编辑：米奇 来源：财经会议圈

蔡崇信 VivaTech 2026 巴黎峰会完整对谈实录（中英对照，官方整合全文）

时间：2026 年 6 月 18 日 巴黎 VivaTech 

主论坛形式：圆桌深度访谈（主持人：Publicis 集团 Maurice Lévy）

嘉宾：阿里巴巴集团主席 蔡崇信（Joe Tsai）

主题：AI 全栈战略、50 万亿 AI 市场、开源、欧洲数据主权、产业落地

蔡崇信在VivaTech欧洲科技峰会上对AI的核心观点，以及阿里全面投入AI的战略逻辑，核心内容可总结为6点：

蔡崇信VivaTech现场全文

主持人开场

Maurice Lévy：欢迎来到 VivaTech，今天我们对话阿里巴巴主席蔡崇信。过去 25 年阿里从中国电商起步，如今站在 AI 变革的关键节点。全球都在关注两件事：第一，AI 到底有多大市场空间；第二，阿里巴巴如何构建 AI 长期竞争力，以及中国科技企业能给欧洲带来什么合作机会。Joe，先请你分享对 AI 市场规模的核心判断。

蔡崇信完整发言（中文完整版 + 对应英文原文）

AI 的 50 万亿美元总潜在市场（核心开篇观点）

蔡崇信（中文）：很多行业分析师习惯用企业 IT 支出、软件市场规模去测算 AI 空间，这个视角太狭窄了。AI 的本质，是创造一套等同于人类智能、人类生产力的全新价值单元。当前全球 GDP 总量超过 100 万亿美元，其中至少一半，也就是 50 万亿美元，全部来自人类脑力劳动、生产效率创造的价值。这 50 万亿美元，就是 AI 完整的潜在市场（TAM）。这也是阿里巴巴选择全面 all in 人工智能的根本逻辑。如果只把 AI 理解为聊天机器人、云服务售卖，上千亿的投入会显得成本极高；但如果把 AI 看作电力、铁路一样，支撑未来数十年全球经济的底层基础设施，今天的大规模投入，就是在搭建全人类下一代生产力底座。

英文原文：Most analysts measure AI market size based on enterprise IT or software revenue, which is far too narrow.AI creates a new unit of value equivalent to human intelligence and human productivity. Global GDP exceeds $100 trillion, and at least half of that — $50 trillion — comes from human cognitive labor and productivity gains. That $50 trillion is AI’s total addressable market.This is why Alibaba is fully committed to AI. If you only see AI as chatbots or cloud subscriptions, multi-billion-dollar investment looks costly. But if you view AI as foundational infrastructure like electricity or railroads that will power the global economy for decades, today’s heavy investment is building the productivity backbone for the next generation.

阿里千亿级 AI 投入与现金流支撑

蔡崇信（中文）：抓住 50 万亿级机遇，必须持续、不计周期的大额投入。阿里巴巴核心电商业务每年稳定产生约 250 亿美元自由现金流，这是我们持续加码 AI 的坚实资金底盘。早在 2025 年 2 月，集团正式承诺：未来三年，累计投入超 3800 亿元人民币（折合 530 亿美元），全部用于 AI 算力、云基础设施、自研芯片与大模型研发。放到中国产业视角看，当前国内 AI 算力、全产业链基础设施的整体投入依然不足，所有中国科技企业都应当加大长期布局投入。阿里是国内少数具备稳定现金流、能长期扛住重资产投入的企业，我们有责任持续加码。

英文原文：Capturing this$50 trillion opportunity requires sustained, long-cycle massive investment.Our core e-commerce business generates roughly $25 billion in free cash flow every year, forming a solid financial foundation for continuous AI investment. Back in February 2025, Alibaba committed to investing over RMB 380 billion ($53 billion) across three years, fully dedicated to AI computing power, cloud infrastructure, self-developed chips and foundation models.From China’s industrial perspective, overall investment in AI computing and full industrial chain infrastructure remains insufficient. All Chinese tech companies should ramp up long-term R&D spending. Alibaba is among the few domestic firms with steady cash flow to sustain heavy capital expenditure, and we bear the responsibility to keep investing aggressively.

