

[0:00] Good morning and welcome to Stripe [0:03] Sessions. [0:06] >> [cheering] [0:07] [applause] [0:10] >> It's April 29th, [0:12] otherwise known, of course, as day 119 [0:16] of the singularity. And we'll come back [0:18] to this in just a few moments. This is [0:20] the biggest Sessions, as you can see, [0:22] that we've ever put together, and we're [0:24] very grateful to all of you for joining [0:27] us, for spending the time with us, and [0:29] we're extremely excited to show you [0:31] what's been cooking. [0:33] Whether you're here for the talks, or [0:35] the demos, or the hallway conversations, [0:37] or anything else, Sessions is built for [0:40] all of us, and there's really a lot [0:42] going on in the internet economy. [0:45] Now, part of what I love about Sessions [0:47] is that we businesses from so many [0:49] categories and domains. We've AI [0:52] leaders, platforms, enterprises, [0:55] startups, e-commerce giants, solo [0:58] developers, [1:00] you all, across the full spectrum of [1:03] commerce and industry, are ensuring that [1:06] the new miracles of science and [1:07] technology are actually deployed in the [1:10] real world, and we feel very honored to [1:12] support you. [1:15] To give you a quick update, Stripe now [1:17] serves more than 5 million businesses [1:20] around the world, [1:22] including 86% [1:24] of the Forbes AI 50, 90% of the [1:27] companies in the Dow Jones index, and 80 [1:30] of the Nasdaq 100. [1:34] Businesses on the Stripe platform, all [1:36] of you, [1:37] collectively generated revenue that's [1:39] equivalent to about 1.6% [1:42] of total global GDP last year, [1:46] quite remarkable. And this works out to [1:48] about 1.9 trillion dollars in total [1:51] payment volume, and this figure is up [1:53] 34% year-over-year. [1:57] >> [applause] [2:02] [applause] [2:03] >> And that's particularly striking when [2:04] you set it alongside global growth of [2:07] just 2%. And so businesses that build on [2:09] Stripe, again, all of you, are [2:11] dramatically outpacing the rest of the [2:13] economy. [2:15] But it's not just that growth. [2:17] We work really hard to ensure that [2:19] Stripe remains the steadiest and the [2:21] most reliable part of your stack. And [2:24] last year, our critical APIs retained [2:27] 99.9996% [2:30] availability while we on average [2:35] >> [applause] [2:39] >> While we on average merged well over [2:42] 1,000 production pull requests every [2:45] day. And so Stripe is getting better [2:47] faster than it ever has before. [2:52] So day 119 [2:54] of the singularity. [2:56] Now, we're being a bit tongue-in-cheek [2:57] with this, of course, but but only a [2:59] bit. And [3:01] the reason we kind of frame things this [3:02] way is something, as far as we can tell, [3:05] really has changed over the last couple [3:07] of months. [3:09] So I want to show you a chart we've [3:10] actually never publicly shown before. [3:13] This is the number of new businesses [3:15] launching on Stripe every month. And [3:18] start here in 2019, you can see this big [3:21] spike at the beginning of the pandemic. [3:23] Actually, the interesting thing about [3:24] that is it never went back to baseline. [3:27] It remained at this elevated rate. [3:31] But since the beginning of this year, [3:33] the beginning of 2026, [3:36] we have started to see this incredible [3:38] inflection, this parabolic rise in new [3:42] firm creation. [3:45] And this acceleration that we're looking [3:46] at, no, it's about lots of new company [3:49] formation, [3:51] but what's happening, as you guys know, [3:53] is really not only about new startups. [3:57] Because of AI, the entire economy is [4:00] re-platforming. The entire economy. New [4:03] products, new business models, and [4:06] completely new sectors. [4:09] And every business now faces two [4:11] imperatives. [4:12] One, to figure out how you're going to [4:15] reinvent your product. And then second, [4:19] to reimagine how the business works. [4:23] And interestingly, programmable [4:25] financial infrastructure is crucial for [4:27] both of these. [4:29] You know, from the start, we've been big [4:31] believers in the idea that financial [4:33] infrastructure should converge with [4:36] code. You know, I grew up as a Lisp [4:38] programmer, and in Lisp, the core idea [4:40] is that code is data. [4:42] Well, the basic idea in Stripe is that [4:44] money is data. [4:46] And when we launched back in 2011, you [4:48] know, this was not exactly industry [4:51] orthodoxy, but it turned out to be a [4:53] pretty impactful ideas software came to [4:56] take over the world. [4:58] But, if this thinking was useful in [5:01] pre-singularity times, standing here in [5:04] 2026, [5:06] it's obviously become quite essential. [5:10] And so, Stripe is building the economic [5:12] infrastructure for the AI era, [5:15] and you're going to hear this theme [5:17] interwoven across everything that we [5:19] cover at Sessions. [5:22] And so, this brings me to today. [5:25] We've a lot to show you. [5:26] New products, new capabilities, and new [5:29] ways that Stripe can hopefully help you [5:31] grow. And I won't spoil the details, but [5:34] I will [snorts] say that I think today's [5:36] announcements represent the most [5:38] ambitious single day set of product [5:40] launches in Stripe's history. [5:43] So, let's get into it. [5:46] >> Please welcome Stripe President of [5:48] Technology and Business, Will Gaybrick. [6:00] >> All right. [6:02] So, something fundamental changed at the [6:04] start of this year. [6:05] You know, we could all sort of feel it, [6:07] but I want to show you the moment that [6:09] it snapped into focus for me. [6:11] And because the near vertical inflection [6:14] that Patrick just showed us, it's not [6:16] just happening at the macro level. You [6:17] know, we're seeing the singularity [6:19] within our products as well. [6:22] So, this is the usage of something we [6:24] launched 7 years ago in relative [6:27] obscurity, our command line interface or [6:29] CLI. [6:31] And for a long time, it barely moved. [6:32] You know, the CLI was the province of a [6:34] relatively narrow band of deeply [6:37] technical Stripe aficionados. [6:39] Then usage started ticking up just a bit [6:40] last year. [6:42] But then, early this year, [6:44] usage of the CLI exploded. [6:47] So, why? [6:49] What you're looking at here is the rise [6:51] of agents. [6:53] You know, we'd all been talking about [6:54] the agentic era for years. But in a [6:56] matter of weeks, we suddenly found [6:58] ourselves, all of us, living in it. You [7:00] know, it's here. [7:02] And at Stripe, as for many of you, [7:04] our daily reality is now building [7:06] agents, designing evals, writing skills, [7:09] and of course, redesigning our products [7:12] to be used by agents. [7:15] It's easy to underestimate the [7:16] implications of a near future where [7:18] agents are autonomous economic actors [7:21] responsible for most internet [7:23] transactions. [7:24] And just as agents work faster than we [7:26] do, they're going to spend money a lot [7:28] faster, too. [7:29] So, this is a huge economic opportunity [7:32] if only we all can seize it. [7:34] So, how should we be preparing? [7:37] We have a ton of exciting stuff to show [7:39] you on that score today. So, So, hope [7:40] your agent is diligently taking notes. [7:43] But, Sessions is also about giving you a [7:45] peek into the future. You know, less [7:46] talking and more showing. [7:48] So, what do you think? Should we just [7:49] get the demos started? [7:52] All right. [7:52] >> [applause] [7:55] >> So, we all know that agents can write [7:57] great code. [7:58] Or they can write a ton of code fast, [8:00] and some of it usually works. [8:03] What they can't do is package up an [8:05] application for real users and put it on [8:07] the public internet. Or rather, they [8:09] couldn't do that until last month we [8:11] launched a private preview of Stripe [8:13] Projects. Because Projects handles the [8:15] other side of vibe coding. Vibe [8:17] deploying. [8:18] Let's see what it can do. [8:21] With Stripe, you know we care a great [8:22] deal about APIs, in case you haven't [8:24] noticed. [8:25] So, for this demo, we built an agentic [8:27] app that takes any public API and [8:29] reviews it for legibility and developer [8:31] developer ergonomics. And all for a flat [8:34] rate of $2 per review. [8:37] You can see here that the app is just [8:38] being hosted locally on my computer. [8:41] Let's see if Claude code can one-shot [8:43] deploy this using Stripe Projects. [8:45] So, please [8:48] quickly [8:49] deploy my [8:51] app using All right, you're all watching [8:54] me type. [8:56] and agent [8:58] sandbox and a hosting. [9:03] Excuse me. [9:05] All right. [9:06] And please work because I am on stage. [9:11] Okay, so while Claude does its thing, [9:13] let me tell you a little bit more about [9:15] Stripe Projects. [9:17] So, the idea for Projects came from our [9:19] favorite source of inspiration. You all. [9:22] You know, for example, this tweet last [9:24] year from Andrej Karpathy. [9:27] And you know, he's right. [9:28] In the era of agentic coding, building [9:30] an app shouldn't feel like assembling [9:33] You're juggling a dozen different [9:34] services, creating accounts with each of [9:36] them, bouncing between dashboards, [9:38] copying and pasting secrets. [9:41] And agents can't do any of this. You [9:42] know, every single step requires a [9:44] human. [9:45] Well, projects fixes this. You know, it [9:47] programmatically provisions services, [9:49] manages usage, and it even navigates [9:51] billing, all directly from the command [9:53] line. [9:54] And best of all, it empowers agents to [9:56] do this autonomously. [9:58] So, we launched in private preview last [10:00] month, and there are many, many [10:01] thousands of you on the waitlist. [10:04] Well, right now, we're raising the gates [10:06] and making Stripe projects available to [10:08] everyone. [10:11] >> [applause] [10:15] >> So, we already have dozens of great [10:16] partners, and if you want to join the [10:17] ranks, please come see us downstairs in [10:19] the app store. Okay, let's go check in [10:21] on our friend Claude. [10:27] Okay, and it claims it worked, but [10:29] before we check, let's just go through [10:31] the output log here. So, first thing it [10:34] did was choose a set of providers. So, [10:36] Vercel for hosting, we got Supabase for [10:37] the database, and then RunLoop for the [10:39] agent sandbox. Looks like it pulled some [10:42] credentials, some secrets here. [10:45] And again, it says it was successful, [10:48] but you know, agents can be a bit overly [10:50] optimistic, so [10:51] let's just copy the URL. [10:57] All right, and we'll head back to the [10:58] browser. [11:00] Moment of truth. See if it actually made [11:02] it to the web. [11:04] And boom, there you go. The app is live. [11:07] >> [applause] [11:10] >> And I mean actually live. You know, over [11:12] the course of building this, we heard [11:13] from many of you that you actually [11:14] wanted to use it. And so, as of right [11:16] now, [11:17] it's available to all of you, and of [11:18] course, to your agents. [11:21] Okay, so we just one-shotted deployment [11:23] with projects. That's pretty cool. [11:25] But think back to that future that we [11:26] imagined earlier. You know, one where [11:28] agents are autonomous economic actors [11:30] responsible for most internet [11:32] transactions. [11:33] Well, for that future to be real, agents [11:36] need to be able to go buy things [11:37] autonomously. [11:38] And this is highly non-trivial. [11:41] Because commerce on the internet, you [11:42] know, pricing pages, checkout forms, [11:44] these were all designed for us. [11:47] And agents need entirely new modalities. [11:50] So, we built some. [11:52] A few months back, we launched the [11:53] machine payments protocol, or MPP, in [11:56] collaboration with our friends at Tempo. [11:58] Now, MPP is an open standard that lets [12:01] any service tell agents that payment is [12:03] required and how to pay, all directly [12:05] over HTTP. So, I've already used Stripe [12:08] to integrate MPP into this app. [12:11] But, even if the app is ready to accept [12:14] agentic payments, [12:15] how do we give an agent a payment [12:16] credential in the first place? [12:18] And how can we trust them to use it [12:20] responsibly? [12:22] Well, it turns out we already have [12:23] something perfect for this, Link. [12:26] So, today we're launching Link's new [12:29] wallet for agents. [12:30] It lets you securely empower agents to [12:32] spend on your behalf. Your payment [12:34] credentials are never exposed, and you [12:36] approve every purchase. [12:38] So, let's go see if we can close the [12:40] loop. Can another agent autonomously buy [12:43] an API from the agent who built and [12:45] deployed the app? [12:46] So, I was using Claude for the first [12:49] agent, the builder, but now I'm going to [12:50] use Codex. I'm going to do it in dark [12:52] mode just to keep them sort of visually [12:53] distinct. I'm not going to tell it [12:55] anything about MPP or Link. I'm just [12:57] going to ask it to get me a review for [12:58] an API from Privy. So, get [13:01] me a review [13:03] of this API. [13:06] Uh and Privy is our crypto wallet [13:07] solution for anyone not familiar. [13:10] So, [13:11] grab this URL. [13:15] All right, MD for the agent. All right, [13:17] and off you go. [13:18] So, it's going to request the review. [13:20] You can see right away [13:22] it notices that the app is using MPP and [13:24] accepts card payments. [13:27] And here you can see that the agent has [13:29] realized that it can use link, and [13:30] that's what it's attempting to do. [13:33] It's extracted the amount, so 200 cents, [13:35] so $2. [13:43] It's taking its time. [13:46] All right, and this is important. You [13:47] can see it's sending me over to the link [13:50] app for my approval. [13:52] I wish I could just command double [13:53] click, but [13:55] c'est la vie. [14:00] So, here we are in the link app. [14:02] I can see what I'm buying, a $2 review. [14:05] Uh and my my Visa card here, so let me [14:07] approve that. [14:09] Give it my thumb print. [14:13] Okay, so if you've been to the sessions [14:14] before, you've seen a payment or two on [14:16] this stage, but you've never seen one [14:17] like this, because this is fully [14:19] autonomous. Two agents attempting a real [14:22] value exchange right before our eyes. [14:25] So, fingers crossed, it's got the [14:27] approval. [14:35] And there we go. Autonomous payment [14:37] accepted. [14:39] >> [applause] [14:43] [applause] [14:45] >> So, you know, that's one small step for [14:46] these two agents, but one giant leap [14:48] towards making agents full economic [14:50] actors. [14:52] And even a few months ago, you know, [14:53] that very idea would have sounded [14:54] futuristic. But now, when you string [14:56] together straight projects, the machine [14:58] payments protocol, and the link wallet [15:00] for agents, [15:01] we're really on the precipice of agents [15:03] building, buying, and selling software [15:06] autonomously. [15:08] >> [snorts] [15:09] >> You know, we're we're in a sort of funny [15:10] moment, because we're all being called [15:12] to reimagine our businesses from the [15:14] ground up. [15:15] We still have to do the day-to-day work [15:17] of serving all of our customers. [15:19] So, So working hard to help you with [15:21] both. [15:23] And over the next hour, we'll run [15:24] through a ton of upgrades across [15:25] payments, [15:27] Radar, [15:28] revenue, [15:29] money management, and embedded finance. [15:32] And as we do, you'll notice three [15:34] themes. [15:35] So how we're making Stripe more [15:36] programmable, [15:38] how the Stripe network is protecting and [15:40] propelling your business, [15:43] and of course, how Stripe's economic [15:45] infrastructure for AI can boost your [15:47] growth in this new agentic era. [15:49] Let's dive in, starting with payments. [15:51] Here's Kevin. [16:03] >> Thanks, Will. [16:04] Payments has always been the beating [16:06] heart of your business and ours. [16:10] And as you've just seen, payments is [16:11] about to meet a new kind of buyer, [16:14] agents. [16:15] Today, I'll show you how we can help you [16:18] stay ahead. [16:19] I'll share how you can get ready for [16:21] agentic commerce, and how you can boost [16:24] your revenue with high-performing [16:25] checkouts and global payments [16:27] optimizations. [16:29] Let's start with agentic commerce. [16:31] We think of this as a spectrum, where at [16:33] one end, agents help customers discover, [16:37] but the human still chooses and buys. [16:40] In the middle, the human chooses and the [16:42] agent checks out. [16:43] And at the far end, the agent discovers [16:46] and buys. [16:47] But what we don't know yet is when and [16:50] how consumer behavior will change. [16:53] But you shouldn't have to bet on that. [16:55] And so we built the infrastructure to [16:57] help you thrive, no matter how it [16:59] unfolds. We're already seeing demand [17:01] across verticals, from retail and media [17:05] to SaaS and travel. [17:07] And let me give you a rundown of what [17:08] we've been up to. [17:10] With Microsoft and OpenAI, we're making [17:13] products discoverable inside Copilot and [17:15] ChatGPT. [17:17] With Meta, we're powering checkout right [17:20] inside ads. [17:22] And to reach customers everywhere, you [17:24] need to be where billions of people [17:26] search every day. [17:29] And so today, we're going even further [17:30] with a big new announcement. [17:32] We have partnered with Google, so your [17:35] customers will soon be able to buy from [17:37] you right inside AI mode and the Gemini [17:40] app. [17:42] >> [applause] [17:46] >> Quince, Fanatics, and JD Sports are [17:48] coming soon. [17:50] So that's Google, OpenAI, Microsoft, [17:53] Meta. Four of the biggest AI players all [17:57] shaping new commerce experiences with [17:59] Stripe. [18:01] But now, let's go to the other end of [18:02] the spectrum, agents buying directly [18:04] from you. [18:05] For that, we built the machine payments [18:07] protocol, MPP. [18:09] If you offer services via APIs, MPP is a [18:12] great way to acquire new customers. [18:15] And as you can see, MPP has already [18:18] achieved the highest honor for a [18:20] payments protocol. People are [18:22] voluntarily posting about it on the [18:24] internet. [18:26] So whether you're selling inside AI [18:28] surfaces or directly to agents with MPP, [18:31] it all runs through the Agented Commerce [18:33] Suite. [18:34] You can integrate with Stripe once, [18:36] your products become legible to agents, [18:39] you remain the merchant of record, [18:41] and Stripe handles checkout, payments, [18:44] and fraud prevention. [18:46] And today, I'm pleased to announce that [18:48] Shopify will become the preferred [18:50] catalog provider for retailers that want [18:52] to integrate with the Agented Commerce [18:54] Suite. [18:56] And we're also adding support for [18:57] platforms on Stripe. [18:59] We're making it easy for platforms to [19:01] offer this new growth channel to their [19:02] customers. And those businesses can sell [19:05] without having to integrate with agents. [19:07] Platforms like Wix and Commerce Tools, [19:10] and enterprises like Best Buy and Kate [19:12] Spade are already building with the [19:14] Agentic Commerce Suite. [19:16] And as more agents and modalities [19:18] emerge, our suite will expand with them. [19:21] So, no matter what you sell, how you [19:23] sell it, or who you sell it to, our [19:26] suite will have you covered. [19:28] But none of this scales if your [19:30] customers can't empower their agents to [19:32] make payments. And as we showed earlier, [19:35] we're making Link, the wallet for [19:37] agents. [19:39] It's already a network of more than a [19:41] quarter billion users. It drives a 5% [19:44] lift in conversion for returning [19:45] customers. And now, right in the Stripe [19:48] dashboard, you can easily see how Link [19:50] is performing for your business. [19:53] Link now supports UPI, used by over 500 [19:56] million people in India. [19:58] Pix, used by over 160 million people in [20:01] Brazil. [20:02] And stablecoins to provide global reach. [20:05] So, you can serve more customers with [20:07] the lowest cost payment methods they [20:09] already use. [20:13] Now, let's talk about what's new with [20:14] our optimized checkout suite. [20:17] So many of you use it for payment [20:19] surfaces, local currency presentment, [20:22] language localization, and more to boost [20:25] your conversion globally. [20:27] Businesses using Stripe Checkout, our [20:29] fully hosted checkout page, grew their [20:31] revenue two times faster on average. Two [20:34] times faster. [20:36] And we often hear you talk about a [20:37] flywheel to increase conversion. You can [20:40] figure new checkouts, you analyze them, [20:43] and then you optimize them. [20:44] And we want to help make that flywheel [20:46] spin faster. [20:48] And so today, I'm very excited to [20:50] announce Checkout Studio, a new place [20:53] for you to configure and optimize [20:55] checkouts for conversion. [20:59] Now, you can see all your checkouts side [21:01] by side, analyze performance. I think of [21:04] it as mugshots for your checkout flows. [21:06] Who did it? Who lowered our conversion [21:08] yesterday? [21:10] You can build your custom rules and you [21:12] can run AB tests for payment methods. [21:16] And now, many of you use Payment Element [21:18] to build checkouts on your websites with [21:20] our UI components. But you also want a [21:23] highly adaptive checkout that can show [21:25] up anywhere, from apps to chatbots to [21:29] agentic surfaces. [21:31] And so today, we're introducing a more [21:33] customizable embedded form. [21:36] And to make it really easy to build your [21:38] checkouts, we're also introducing a [21:40] visual builder. [21:41] Felix, thanks for joining us at the demo [21:43] desk. How quickly can you design an [21:45] embedded form? [21:46] >> Let's find out. [21:48] I just have to hit create checkout. And [21:51] immediately, you see the new option that [21:53] Kevin's been talking about, the embedded [21:55] form. So, let's just make one for my [21:57] support chat UI. I'll just call it [22:02] support chat UI. Easy. Now, [22:05] in the middle here is the new visual [22:07] builder. And down all the left are all [22:09] my controls and tools. So, we can, let's [22:12] say, collect customer names. And you can [22:15] immediately see the preview updating [22:17] with that. Now, it's honestly a little [22:20] bright up here, so I kind of want to [22:22] turn it down a bit. And so, let's go [22:25] dark mode. All right, that's much better [22:27] for me. [22:28] Now, I can also add custom components. [22:31] So, in this case, let's say I want to [22:33] add a checkbox and we'll say [22:37] uh [22:38] join my newsletter. [22:42] And I can even use these drop zones and [22:45] add it right here on the checkout. I [22:47] just hit save [22:50] and my checkout looks pretty good. So, [22:53] when