3四层全栈 AI 战略（阿里核心差异化优势）

蔡崇信（中文）：我们的核心战略是全栈一体化布局，没有人能精准预判未来 AI 价值会沉淀在哪一层，完整掌控产业链每一环，才能随市场变化灵活调整、持续沉淀壁垒。整个 AI 产业链分为四层：第一层是能源层：中国拥有高效、低成本的绿色能源供给，这是我们发展算力得天独厚的基础优势，海外多数企业不具备同等条件；第二层是基础设施层：平头哥自研 GPU 芯片，累计交付 56 万片，搭配阿里云全球分布式算力集群，构建底层算力底座；第三层是模型层：通义千问（Qwen）开源大模型，已开放 300 + 系列模型，全球下载量突破 10 亿次，是全球最主流的开源模型之一；第四层是应用层：阿里拥有电商、即时零售、本地生活、地图、文旅等海量真实商业场景，为 AI 提供规模化落地、持续迭代的数据土壤。

全球范围内，同时完整覆盖算力、基础大模型、大规模真实商业场景的科技公司非常稀少。全栈布局最大价值不是 “什么都做”，而是内部协同闭环：业务场景产生真实业务数据，反哺大模型持续优化；云与芯片降低模型训练、推理成本，反过来让 AI 应用大规模普及，形成正向循环。

英文原文：Our core strategy is end-to-end full-stack AI. No one can accurately predict which layer will capture most AI value long-term. Controlling every segment of the industrial chain lets us adapt flexibly as markets evolve and build sustainable moats.The full AI stack has four layers:

Energy layer: China provides efficient, low-cost green energy supply, a unique advantage for computing power that most overseas competitors lack.

Infrastructure layer: Our T-Head self-developed GPUs have delivered over 560,000 chips, paired with Alibaba Cloud’s global distributed computing clusters to form the underlying computing foundation.

开源 AI：回应欧洲技术主权、数据隐私诉求

蔡崇信（中文）：今天欧洲企业最关心两个问题：技术自主可控、本地数据隐私安全，开源模型恰好是最优解决方案。企业基于开源模型，完全可以在本地机房独立部署、自主微调，所有核心业务数据全程留在企业自有防火墙内，不需要把敏感数据上传至第三方外部平台，完美契合欧盟 GDPR、欧洲数据主权政策。我常打一个比喻：全球技术格局下，企业不能把所有鸡蛋放进同一个技术供应商的篮子。开源生态就是企业的安全备份，摆脱单一厂商锁定、规避 “杀戮开关” 风险。客观来讲，当前全球 AI 开源浪潮的核心推动力量来自中国企业。过去几年阿里团队持续开放前沿模型，未来我们还会持续加码开源投入，向全球企业开放轻量化、可本地部署的 Qwen 系列模型。

英文原文：European enterprises today focus on two core priorities: technological sovereignty and local data privacy. Open-source models deliver the optimal solution.Companies can fully deploy and fine-tune open models on-premises, keeping all sensitive business data behind their own firewalls, without transferring confidential data to third-party external platforms — fully aligned with GDPR and Europe’s data sovereignty agenda.I often use a simple analogy: in today’s global tech landscape, enterprises should not put all eggs in one vendor’s basket. Open-source ecosystems act as a safety backup, freeing businesses from single-vendor lock-in and eliminating the risk of remote kill switches.Objectively, Chinese enterprises are the primary driving force behind global AI open-source development. Alibaba teams have continuously released cutting-edge open models over the past years, and we will keep investing heavily in open-source, offering lightweight, deployable Qwen variants for enterprises worldwide.

AI 产业落地：制造业是下一个爆发赛道，欧洲深度合作

蔡崇信（中文）：大众目前看到的 AI 应用大多面向 C 端消费者，但真正会拉动万亿级增量的，是工业制造业 AI。阿里巴巴已经和博世等欧洲头部制造企业落地合作，用 AI 优化产品研发设计、自动化产线测试、全流程质量管控、供应链预测。制造业拥有海量设备数据、生产流程数据，是大模型落地价值最高的场景。中欧在 AI 产业上具备极强互补性：欧洲拥有顶尖工业制造、精密工艺、合规治理经验；中国拥有成熟算力基础设施、海量市场场景、规模化 AI 落地实践。双方完全可以建立产业协同，欧洲企业提供工业 know-how，阿里输出算力、开源模型、云平台，共同打造面向全球的工业 AI 解决方案。