I'm ready, I just hit implements [22:56] and they give me this lovely code [22:58] snippet right here. Except, I barely [23:01] remember how to code anymore. So, [23:05] this handy-dandy copy for LLM button, I [23:08] hit that and I have something to hand to [23:10] my agent. [23:14] >> [applause] [23:19] >> So, I just went from start to visualizer [23:22] to prompt on a few clicks. [23:25] >> Thanks, Felix. [23:27] With the new checkout studio, you can [23:29] configure and optimize a new checkout in [23:31] just a few minutes. [23:33] But now let's talk about global [23:34] expansion. [23:35] It sounds so great on paper, and then [23:38] reality sets in. Suddenly, it's taxes, [23:41] compliance, support, a very long to-do [23:43] list. [23:44] That's why we built Stripe Treasury [23:46] payments, our merchant of record [23:47] solution. [23:48] It helps companies like Unity, [23:50] RevenueCat, and Lovable to expand in 195 [23:54] markets with far less operational [23:56] complexity. [23:58] And as of today, Stripe Treasury [23:59] payments is generally available. [24:02] And we also know that localizing your [24:03] prices globally is hard. You have to [24:06] manage FX, currency exposure, and [24:09] payment method compatibility. [24:11] That's where adaptive pricing comes in. [24:13] And we've actually been surprised by its [24:15] results. A 17.8% [24:18] average increase in cross-border [24:19] revenue, [24:20] and a 5% average increase in [24:22] authorization. [24:24] It now also localizes your subscription [24:27] prices. So, subscription businesses from [24:29] startups to enterprises have seen a 4.7% [24:33] uplift with adaptive pricing. [24:35] That is a lot of additional revenue with [24:38] very little effort. [24:40] We've also done a lot of hard work to [24:42] make payment methods operate [24:43] cross-border, as well as support [24:45] recurring payments, and just behaving [24:47] the way you expect. [24:49] That's a level of standardization that's [24:51] only available on Stripe. [24:54] When Gamma, one One the fastest-growing [24:56] AI companies, started accepting UPI. [24:59] Their revenue in India jumped 22% [25:02] in the same month. Amazing. [25:05] Now, let's talk about selling in person [25:07] globally. [25:08] Over 900,000 businesses, including [25:11] enterprises like Hertz, use Stripe [25:13] Terminal for in-person payments in 25 [25:15] countries. [25:17] This year, Stripe Terminal is expanding [25:19] to 15 additional markets. [25:21] And we have an incredible roster of [25:23] terminal devices. [25:25] The S700 is our handheld device with [25:27] cellular support, and it's been a hit. [25:30] But, today I'm pleased to announce the [25:32] Stripe Reader T600. [25:35] Now, first off, I'm sure you can tell [25:36] that we had to burn a lot of tokens to [25:37] come up with that name. [25:40] But, it's a customer-facing display, and [25:43] it can power some premium checkout [25:44] experiences like loyalty programs and [25:47] upsells with custom apps. [25:50] But, revenue optimization doesn't end at [25:51] checkout. And last year, we introduced [25:54] Auth Boost. [25:55] It automatically updates cards, routes [25:57] transactions across networks, and [25:59] determines when to retry failed [26:01] payments. [26:02] Let's take a look at how it's doing. [26:04] We found businesses can see a 3.8% lift [26:07] in auth rates, [26:09] and up to 3.3% in cost savings. [26:13] And now, you can use an AB test to [26:16] measure Auth Boost's impact. [26:18] With Auth Boost, you can drive more [26:20] revenue and lower your costs. [26:23] But, now I want to give you a little [26:24] glimpse of something we're actively [26:26] developing. [26:27] We are taking everything we know about [26:29] growing businesses faster on Stripe, and [26:32] making it all available through an [26:34] agentic expert called Stripe Console. [26:38] Felix, can you give us a sneak peek? [26:40] >> This is the Stripe Console. So, I can [26:42] ask it something like, [26:45] "How many payments were refunded [26:47] yesterday?" [26:49] And it will lovingly tell me [26:52] 17. [26:54] Neat, helpful, but Console's not a [26:57] refunds lookup tool. It's a fully [26:58] fledged AI agent for your Stripe [27:00] account, capable of answering questions, [27:02] investigating, and taking action on your [27:04] Stripe data. So, let's push it a bit [27:07] harder, and I want to know [27:10] how is my business doing month-to-date? [27:15] Now, Console is running a bunch of tasks [27:18] in parallel, and you can see that down [27:19] here, and it's pulling all my data on [27:21] Stripe to generate charts for me. [27:24] And so, whoop, here's transactions, [27:27] here's revenue, and here's conversion [27:29] rate by country. That's great, but it's [27:31] not just pulling data, it's also done [27:34] some thinking. So, if you look here, [27:38] it's noticed that conversion rates in [27:40] Brazil are lower than other regions, and [27:43] it's even encouraged me to enable Pix, [27:45] which I think that makes sense. [27:48] But, most proactively, it's even offered [27:50] to set up an AB test for me, [27:52] which sounds great to me, so let's just [27:54] tell it, let's do it. [27:57] Exclamation mark. [27:59] For my excitement. Now, [28:03] Console's going to figure out the tool [28:05] calls it needs to make, and it's going [28:07] to ask me for confirmation. It's [28:08] checking me to proceed because this is [28:10] going to be a user-facing change. And [28:13] so, let's do it live. [28:15] It's loading up my payment [28:16] configuration. [28:19] It's adding Pix to my checkout in [28:21] Brazil, and [28:23] there it is. [28:26] >> [applause] [28:31] >> Our AB test is live and created [28:34] right from the Stripe console. [28:35] >> Amazing. Thanks, Felix. [28:39] Now, yes. [28:41] >> [applause] [28:43] >> Some agents are optimized to be [28:44] friendly, and some are optimized to be [28:46] creative. Console is a general purpose [28:48] agent optimized to look at your Stripe [28:51] account and ask, "How can I grow [28:53] faster?" [28:54] So, whether you're selling to humans or [28:56] agents, expanding globally with [28:58] localized experiences, or optimizing [29:01] your payments performance, Stripe helps [29:03] your business thrive. [29:05] Now, to tell you more about the future [29:06] of commerce, please welcome Ginger [29:08] Baker, Vice President of Product for [29:10] Meta, and Will back to the stage. [29:24] >> So, Ginger, thank you for being here. [29:26] That's right question. [29:27] >> thanks for having us. It's like the [29:29] full fintech family reunion again out [29:31] here. [29:31] >> And so, you know, when we all think [29:33] about Meta, first thing comes to mind, [29:34] you know, world's dominant social [29:36] platform. Don't always think about [29:38] commerce right away, but here you are at [29:39] Stripe Sessions, so why are you here? [29:42] >> Yeah. Meta is many things to many [29:44] people, but at our core, we build [29:46] technology that helps people to connect. [29:48] Uh we have 3.5 billion people that use [29:51] one of our apps each day, and when they [29:52] use those apps, they're looking to [29:54] connect with others, whether or not it's [29:56] through text they post on Threads, or a [29:58] video they may post on Facebook, or [30:01] content they create on on Instagram. [30:03] It's all about using different forms of [30:05] digital content to connect to one [30:06] another. And what is money other than [30:08] another form of digital content? And so, [30:10] if we aren't able to make money move as [30:12] seamlessly as other forms of digital [30:14] content through our surfaces, it feels [30:16] like we're missing out on a real [30:17] opportunity to help people connect with [30:19] one another. [30:20] >> And so, [30:21] social commerce, you know, I think when [30:24] we hear that, you know, scrolling and, [30:26] you know, maybe tapping on an ad. But [30:28] now we're in this agentic era. [30:30] How's that going to change? [30:32] >> Well, let's talk about what it looks [30:33] like today. So, for folks in the [30:35] audience, how many people used Instagram [30:37] this week? No judgment. No judgment, [30:38] Instagram users. Okay. So, you know the [30:41] experience then for you see something [30:44] that your friend has posted. Maybe it's [30:45] a pair of gold snake earrings and you [30:48] really want that item. But what are you [30:49] going to do today? Today you have to [30:51] reverse image search, try to figure out [30:54] what brand the item came from, find the [30:56] website, create an account, you know, [30:59] enter your credit card information to [31:00] check out. And I just think it's really [31:03] exciting to think about a world in which [31:04] you don't have to do that. Where you can [31:06] go through your friend's photo and even [31:08] just on clicking on the photo, you can [31:10] get information about the brand that [31:12] made the item, [31:14] information about the price or the size [31:15] or whatever you need. And without ever [31:17] leaving the experience that you're in, [31:20] you check out. It's related to what [31:22] Kevin was just talking about that we [31:23] launched together. [31:24] >> I didn't know people were judging me for [31:26] using Instagram. [31:26] >> Yeah, no. I don't think it's judging. Um [31:29] clearly I shop on Instagram pretty [31:30] quickly. [31:31] >> It's a [31:31] more in situ experiences. So, you're not [31:33] going to be bouncing around. You're [31:34] thinking right there on the page. [31:36] >> That's right. I mean, from discovery to [31:38] purchase, it all takes place in the same [31:40] experience. And that's what the power of [31:41] AI I think really is is all of those [31:44] steps that you used to need to go [31:45] through to satisfy your desire of the [31:48] thing that you were inspired by can now [31:51] be done everything from inspiration to [31:52] purchase happens in a single seamless [31:54] experience. [31:56] >> You guys have of course a uniquely [31:58] global platform. And as we've discussed, [32:00] you're trying to do commerce worldwide. [32:03] So, what are the challenges and how are [32:04] you, you know, getting around them? [32:06] >> Yeah, we operate in just about every [32:09] country in the world. Um and because we [32:11] operate in all those countries, as I [32:13] mentioned, money also needs to move [32:15] seamlessly and safely in all of those [32:17] countries. And so, given the vast [32:19] breadth that we have not only in our [32:21] surfaces that in our products and our [32:23] wearables and everything that we have, [32:25] we must rely on partners to build out [32:28] this infrastructure with us to ensure [32:29] that uh we connect businesses and people [32:31] in the way that make matters most to [32:33] them with money. Um and so, as a result, [32:35] we rely very heavily on partners. I'll [32:36] use the example of creators cuz they're [32:39] a pretty diverse group. Um so creators [32:41] today you can have a small solo artisan [32:45] in Indonesia who manages their business [32:47] almost solely on WhatsApp. Or you can [32:49] have a brand celebrity brand in the [32:52] United States who manages a [32:53] multi-million dollar business across [32:55] many of our products and services. And [32:57] all of them need something different. [32:59] They're going to get paid by us in a [33:01] different way. They're going to receive [33:03] payment from their um the people that [33:05] they interact with in a different way. [33:06] And so we rely very heavily on our [33:08] partners like Stripe. Um one example [33:10] here is for creators specifically, yeah, [33:13] we're working with Stripe as a partner [33:14] to enable creators to get paid out in [33:16] stable coins in select markets now. And [33:19] it's just an example of trying to meet [33:20] people where they are and give them the [33:23] payment methods and tools that they [33:24] want. And for us it's super important [33:26] because Stripe can handle the licensing, [33:29] the on-ramps, the custody, the [33:30] off-ramps, all of the things that we [33:31] don't want to do and we can focus on the [33:33] experience and the relationship that's [33:35] so important for us with creators. [33:37] >> So how many markets will stable coins [33:38] get you to? [33:39] >> Um well, I think it You tell You tell [33:42] me, Will. I think that's the question [33:43] for you. [33:44] >> a bit later, [laughter] but at least 160 [33:45] by the end of the year. [33:46] >> Okay, fantastic. Yeah. [33:48] >> Um all right, one last one for you. So [33:50] AI keeps surprising us. Timelines keep [33:52] compressing. What's one thing that we [33:54] may all think is sort of radical [33:56] sounding today that in two years might [33:58] seem obvious? [33:59] >> Mhm. [34:00] Oh, I think that payments will pivot [34:03] from being a moment to being a policy. [34:06] >> Mhm. [34:07] >> And I think it feels radical today [34:10] because that policy will be something [34:12] that you set with your agent most [34:14] likely. And that policy will be scripted [34:17] with guardrails that say, [34:19] "Please spend, you know, only up to $50 [34:21] on groceries this week. [34:23] Always preference this card. Never do an [34:26] authorization for a payment over $500." [34:29] And those policy and guardrails are [34:30] things that someone other than yourself [34:32] or some thing other than yourself is [34:34] going to be operating by. And so I I [34:37] think it feels radical today to suggest [34:39] that we would just completely uh, you [34:41] know, give something else permission to [34:43] manage our our financial transactions, [34:45] but I think that that's going to be the [34:47] reality we're in. [34:47] >> Yeah, moment to policy. I like that. [34:49] >> Yeah. [34:49] >> Well, thank you again. We love our [34:50] partnership. [34:51] >> Yes. [34:52] >> And uh, next up, we've got Emily to tell [34:55] us about Stripe Radar. [34:59] >> [music] [35:08] >> AI is creating entirely new terrain for [35:11] fraud. [35:12] Bad actors are no longer just stealing [35:14] money or credentials. [35:16] Increasingly, they're also stealing [35:18] tokens. [35:20] In AI, every prompt, every API call, [35:23] every token has a real cost. And as more [35:26] of you embed AI into your products, [35:28] token theft has become one of the most [35:30] urgent challenges you face. [35:33] Radar is our fraud prevention product. [35:36] It started with protecting you from [35:38] fraud on card payments. [35:40] Today, I'll show you how we transformed [35:41] Radar to protect far more of your [35:43] business. Your tokens, your funnel, [35:46] every transaction, and with more control [35:49] than ever. [35:51] The most ambitious AI companies trust [35:53] Radar. OpenAI, Anthropic, ElevenLabs, [35:57] Cursor. [35:58] So do some of the world's most [35:59] established businesses, like Pepsi and [36:01] Hertz. [36:03] So, let me walk you through three of the [36:04] fastest growing fraud patterns we see [36:07] across the Stripe network. And by the [36:09] way, also in your tweets. [36:12] The first is multi-account abuse. And [36:15] this happens well before any payment [36:17] credentials are entered. So, bad actor [36:19] signs up again and again and again, [36:22] stealing new user credits every time. [36:25] Across Stripe, one in six sign-ups at AI [36:28] companies are involved in multi-account [36:30] abuse. [36:31] And they've been nearly impossible to [36:33] stop [36:34] until now. [36:36] Radar evaluates each new account in real [36:39] time, so you can block before a single [36:41] token is burned. [36:43] Our solution is drawing on signals from [36:44] prior abuse across the entire Stripe [36:47] network. So, device fingerprints and IP [36:49] addresses and email domains and a whole [36:51] lot more. [36:52] For one AI company, 80% of the bad [36:55] actors Radar caught had sailed right [36:57] through their existing provider. [37:00] Another fraud pattern that's escalating [37:02] quickly is free trial abuse. [37:05] In just the last 6 months, free trial [37:07] abuse has more than doubled. And the [37:09] last month, [37:11] particularly ugly. [37:13] For AI businesses, every abusive trial [37:15] burns real inference costs. One company [37:18] we work with was bleeding over $500 on [37:21] tokens to acquire each paying customer. [37:24] How, you might ask? They were [37:26] bankrolling 25 trials just to get one [37:30] conversion. [37:31] Now, not every failed trial is abusive, [37:34] but enough of them are that the [37:36] economics clearly break. [37:38] And you're left with a painful [37:39] trade-off. Either you keep the free [37:40] trial and let abuse eat your margin, [37:44] or you kill it and growth slows. [37:47] Well, Radar now removes that trade-off. [37:49] You can turn on free trial abuse [37:51] prevention in the Stripe dashboard with [37:53] one click. [37:54] Radar predicts which trials are abusive, [37:56] letting good users through while keeping [37:58] bad actors away from your tokens. [38:01] You just saw free trial abuse [38:02] accelerating, right? Now, watch what [38:04] happens when we isolate the businesses [38:07] that adopted our solution. [38:09] A similar exponential curve, and then at [38:11] adoption, [38:13] a dramatic drop in abuse. [38:18] >> [applause] [38:22] >> Across eight high-growth AI businesses, [38:24] Radar blocked more than 3.3 million [38:26] risky free trial attempts in the last [38:28] month alone. [38:30] Okay, and then there's pay-as-you-go [38:32] abuse. So, this is a customer coming in [38:34] consuming thousands of dollars of tokens [38:36] over the course of the month, getting [38:37] billed at the end, and then never [38:39] paying. [38:41] It's like a Michelin starred [38:42] all-you-can-eat dine and dash, [38:44] but for tokens. [38:46] Radar now predicts non-payment as that [38:49] usage accumulates, which means well [38:51] before the bill comes due, you can [38:52] require a top-up, or cut off service, or [38:54] take whatever action fits your risk [38:56] tolerance. [38:58] So, stopping token theft solves a [39:01] growing part of the problem, but most [39:03] fraud today still happens at the [39:05] transaction level. [39:06] So, we've also expanded Radar's coverage [39:09] of your transactions. [39:11] Last year, we took Radar beyond cards to [39:14] ACH and SEPA. [39:16] Today, we're going further, a lot [39:19] further, expanding Radar to block [39:21] high-risk transactions across more [39:23] payment methods. [39:24] Now, [39:25] I'm not going to read them all out [39:26] because [39:28] well, it's all of them. [39:32] >> [applause] [39:35] >> Yes, Radar now protects all of your [39:38] Stripe payment volume. And by the way, [39:40] this extends beyond Stripe, too. Radar's [39:42] same risk scores are available via API [39:45] for your off-Stripe volume. So, one [39:48] unified fraud layer for your entire [39:50] business. [39:56] You guys are my people. I love it. Okay, [39:58] broader coverage also means the full [40:00] transaction life cycle. So, from the [40:01] moment of purchase all the way through [40:04] to disputes. [40:05] For the disputes that are worth [40:06] fighting, Smart Disputes compiles and [40:09] submits the evidence for you, and it [40:10] recovers 18% more revenue. [40:14] But now, it also tells you exactly what [40:17] would strengthen your case. [40:19] Upload that additional evidence, Smart [40:21] Disputes re-compiles the package, and [40:23] you win three times more often. [40:27] Finally, we're making Radar more [40:29] customizable. [40:31] You can set your risk tolerance and let [40:33] Radar just adapt to you, balancing fraud [40:36] protection with your growth objectives. [40:39] And for those of you who want even more [40:41] control over how you fight fraud, I'm [40:43] excited to share that we're starting to [40:45] offer custom fraud models for your most [40:47] bespoke needs. [40:49] You pass us signals unique to your [40:51] business, product catalog, loyalty data, [40:53] whatever's relevant, and we marry them [40:56] with our network data to deploy a model [40:58] fitted just to you. [41:01] For every one of our early adopters, [41:03] custom models are detecting at least 15% [41:06] more fraud with no increase in false [41:08] positives. [41:10] So, with Radar, you can now defend [41:13] against new types of fraud, including [41:15] token theft. [41:16] Protect all of your payments on Stripe. [41:19] Protect your payments off Stripe with [41:21] that same rigor. [41:23] Win more disputes. [41:25] And use custom models to meet your [41:26] bespoke needs. [41:28] Now, to tell you all about how we're [41:30] upgrading our revenue suite, here's [41:32] Tanya. [41:44] >> For most of the last 20 years, revenue [41:46] models have pretty much remained the [41:48] same. [41:49] But in the last 2 years, that's [41:51] completely changed. [41:53] The market has moved to pricing based on [41:55] usage, on outcomes, on hybrid models, [41:58] and increasingly on tokens. [42:00] This means that your ability to compete [42:03] depends on a revenue system that can [42:05] keep up. [42:06] That's why companies like OpenAI, [42:08] Lovable, Notion, Zoom, and many more run [42:11] on the Stripe revenue suite, which [42:13] includes billing, tax, invoicing, and [42:15] our data products. [42:17] Today, I want to show you three ways [42:18] that we're upgrading it. [42:20] First, we're enabling entirely new [42:23] AI-native business models. Second, we're [42:26] making Stripe more programmable than [42:28] ever. [42:29] And finally, we're making your data more [42:31] accessible in real time. [42:34] So, let's start with business models. [42:36] If you're an AI-native business, or [42:38] really any business leaning into AI, you [42:41] face the same challenge. You need a [42:43] business model that can handle the [42:45] agent-fueled explosion in token maxing. [42:48] Now, the first step is tracking token [42:50] consumption in real time. [42:52] But with agents, this is actually a huge [42:55] distributed systems problem where every [42:57] millisecond matters. [42:59] This is why we acquired Metronome, a [43:01] billing engine that runs as fast as [43:04] agents do. [43:06] With Metronome, we're building real-time [43:08] metering, that's counting every token [43:10] your customers use. [43:12] Real-time rating, that's applying your [43:14] pricing to every single one of those [43:16] tokens as they're consumed. [43:18] And real-time alerting, so you can cap [43:20] customers and agents before they [43:22] outspend their token budget. [43:24] Now, to see it in action, let's welcome [43:26] Will back to the stage. [43:28] >> Thanks, Tanya. [43:30] >> [applause] [43:30] >> Okay, so let's get back to that API [43:32] reviewer app that I built. And you may [43:34] recall I was charging a $2 flat fee per [43:36] review. [43:38] But the problem with that is plenty of [43:39] reviews cost us more than $2 worth of [43:41] compute. So, we're going to change the [43:43] business model. Here's my new pricing [43:45] page. [43:46] Now, it's going to be $3 per 1 million [43:48] tokens. And customers are going to [43:50] prepay for a block of tokens, and when [43:52] they run out, Metronome will stop their [43:54] agents from completing the reviews. [43:57] So, let's say Tanya here is our [43:58] customer, and this time you've asked us [44:00] to review a whole block of Privy APIs. [44:03] You're a real crypto wallets enthusiast. [44:06] And we'll give each agent a $3 token [44:09] allowance, we'll fire them off. [44:12] I've got this dashboard here also so we [44:13] can see a live visualization of them [44:16] working. So, here we go. [44:21] And off they go. [44:23] You see they're each reviewing their own [44:24] API [44:25] and some are finishing like this one [44:27] turning green, green. [44:30] But some are not thanks to Metronome's [44:32] real-time rating. And by telling us [44:34] precisely when an agent is out of [44:36] tokens, Metronome has saved us from [44:38] extending credit to Tanya and burning [44:41] even one extra unpaid token. [44:44] But this is also a problem. [44:45] >> Cuz I might be a great trustworthy [44:47] customer getting incomplete reviews as I [44:49] run out of tokens. [44:51] >> So, I could top you up with more. [44:52] >> But I might also be a fraudster and as [44:55] we heard from Emily earlier, token theft [44:57] is a huge problem. [44:59] >> I've known you a long time and you're a [45:00] bit sketchy. So, I don't want a trail of [45:03] burned tokens and unpaid bills. [45:05] So, how do we solve this? [45:07] What what if I could get paid [45:08] irreversibly in real time for every [45:11] token burned? What if you could stream [45:13] payments the same rate that you can burn [45:15] tokens? [45:16] Well, stepping back, payment systems [45:18] today just simply can't do this. You [45:20] can't do it with cards, you can't do it [45:21] with ACH or UPI or pics. [45:24] And beyond payments, you actually need a [45:25] different kind of money. You can't do [45:28] with fiat. You need stablecoins to be [45:30] able to charge minute sums by the [45:32] millisecond. [45:33] So, we need to process and settle [45:35] payments at machine speed and that's [45:37] exactly why we partnered with Paradigm [45:39] to build Tempo. [45:41] It's a blockchain designed specifically [45:43] for payments. [45:45] So, Tanya, I'm going to change up my [45:46] pricing one more time. [45:49] Here's my new pricing page and I'm going [45:50] to charge you by the token. [45:53] And this time I'm going to collect [45:54] payments in real time for each token and [45:57] I'm going to do it by streaming [45:58] streaming stablecoins on Tempo. [46:02] So, let's reset the dashboard and we [46:04] will fire off these agents again. [46:07] And if this works, you'll see the [46:09] agents' token use ticking up on the [46:11] left. On the right, you'll see real live [46:14] stablecoin micropayments streaming in. [46:17] As far as we know, no one's actually [46:18] done this before. So, we're going to [46:20] launch the world's first ever tokens [46:23] paid as burned business model right here [46:25] on stage. We're going to do it by [46:27] streaming stablecoins on Tempo. [46:30] >> So, if you pull this off, we can say [46:32] we've launched America's next top model? [46:35] >> The world's top new business model. [46:39] Okay. So, I'm going to set these agents [46:41] off. We're going to say just a little [46:42] prayer to the Moscone demo gods. And [46:45] here we go. [46:48] So, you can see the agents burning [46:50] tokens and the payments streaming [46:52] alongside. Now, no agents are getting [46:54] throttled and there's no risk to me. [46:56] Now, let's head over to this is fun, the [46:59] Tempo block explorer. So, this is the [47:01] actual Tempo blockchain. You can see [47:03] here the payment stream that was opened. [47:06] Let's see if we can zoom in here. [47:09] There we go. $3.30 that total. If I [47:12] scroll down, you can see these thousands [47:14] and thousands and thousands of sub cent. [47:17] You know, this is 3/1000 of a cent [47:21] that make up that total. So, let's head [47:23] back. [47:25] All the agents are done and that's it. [47:27] We just launched a world's first [47:29] streaming payments business right here [47:31] on the stage at Moscone. [47:34] >> [applause] [47:40] >> And now, this is only possible because [47:42] of Metronome, Tempo, and stablecoins [47:44] tracking and paying for these tokens in [47:46] real time. [47:48] So, this is a really visceral example of [47:50] how Stripe is building the economic [47:51] infrastructure for AI. [47:54] Now, to hear more about Tempo and [47:57] agentic payments, let's welcome Georgios [47:59] Constant the CTO of Tempo and Paradigm [48:02] to the stage. [48:10] Georgios, good to see you. [48:11] >> Good to see you. [48:12] >> Thanks for being here. [48:13] >> It's wonderful to see our technology [48:14] shown and you launching new business [48:16] models on stage. [48:17] >> Some great technology. [48:19] So, you were a long time leader and [48:21] architect in Ethereum, which is, you [48:23] know, one of the most important projects [48:24] in crypto, but notably not one that was [48:26] ever really fintech or payments focused. [48:29] So, what made you want to bring all that [48:31] crypto expertise over to this problem [48:33] domain? [48:33] >> Yeah, so well, I got started in the [48:35] crypto space almost 10 years ago, um, [48:38] when, you know, I'm born and raised in [48:40] Greece. We went through financial [48:41] crisis. The problem was deeply buried in [48:43] our bodies and when we saw what crypto [48:46] could do, it could enable global 24/7 [48:49] just open financial access, that was [48:51] really motivating. And that was great, [48:54] but it was still very early. We were [48:56] still didn't have the foundational [48:59] pieces of infrastructure to make [49:00] anything really work. [49:02] And if you fast forward to today, [49:04] there's a tsunami coming, which is the [49:07] agents that many people have been using [49:09] this room, if not everyone, and we see [49:11] that this tsunami is just going to [49:13] trample over existing infrastructure. [49:14] We're seeing it already with the Git [49:16] world. We're seeing this with databases. [49:17] We're about to see this with payments as [49:19] well. So, we got together and we thought [49:22] we have this great technology. Let's [49:23] point it to an important problem with [49:25] great people like Stripe, and that was [49:27] the the birth of the Tempo project, the [49:29] blockchain for payments, stablecoins, [49:31] and agentic finance. [49:33] >> So, beyond my amazing app, [49:35] who else is using Tempo and what are [49:37] they using it for? [49:38] >> So, stablecoins are a great solution for [49:41] anyone doing B2B payments, remittances, [49:44] payroll, anything that requires high [49:47] velocity of money, low fees, global [49:49] programmable. So, the three people I [49:52] would the three companies I would [49:53] mention is one, DoorDash is using Tempo [49:56] to scale their Dasher payouts globally [49:59] at low cost, which has a lot of [50:00] infrastructure cost involved right now. [50:03] Klarna has issued Klarna USD, which is [50:05] their own stablecoin that they're [50:06] deploying across their fintech [50:08] infrastructure. [50:09] And thirdly, Visa, uh who we work with [50:12] alongside you guys on the on machine [50:14] payments on the machine payments [50:15] protocol, which they're using to make it [50:18] the most frictionless way to talk to [50:20] agents. And they're also running a [50:22] validator alongside Stripe and Standard [50:24] Chartered and a few others. [50:26] >> Heard of Visa. [50:27] >> Yeah, I've heard of them. Payments [50:27] company. [50:29] >> Um and of course Stripe is using Tempo [50:31] for our financial accounts and a lot of [50:32] our global [50:33] >> Absolutely. [50:34] >> So, you guys shipped fast uh [50:37] deposit addresses earlier this week. [50:38] That was great. Uh if you could just [50:40] name two exciting things coming, what [50:41] would they be? [50:42] >> Yeah. Well, A, I think it's insane for [50:44] anyone to be using stablecoins in clear [50:47] text without privacy. So, the biggest [50:48] problem that we're tackling up next in [50:50] production is going to be privacy, [50:52] private, global, 24/7 trans payments. [50:55] And the other part is further preparing [50:57] for the tsunami. So, further scale, more [51:00] optimization of our software, just [51:02] making sure that the agents can crank. [51:04] >> All getting ready for the tsunami. Thank [51:06] you, Georgios. [51:07] >> And you. [51:08] >> Tanya, back to you. [51:10] >> [applause] [51:13] >> Thanks, Will and Georgios. [51:15] So, now for the second upgrade to our [51:17] revenue suite, making Stripe more [51:19] programmable. [51:20] You heard from Patrick earlier, this has [51:22] always been core to Stripe's identity. [51:25] And there's no place that this is more [51:26] important than in your revenue stack, [51:28] the operating system of your business. [51:31] As a design principle, we strive to make [51:33] all of Stripe's products extensible. But [51:36] we also have products specifically built [51:38] to extend Stripe. [51:40] Like custom apps, which let you build [51:42] bespoke interfaces directly in the [51:44] Stripe dashboard. [51:46] And workflows, a visual builder that [51:48] lets you program actions directly in [51:50] Stripe. [51:51] But here's the thing, no matter how many [51:53] features we build, there will always be [51:56] logic unique to your business. [51:58] Logistics companies run on shipments, [52:00] insurance companies run on policies and [52:03] claims, and none of those things exist [52:05] natively in Stripe. So, you end up [52:08] maintaining separate systems, stitching [52:10] them together with metadata, clicking in [52:11] and out of the Stripe dashboard, and [52:13] writing custom code just to run routine [52:16] tasks. [52:17] Well, that ends today. [52:19] Today, we're introducing custom objects, [52:22] a new primitive that lets you bring your [52:24] business data and your business logic [52:26] directly into Stripe. [52:28] And when you compose custom objects with [52:31] custom apps and workflows, you get [52:33] something really powerful, a [52:35] made-to-order version of Stripe. [52:38] So, let's take a look at a couple [52:39] examples. [52:40] Here's a custom object and method a [52:42] logistics and fulfillment company could [52:45] define in code. [52:46] And here you can see it reflected in a [52:48] custom app they could build inside the [52:51] Stripe dashboard. Now, the fields you [52:53] see here are tracking numbers, [52:55] destinations, carriers, they aren't [52:57] native to Stripe. They're specific to [52:59] how this business actually operates. [53:03] Now, let's take a look at a workflow an [53:05] insurance company could create uh within [53:07] Stripe to handle complex policy [53:09] cancellations. [53:11] It includes everything from triggering [53:13] different approval chains based on the [53:15] size of refund required to the email [53:17] notifying a customer that their [53:18] cancellation is approved. [53:21] Think about what this unlocks for your [53:23] support, finance, and operations teams [53:25] who already spend a bunch of time in [53:27] Stripe. [53:28] They no longer have to use multiple [53:30] tools to do routine work because you can [53:32] tailor Stripe to the shape of your [53:34] business. [53:36] And our commitment to extensibility [53:38] doesn't stop there. We just talked about [53:40] how you can build a custom app. [53:42] You can also build public apps for the [53:44] broader Stripe ecosystem. [53:46] And today we're expanding apps with [53:48] full-page