英文原文：Most visible AI applications today target consumers, but the trillion-dollar growth engine will come from industrial manufacturing AI.Alibaba has launched partnerships with leading European manufacturers including Bosch, applying AI to product R&D design, automated production testing, full-lifecycle quality control and supply chain forecasting. Manufacturing generates massive equipment and production workflow data, making it the highest-value vertical for foundation model deployment.China and Europe hold highly complementary strengths in AI industry: Europe leads in advanced manufacturing, precision engineering and regulatory governance; China boasts mature computing infrastructure, massive market scenarios and large-scale AI commercialization experience. We can build deep industrial collaboration: European firms contribute industrial domain expertise, while Alibaba provides computing power, open-source models and cloud platforms, co-developing global industrial AI solutions together.

回应 AI 泡沫争议：短期波动不改变长期价值

蔡崇信（中文）：市场频繁讨论 AI 投资泡沫，但泡沫更多是金融层面的短期情绪，而非技术本身的长期价值。类比 2000 年互联网泡沫：资本市场出现剧烈波动，但互联网最终重塑了全球商业。AI 同理，短期算力、芯片赛道估值起伏，不会改变 AI 重构生产力的底层逻辑。判断 AI 技术价值的唯一标准，是能否落地真实产业场景、创造可量化的经营效率提升。阿里所有 AI 投入，全部围绕真实业务落地，拒绝脱离产业的纯概念研发，这也是我们抵御行业周期波动的核心底气。

英文原文：Many market participants debate an AI investment bubble, yet volatility stems mostly from short-term financial sentiment, not the long-term intrinsic value of the technology.We can draw a parallel to the 2000 internet bubble: capital markets saw extreme swings, but the internet ultimately reshaped global commerce. AI follows the same logic — short-term valuation fluctuations in chips and computing power will not alter AI’s fundamental ability to redefine productivity.The only metric to judge AI’s real value is its ability to deploy in real industrial scenarios and deliver measurable efficiency gains. Every dollar Alibaba invests in AI targets tangible business use cases; we reject pure conceptual R&D disconnected from real industries, which forms our core resilience against industry cycles.

结尾：全球化 AI 合作，拒绝技术割裂

蔡崇信（中文）：AI 是属于全人类的通用技术，技术割裂、技术阵营化只会抬高全球企业创新成本，延缓生产力升级。阿里巴巴坚持全球化开放路线，开源、云、AI 算力服务面向全球所有地区企业，包括欧洲本土中小企业。我们希望搭建中立、开放的技术合作平台，让不同国家、不同行业的企业共享 AI 基础设施红利，共同挖掘 50 万亿美元的生产力增量市场。

英文原文：AI is a general-purpose technology for all humanity. Fragmented tech blocs only raise global innovation costs and slow productivity progress.Alibaba adheres to a global open strategy: our open-source models, cloud and AI computing services are available to enterprises across all regions, including European small and medium businesses. We aim to build a neutral, open collaboration platform, allowing companies from all countries and industries to share the benefits of AI infrastructure and jointly unlock the $50 trillion productivity market.

主持人追问 & 补充问答精简实录

主持人：欧洲企业担心数据出境，开源模型如何彻底规避风险？蔡崇信：Qwen 全系列支持本地私有化部署，模型权重完全交付企业本地服务器，不需要向阿里传输任何原始业务数据；企业自主掌握微调、推理全流程，数据不出本地机房，完全符合欧盟数据合规要求。

主持人：阿里 3800 亿 AI 投入，短期会挤压利润吗？蔡崇信：电商稳定现金流可以平滑长期重投入，我们不追求短期利润最大化，优先完成 AI 基础设施建设；中长期 AI 会反向赋能电商、制造、本地生活全业务，形成新增长曲线。

主持人：中美 AI 竞争背景下，开源能否保持中立？蔡崇信：开源本身无国界，代码对全球开发者公开；阿里开源策略纯粹面向产业落地，不绑定单一区域政策，欧洲、美国、东南亚企业均可无差别使用、二次开发。

官方原文获取渠道