experiences. [53:50] App builders, the Stripe dashboard is [53:52] now your playground. [53:55] Now let's go to the third upgrade, [53:57] making your data accessible in real [53:59] time. [54:01] Some of the most valuable data about [54:03] your business already lives in Stripe. [54:05] And we've been on a journey with many of [54:07] you to make that data easier to use [54:09] wherever your business runs. [54:11] You told us you wanted to query your [54:12] Stripe data to generate reports, so we [54:15] built Stripe Sigma. [54:17] You told us you wanted to sync your data [54:19] with your own warehouse, so we built [54:21] data pipeline. [54:23] And more recently, you told us you [54:25] wanted real-time access to your Stripe [54:28] data in a nice, structured database. [54:31] So today, we're launching Stripe [54:34] database. [54:35] It's a [54:38] >> [applause] [54:40] >> It's a Postgres database you and your [54:42] agents can use to understand your [54:44] business as it is right now, not as it [54:46] was a few minutes ago. [54:48] And coming soon, we're adding write [54:50] access, custom objects, and third-party [54:52] storage. [54:54] Stripe database is available in private [54:55] preview, and you can go to our docs to [54:57] request access today. [54:59] We're shipping a lot of updates to our [55:01] revenue suite, plenty more than we have [55:03] time to cover right now, but I will [55:05] mention one more that you've been asking [55:06] for. [55:08] Stripe billing support for contracts. [55:10] You'll soon be able to model everything [55:12] from custom pricing overrides to prepaid [55:15] credit commitments to amendments all [55:17] natively in Stripe. [55:19] So, to recap what we've covered, [55:22] Metronome and Tempo for streaming [55:24] payments, [55:25] custom objects, workflows, and apps to [55:27] customize Stripe, [55:29] full-page public apps in the dashboard, [55:33] Stripe database to put all your data to [55:35] work in real time, [55:37] And billing support for sold contracts. [55:40] So, that's it for revenue. Now, to tell [55:42] us about managing your money on Stripe, [55:44] here's Nithya. [55:49] >> [music] [55:56] >> A new breed of business has emerged. [55:59] Startups that are hyper-scaling, super [56:01] lean, and instantly global. Many of you [56:04] are here today. You're growing faster [56:06] than any generation of companies before [56:07] you. But, the one thing that can't seem [56:10] to keep up is how you manage your money. [56:13] Your finances are still scattered across [56:16] half a dozen apps and spreadsheets. [56:18] It's a hassle that just doesn't scale. [56:23] That's why last summer we introduced [56:25] financial accounts. So, you can store [56:27] money in the same place where you [56:29] already accept payments. [56:32] But, if you're a rapidly growing global [56:34] startup, it's not enough to just have [56:37] your money in one place. You should be [56:39] able to move the funds instantly [56:41] whenever and wherever you need it. [56:44] Your idle cash should be earning while [56:46] you sleep. [56:48] And you should be able to automate most [56:50] workflows, freeing up precious time to [56:52] focus on the rest of your business. [56:57] Today, we're solving all all of that. [56:59] I'm super excited to announce the new [57:02] Stripe Treasury. [57:06] >> [applause] [57:10] >> It's a new financial operating system [57:12] that gives startups all the capabilities [57:14] they need for storing, spending, and [57:16] managing money. Powered by both fiat and [57:19] stablecoin rails. [57:21] Let's take a look. [57:23] You get multi-currency financial [57:24] accounts, which enable you to hold [57:26] dollars, pounds, euros, and stablecoins [57:29] side by side. [57:30] You can transfer between balances and [57:33] convert currencies instantly whenever [57:35] you need yes even on weekends and bank [57:37] holidays. Now I don't know what you do [57:38] on weekends I only convert currencies. [57:42] Your accounts come with local routing [57:44] numbers for getting paid or receiving [57:46] money. [57:47] And in the US your balance will now also [57:50] be FDIC insured. [57:53] In addition to that you can create [57:54] stripe cards powered by MasterCard to [57:57] spend directly against your balances. [58:00] You can set limits and manage controls [58:02] right from the dashboard. [58:04] And starting today businesses using [58:06] treasury will earn 2% cash back on all [58:09] card purchases. There are no special [58:11] categories no blackout windows. [58:15] >> [applause] [58:18] >> We're also adding rewards on balances. [58:21] Businesses will automatically earn [58:23] credits when they hold money in treasury [58:25] and they can use these credits to pay [58:27] down their stripe fees. [58:30] So with all of that if you're thinking [58:32] this sounds a lot like a business [58:33] banking account you're right. In fact [58:36] entrepreneurs in more than 50 countries [58:38] are already using stripe treasury as [58:40] their primary business account. [58:44] We're also making treasury the fastest [58:47] easiest and the cheapest way to move [58:48] money anywhere in the world. [58:51] You can pay or send money to anyone [58:53] anywhere using just an email address. [58:56] This year we've more than tripled the [58:58] number of recipient countries from 50 [59:01] and we'll already mentioned this to 160. [59:05] And we're not solving for just the [59:07] outbound leg. [59:08] As you heard from Ginger meta wanted to [59:10] use stable coins to send fast and [59:12] reliable payments to creators in the [59:14] Philippines and Colombia. [59:16] But they couldn't send stable coins [59:17] without having crypto wallets on the [59:19] other end to receive them. [59:21] That's where link comes in. [59:24] Link can now store stable coins. [59:28] And this is a big deal. This means [59:30] businesses can simply send stable coin [59:32] payouts directly to the 250 million [59:35] consumers on Link [59:36] without having to deal with the [59:38] complexities of crypto. [59:42] Now, if you conceive of Link as a big [59:45] network of consumers, then Stripe itself [59:48] is a big network of businesses. [59:51] From the world's largest brands to all [59:52] the fastest-growing startups, you are [59:55] building on Stripe. And along the way, [59:57] you're buying from each other. [59:59] A lot, like 4.8 million times every [60:02] single day. This is how often one Stripe [60:05] business is paying another. [60:08] So, I'm very excited to share that today [60:10] we're launching instant transfers. [60:15] >> [applause] [60:18] >> Thank you. [60:19] Starting today, businesses on the Stripe [60:21] network can send money between treasury [60:24] accounts at zero cost. [60:26] And just to make it super clear how [60:28] any business on Stripe can now send [60:31] money to any other business on Stripe [60:33] instantly and for free. Instant and [60:36] free, two words you don't usually hear [60:37] together in finance. [60:39] >> [laughter] [60:41] >> Now, if you're a high-growth startup or [60:43] if you aspire to be, [60:44] you don't want to just manage your [60:45] money, you want to program it. So, we're [60:47] making all of treasury, all the treasury [60:49] APIs available via the MCP. Felix, can [60:52] you show us? [60:54] >> Of course. Now, whether you're using [60:56] your agents off the shelf or you've [60:58] built your own interfaces to agents, [61:00] your team can connect it to Stripe with [61:02] MCP. And to illustrate it, we say hello [61:04] to ChatGPT. [61:06] Now, [61:08] I've already connected to Stripe's MCP [61:10] server and have granted it access to my [61:12] treasury account. But, we're live here, [61:14] so let's make sure everything is hooked [61:15] up. So, chat, [61:17] show me the money [61:20] in my Stripe balance. [61:26] Now, ooh, here it is. Perfect. That was [61:28] quick. What we're seeing here is a [61:30] Stripe embedded component, and it looks [61:32] just like what I'm used to in the Stripe [61:34] dashboard. But, let's try something a [61:36] bit more complicated. I want to ask my [61:38] agent to help me pay an invoice. So, I [61:42] have this PDF. [61:44] I'll attach it there. [61:46] Pay this invoice. [61:47] Two exclamation marks. [61:50] So, now the agent is checking the [61:52] invoice details. It's going to do thing [61:54] do all the things you expect like the [61:56] check the due date, check available [61:57] payment methods, things like net [61:59] settlements. [62:00] Um uh figure out what I'm paying for, [62:02] whether I've paid for it before. [62:04] >> You didn't use all caps, Felix. That's [62:06] slow. [62:06] >> Hm? [62:07] >> You didn't use all caps. [62:09] >> You know. [62:10] So, it's detected that the recipient has [62:14] a Stripe profile. [62:20] It's thinking really hard today. [62:22] >> [laughter] [62:22] >> Which I appreciate. But, I thought I [62:25] wish it would think faster. Oh, great. [62:27] So, it's detected that the recipient is [62:29] on the Stripe network, so we see the [62:31] Stripe profile, so we know we can send [62:33] them a transfer instantly and for free. [62:35] And beyond that, it's going to even [62:39] check my previous payments. It's [62:40] checking things like whether my [62:42] financial accounts have money. [62:45] And it's confirming whether I really [62:46] want to send $1,300, which it's looking [62:49] out for me. [62:55] Sending the payment now, or so it says. [62:59] Confirming a second time. [63:12] Confirming a third time. [63:14] >> [laughter] [63:15] >> AI safe. [63:24] >> And payment set up successfully. [63:27] >> [applause] [63:33] >> And because it was on the Stripe [63:34] network, that means the recipient can [63:36] start using their funds right away from [63:38] the Treasury account. So, I've got the [63:41] demo app here. [63:42] Uh it's open to my recipient's Treasury [63:45] account. And so, if everything worked, [63:47] I'll just hit refresh. [63:49] And it's landed. [63:51] >> [applause] [63:56] >> You just saw an instant free [63:58] business-to-business transfer on the [64:00] Stripe network run entirely via agent [64:02] and MCP. [64:04] >> Thanks, Felix. [64:06] >> [applause] [64:09] >> All the power of Treasury via MCP means [64:12] that you can manage your money across [64:13] any AI surface where you work. [64:16] So, let's recap. [64:18] With Treasury, you're getting a new [64:19] financial operating system for all your [64:21] money management needs. [64:23] New this year, we've added ways to earn [64:25] free instant transfers on the Stripe [64:27] business network, as you just saw. [64:29] Global payouts to 160 countries, [64:32] stablecoin payouts to Link, and [64:34] agent-ready APIs. [64:38] And because Treasury is for global [64:40] companies, we have an incredibly [64:41] ambitious global road map. [64:44] Today, Treasury is live in 119 [64:46] countries. That includes the US and UK [64:49] and a preview in Europe. [64:51] It's powered by stablecoins in 100 [64:53] countries. [64:54] We'll also expand Treasury to Australia [64:56] and Canada and add stablecoin support [64:59] for another 41 countries this year, [65:01] bringing up bringing the total up to [65:02] 160. [65:05] Now, to tell you about how stablecoins [65:07] are powering Treasury and many more of [65:08] our products, here's Henry. [65:21] >> All right. [65:22] Well, we couldn't talk about money [65:24] management in 2026 without bringing up [65:26] stable coins. You've heard a little [65:27] about it. What we're seeing is a really [65:29] profound shift in their adoption with [65:32] businesses large and small building on [65:33] top of stables the way they adopted the [65:35] cloud a couple decades ago. [65:37] And Stripe's been stepping up its [65:39] investment, too. [65:40] Bridge's orchestration platform enables [65:42] any business to move money globally with [65:44] stable coins. Privy builds better [65:46] wallets so any developer can easily [65:48] integrate crypto into their products. [65:49] That's how I got here. [65:51] And together with Paradigm, we're [65:52] incubating Tempo, a blockchain [65:54] purpose-built for payments. [65:55] Now, we're already using many of these [65:58] primitives in Stripe's platform today [66:00] and the results are showing up. [66:02] Instead of some payments taking T+2, [66:04] they're immediate. [66:05] Instead of integrating banking rails one [66:07] at a time, you use one system that works [66:09] for the entire world. [66:11] Instead of serving customers based on [66:12] the country or currency where they live, [66:14] you can serve them no matter where they [66:15] are. [66:16] And these are programmable rails so you [66:18] can move millions of dollars just as [66:20] easily as you move fractions of a cent. [66:22] This is truly what internet-native money [66:24] can do. [66:26] Now, just like everything else at [66:27] Stripe, we don't want to keep these [66:29] capabilities locked inside our own [66:30] products. Sharing is caring and, you [66:32] know, we care about all of you. And so [66:34] we want you to use the same primitives [66:36] that we use across Bridge, Privy, and [66:37] Stripe to build out the best experiences [66:40] for your users. [66:42] Stable coins have matured beyond [66:43] recognition and today they're used by [66:45] some of the busiest businesses in the [66:47] world. [66:48] But you don't just need a way to store [66:50] assets, you also need to transact [66:52] securely, earn yield, trade, issue [66:54] cards, and manage balances. [66:57] That is a lot to build. Today that [66:59] requires dozens of integrations and [67:01] we're changing that. [67:03] Now, I'm extremely excited to to you [67:05] something we've been building. [67:07] These are digital asset accounts. It's [67:09] everything you need to build financial [67:11] products with crypto and stablecoins all [67:13] in one API. [67:15] So, you get payments across tokens and [67:17] chains, built-in fiat on and off-ramps, [67:20] on-chain effects, integrated yield, and [67:22] flexible custody. And again, this is all [67:24] one API, so you can just get rolling. [67:27] I can see from the lack of applause that [67:29] many of you have not yet touched [67:31] stablecoin rails, and we will change [67:32] this because it's become so much easier. [67:36] >> [applause] [67:40] >> We'll get there. [67:42] We're rolling out these accounts to [67:43] developers now, and you've already seen [67:45] them in action powering Stripe Treasury [67:47] and Link. And the same tech serves some [67:50] of the biggest names in fintech today. [67:52] Rams is using our stack to accelerate [67:54] their international expansion. You'll be [67:56] able to hear about it later. [67:57] Deel is leveraging it to pay global [67:59] workers instantly. And DoorDash works [68:02] with us to pay restaurant owners more [68:03] quickly with stables. [68:05] It's really incredible to see how far [68:07] the space has come looking at these [68:08] names here. [68:10] But I'm a startup nerd at heart. And so, [68:13] equally exciting to me is seeing our [68:15] APIs power a new generation of startups. [68:18] I'm going to give you three more [68:19] companies that we work with. These are [68:21] the leaner, faster, global startups that [68:23] Nima was talking about earlier, and [68:24] they're quickly becoming some of [68:26] Stripe's biggest customers. [68:28] These are companies like Hyperliquid, [68:29] Chipper Cash, Art Finance, all building [68:32] global businesses on stablecoins. And [68:34] okay, I lied. I'm going to give you a [68:36] fourth name. This is Felix. [68:38] Not to be confused with the very [68:39] talented Felix who's been standing [68:41] behind the demo desk all day. Felix is [68:44] building remittances on WhatsApp, and [68:46] today they power more than 5% of the [68:48] US-Mexico remittances corridor. This is [68:51] the largest remittances corridor in the [68:52] world, and this means that one in every [68:54] $20 going from the US to Mexico is piped [68:58] through a single startup thanks to [69:00] stablecoins. [69:04] >> [applause] [69:05] >> Now, [69:07] importantly, this work is not about [69:09] crypto for crypto's sake. It's about [69:12] enabling you to build fundamentally [69:14] better experiences for your customers, [69:16] your users, and their money. [69:18] Stable coins are here. And this is just [69:20] the start for us. We have a lot to build [69:22] together. With that, back to Nideka to [69:24] talk about what we're shipping for [69:25] platforms. [69:37] >> Here's my favorite part. We've saved the [69:39] best for the last. [69:40] Today, Stripe powers more than 16,000 [69:43] platforms serving over 11 million [69:46] businesses. Some of the biggest [69:47] companies in the world, from Shopify to [69:49] DoorDash to Substack, use Stripe Connect [69:52] as the financial backbone of their [69:54] software. [69:55] Platforms are very core to getting [69:57] financial services distributed to every [69:59] corner of the economy. And so, we care [70:02] deeply about helping you succeed. [70:05] And I know you've given us feedback on [70:07] this, so almost every product we've [70:09] talked about today, we've built it [70:11] natively for platforms. [70:14] >> [applause] [70:18] >> I'll give you just two examples, [70:19] starting with Treasury. [70:21] We're rolling out embedded components so [70:23] you can offer your customers all the [70:25] Treasury capabilities I showed you [70:26] earlier. Companies like Squarespace, [70:29] TechCrunch, and more are already using [70:31] Treasury to become the financial home [70:33] for their customers. [70:35] Last year, we announced Radar for [70:37] platforms, which combats fraudulent [70:39] accounts in addition to fraudulent [70:40] transactions. New this year with Radar [70:43] are AI-generated risk scores, along with [70:46] detailed explanations and recommended [70:48] actions, so you can move faster to shut [70:50] down bad actors. [70:52] We're seeing platforms using Radar like [70:54] StyleSeat, Vimeo, Fair Harbor. [70:56] They're seeing 5.3 times lower losses [70:59] per fraudulent account, [71:01] all powered by the Stripe network. [71:05] But, here's the thing. As your business [71:07] scales, growth gets harder. [71:11] That's why today I'm super excited to [71:13] announce that we're launching the [71:15] Platform Growth Studio. [71:19] >> [applause] [71:22] >> It's a powerful recommendation engine [71:24] that analyzes all the data we see across [71:26] thousands of platforms running on Stripe [71:28] and surfaces the best growth [71:30] opportunities for your business. [71:32] Let's take a look. [71:33] >> Welcome to the Platform Growth Studio. [71:36] Now, right away you can see your top [71:38] metrics, things like your payment [71:39] volume, your revenue, margin, and take [71:42] rate. But, also your top accounts. But, [71:45] what's new this year is recommendations. [71:48] These are [snorts] personalized [71:49] suggestions for how you can grow your [71:51] business. [71:52] The recommendations with the greatest [71:54] impact are always right on top. Now, for [71:56] example, it's suggesting that I can [71:58] improve my ACH pricing. [72:00] But, how does Stripe know that? Let's [72:02] take a look. [72:03] And it looks like I've lost $13,000 [72:06] last month and 75 bips on every single [72:10] ACH transaction. [72:12] Now, here's my take rate over time in [72:14] green, but more interestingly is this [72:17] line in orange, the benchmark, generated [72:19] from all the data across all the [72:21] businesses on Stripe. [72:23] My peers are just doing a much better [72:25] job at monetizing than I am. [72:27] And so, to pick my new pricing, I can [72:29] use the simulator [72:31] and I can play with a few different [72:32] options. I could set it myself or I [72:34] could just choose a Stripe benchmark and [72:36] I hit save and you can see it pop right [72:39] in using my historical data. Now, [72:42] let's go back to recommendations. [72:45] A lot of growth is going to come from [72:47] helping your users adopt new high-margin [72:49] products. And so, here for example, I [72:51] can see that with BNPLs, [72:53] I could stand to make another $12,000 a [72:56] month. Now, only 13% of my accounts are [73:00] using it, so there's just a lot more [73:02] opportunity. So, to encourage adoption, [73:05] I can create a campaign. [73:07] I want to offer promotional pricing for [73:10] these payment methods. [73:12] I will target everyone who's eligible. [73:15] And in just Oh, and I want to price it [73:18] at the great rate of 1.5%. [73:22] And in just a few clicks, the campaign [73:24] is live and the new pricing is effective [73:26] immediately. [73:31] >> [applause] [73:33] >> Thanks, Felix. What you just saw is the [73:36] new platform Growth Studio. You can [73:38] discover opportunities and take action [73:40] on them all from the Stripe dashboard. [73:43] Now, I want to mention two additional [73:45] products, Capital and Issuing. [73:47] Stripe Capital is one of the most [73:49] effective ways to stimulate growth on [73:51] your platform. In fact, we ran a [73:53] two-year study and we found that when [73:55] businesses accepted Capital offers, they [73:57] grew 27 percentage points faster than [74:00] businesses that didn't. [74:03] Today, I'm excited to announce a new [74:05] form of Capital, lines of credit. [74:08] With a line of credit, small businesses [74:10] get a ton of flexibility. They can pull [74:12] funds only as they need to. [74:15] We're expanding Capital today to France [74:17] and Germany with Australia and Canada to [74:20] follow soon. [74:23] Now, Issuing represents another massive [74:26] opportunity for both platforms and [74:28] fintechs. Hundreds of companies like [74:30] Navan, Rippling, Instacart, Affirm, and [74:32] more use Stripe Issuing to distribute [74:34] more than 350 million cards. [74:38] Today, I'm going to share the largest [74:40] set of upgrades that we've made to [74:41] Issuing since we launched it back in [74:43] 2018. [74:45] We've recently added Radar for issuing. [74:48] Using radar, Navan was able to reduce [74:50] fraudulent disputes by 36%. [74:53] We're also adding debit cards to our [74:55] consumer program. [74:57] And we're launching issuing for agents [74:59] so that developers can now easily and [75:01] more importantly safely, you saw how [75:03] safe AI is, gift cards to whole new kind [75:06] of economic actor. [75:08] And finally, the same API developers use [75:11] for fiat issuing can now be used for [75:13] stablecoin card issuing. [75:16] We're seeing incredible traction on [75:18] stablecoin back cards. Fintechs like [75:20] Phantom, Ramp, Jeeves, and Zaps are all [75:23] empowering their users to spend [75:25] stablecoins anywhere Visa is accepted. [75:30] Using stablecoins, we're now able to [75:32] offer Stripe issuing in 100 countries. [75:37] So, to bring it all back, [75:39] Stripe is built for platforms. [75:42] We're shipping updates to Treasury and [75:43] Radar optimized to your needs. We're [75:46] giving you a new growth studio powered [75:47] by Stripe's data. And we're expanding [75:50] the capabilities of both Capital and [75:52] Issuing. [75:53] And as Henry shared, developers [75:54] everywhere can now build great products [75:56] on stablecoins with digital asset [75:59] accounts. [76:00] Now, to hear from one of the world's [76:03] leading platforms, Shopify, please [76:05] welcome Kate Royer, Chief of Staff at [76:07] Shopify, and Will. [76:20] >> Kate, welcome to Sessions. Thanks for [76:21] being here. [76:22] >> Thanks for having me. This is exciting. [76:24] >> So, we just heard from Nitika and sort [76:26] of self-evident, Shopify, one of the [76:27] great platform companies in the world. [76:29] But, I feel like it's still almost [76:31] slightly misunderstood because people [76:33] think a lot about software and SaaS when [76:35] in fact, you guys have quietly become [76:37] one of the world's largest financial [76:38] institutions over the past decade or so. [76:41] So, how does that factor into the [76:43] strategy and what do you have to think [76:44] about day-to-day to make that happen? [76:47] >> Yeah, I mean I think [76:49] Stripe has been a really great partner [76:50] in all of our financial work, or most of [76:53] it anyway. Um [76:54] we really [76:56] we discovered that merchants, especially [76:58] those at scale, you know, they need [77:01] global acquiring, [77:03] they need [77:04] sophisticated fraud tooling, they need [77:07] multi-currency payouts. [77:09] And [77:10] you know, so we had to adapt our [77:12] product. We, you know, I I guess among [77:14] the the many things on the Shopify [77:17] finance suite, we have Shopify Balance, [77:20] which is on top of Stripe. [77:21] >> a couple a couple three years ago, four [77:23] years ago? [77:23] >> Something like that, yes. Um [77:25] and it's only currently in the US, um [77:28] but merchants love it. [77:30] Uh [77:31] you know, it just simplifies their [77:32] business banking needs. Anyway, we're [77:34] hoping to expand and uh Stripe's global [77:37] footprint definitely makes that possible [77:39] faster. [77:40] >> You guys also do Shopify Capital, so [77:42] you've got like a whole suite of [77:44] financial services tools that you're [77:45] bringing to your merchants. [77:46] >> That's right. We're also partnering on [77:50] money movement with digital assets. [77:52] Our teams have been collaborating on [77:56] stable coin storage and [77:59] I think I heard [78:01] we will be coming we will be becoming a [78:04] validator node on the Tempo network. [78:07] >> The Tempo network, yeah, I heard that as [78:08] well. [78:08] >> Yeah. [78:09] >> And so one of the other things that um [78:12] Shopify's known for is driving growth [78:14] for its users. And we made, you know, [78:16] kind of a fun but quick announcement [78:17] earlier that the Shopify catalog will [78:19] now be able to be discovered through [78:22] ACS. So, basically, if you're using the [78:25] Agnostic Commerce Suite, [78:27] agents will be able to sort of surface [78:29] Shopify products directly in their [78:31] experiences. Tell us more about the [78:33] catalog. How's it work? [78:34] >> Yeah, so it's our global product [78:37] catalog. [78:38] But I think what makes it special is [78:41] that we have put in exceptional work to [78:44] organize products with structured data, [78:47] clean attributes, real-time accuracy. [78:50] These things are crucial for AI commerce [78:54] so that agents can find quality results. [78:58] You know, black Vuori leggings shipping [79:01] to Canada [79:03] or [79:04] Glossier Boy Brow in Auburn. [79:06] >> Yeah. [79:07] >> We're now world-class at these kinds of [79:09] queries [79:10] and increasingly complex queries will be [79:13] supported. [79:14] >> And are we talking tens of millions, [79:16] hundreds of millions of products? [79:18] >> No, more than that. [79:20] >> [laughter] [79:21] >> Um the catalog API will give you access [79:25] to millions of shops and billions of [79:28] products. [79:29] >> So Shopify has pretty big. [79:30] >> Yeah, go build amazing new commerce or [79:33] shopping experiences. [79:36] >> So um [79:37] an announcement which we didn't mention [79:38] today, but maybe people saw over the [79:41] past week is that Stripe is excited to [79:43] join the general committee of the [79:45] Universal Commerce Protocol, something [79:47] that you guys pioneered. So thank you [79:49] for having us and thank you to Google as [79:51] well. Uh [79:53] tell us about UCP. Just [79:56] how did it come about and what makes it [79:57] special? [79:58] >> Yeah, when we first started [80:01] UCP with Google, we had one question. [80:05] Can you build a single protocol across [80:08] the full commerce journey [80:11] that supports any platform and [80:14] has all the things that need to work to [80:16] make a transaction trustworthy? [80:18] But a protocol only really matters if [80:20] the industry adopts it. And so that's [80:22] what's most exciting exciting. Just last [80:25] week it was announced that Meta, Amazon, [80:28] Salesforce, and Microsoft are joining [80:30] UCP [80:31] and Stripe, too. [80:32] >> Yeah. [80:33] >> Um so so yeah, I think [80:36] you know [80:37] this for merchants and developers [80:40] eliminates fragmentation. It's a one [80:41] standard to build on, one specification [80:44] to implement, [80:46] one community to support its evolution. [80:49] And [80:50] I mean, I think this sort of marks an [80:53] inflection point. We're going to see an [80:55] even further acceleration in agent [80:56] commerce. [80:57] >> Yeah. So on that note, you guys have [80:59] always been sort of looking around [81:00] corners with commerce experiences. [81:03] What do you foresee in the next, you [81:04] know, couple of years for agent [81:05] commerce? [81:08] >> I mean [81:09] >> Easy question. [81:10] >> to know that far but um [81:13] I'm sure everyone here has used an AI, [81:17] talked to an AI about, I don't know, [81:20] like [81:21] what to wear to a tech industry [81:24] conference. [81:25] Your agent did a great job, by the way, [81:27] Will. [81:28] >> Thank [laughter] you. [81:29] >> [gasps] [81:30] >> Um [81:31] No, I mean, your agent knows your your [81:33] style, your budget, your sizing, other [81:36] relevant context. [81:39] But lately, what has happened and I [81:41] think we'll see more of this is sort of [81:42] these unique finds. Um [81:45] the the sort of items you delight in, [81:48] things that make for really thoughtful [81:49] gifts. [81:51] You know, recently, as an example, an [81:54] engineer on our on my team [81:56] had had a baby and [81:59] my agent helped with baby gift ideas and [82:02] one of them was this [82:04] book, a super small production run, but [82:06] it's called Computer Engineering for [82:08] Babies. [82:08] >> [snorts] [82:10] >> It's a baby book. It's just adorable and [82:13] >> um [82:14] it's the perfect gift. So anyway, I [82:17] think these things used to be hard to [82:20] find or serendipitous at best. And [82:23] agents will do this way better. But I [82:26] should note that [82:28] it's [82:29] a lot more than just syncing a product [82:31] feed to agents, you know, someone needs [82:34] to check the inventory, [82:36] uh calculate the taxes, [82:38] loyalty programs and discounts, process [82:41] a payment, prevent any fraud, and oh [82:44] yeah, get the things to your door. [82:46] >> Yep. [82:47] >> And this is really where Shopify excels. [82:50] For [82:51] over 20 years, we've been [82:54] unifying this kind of complex commerce [82:56] life cycle [82:58] into, [83:00] you know, what we hope is one magical [83:02] experience. [83:03] >> Yep. [83:04] We love the partnership. Thank you for [83:05] being here. [83:06] >> Thanks for having me. [83:07] Thanks, everyone. [83:09] >> [music] [83:15] >> All right. [83:18] So, we started off this morning talking [83:20] about the rise of agents. [83:22] But agents are only as capable as the [83:25] models and the data that power them. [83:28] As Patrick shared earlier that last [83:29] year, 1.6% of global GDP ran on Stripe. [83:34] Well, that amounted to last year 70 [83:36] trillion commercial data points flowing [83:39] through our systems. [83:42] So, for years, our models trained on all [83:44] that data have worked quietly in the [83:46] background powering many of the products [83:47] you saw today. [83:50] We've heard from you that you want more [83:52] than that. You don't just want the [83:53] results in the products, you want the [83:56] actual intelligence behind them as a [83:58] programmable, flexible building block. [84:01] So, today, [84:03] we're introducing something very [84:04] exciting [84:05] that gives you and your agents exactly [84:07] that. [84:09] It's called Stripe Signals. [84:12] So, Stripe Signals gives you direct [84:14] access to our network intelligence. [84:16] Send us a customer, a transaction, a [84:19] business on or off Stripe, and we return [84:22] a score and an explanation in real time. [84:26] Is this transaction risky? Can I trust [84:28] this business? How can I help convert [84:30] this customer? [84:32] Stripe's network data is one of the most [84:35] valuable assets in commerce. [84:38] And Signals distills that value and puts [84:41] it in your hands. [84:44] Cursor, DoorDash, Navan, Cloudflare, and [84:46] 11:11 Labs are all already building the [84:48] Signals and we're launching with 10 [84:50] Signals today and many more are on the [84:52] way. [84:56] Now, we've covered a lot today across [84:57] payments, Radar, Revenue, Money [84:59] Management, and Embedded Finance. [85:02] And all of it, every single launch has [85:04] been inspired by your ideas and your [85:06] feedback. [85:08] But believe it or not, there's even [85:10] more, a lot more that we didn't have [85:12] time to get to. [85:14] So, I'm excited to announce that there's [85:16] finally somewhere you can see it all, [85:19] our brand new public roadmap. [85:22] >> [applause] [85:23] >> Yeah. [85:27] Yeah, you've been asking us for this for [85:28] years and candidly, it's kind of my [85:29] fault you don't have it, so I'm sorry, [85:31] but it's live now at stripe.com/roadmap. [85:35] You know, we know that many of you are [85:37] designing your roadmaps around ours, so [85:39] we really care that you know what's [85:41] coming. [85:42] This roadmap is the real deal. It's a [85:44] precise, itemized list with hundreds of [85:47] detailed entries going through Q1 2027. [85:50] And of course, it won't stay static. You [85:52] know, we're moving at breakneck speed, [85:53] so we'll be adding to it consistently. [85:57] Now, we strive to make Stripe the [85:59] fastest improving infrastructure in the [86:01] world. [86:02] So, to that end, if you look underneath [86:04] all the individual products here, we've [86:07] made some deep architectural upgrades to [86:09] Stripe. [86:10] We're making Stripe even more [86:11] programmable, so you can run as much of [86:13] your business on Stripe as you want and [86:15] build entirely new business models on [86:16] our primitives. [86:18] We're activating the Stripe network. You [86:20] know, you all are a coalition of the [86:23] world's fastest growing companies. And [86:25] this network should serve to protect and [86:28] propel your business. [86:30] And of course, we're building the [86:31] economic infrastructure for AI. [86:35] You know, the AI gigantic era isn't [86:37] coming soon. [86:38] It's happening right now. [86:40] So, I hope you leave today with a bunch [86:42] of specific ideas for how to seize this [86:44] moment. [86:45] It will be shaped by what all of you in [86:47] this room build and what we can build [86:50] together. [86:51] And it's going to happen fast. [86:54] Thank you everyone and enjoy Stripe [86:56] Sessions. [86:57] >> [applause] [87:01] [music]