内容来源：https://aiweekly.co/issues/473

内容总结：

人工智能领域先驱人物杨立昆（Yann LeCun）近期离开Meta后，在短短四个月内为其新创立的巴黎人工智能实验室AMI筹集了高达10.3亿美元种子轮资金，公司估值达35亿美元，创下欧洲史上最大规模种子轮融资纪录。本轮投资方阵容凸显产业战略布局，包括凯辉基金、格雷克罗夫特、贝索斯探险、英伟达、三星、淡马锡及丰田风投等机构。

AMI实验室的核心技术路线区别于当前主流的大语言模型，专注于开发能够理解物理世界的“世界模型”。该技术基于杨立昆提出的JEPA架构，旨在通过抽象表征预测来学习物理规律，目标应用于机器人、硬件设计、医疗等需要理解因果关系与物理原理的产业领域。团队透露已与Meta就智能眼镜等短期应用场景展开技术部署探讨。

值得注意的是，AMI核心团队几乎全部来自Meta FAIR实验室，形成了一支独立于超大规模科技公司体系的前沿研究力量。杨立昆将公司定位为“中美之外的第三极”，凭借巴黎总部的地理位置与欧洲深厚的人工智能人才储备，试图在欧洲人工智能投资热潮中打造兼具技术领导力与区域自主性的产业标杆。

此次融资不仅标志着欧洲人工智能领域获得资本高度关注，也反映出产业界对能够深度融合现实世界的人工智能技术路线的战略押注。在开源研究原则与商业化应用并行的框架下，AMI的后续发展将成为观察人工智能技术演进与全球格局变化的重要案例。

中文翻译：

杨立昆刚刚离开Meta，在四个月内筹集了十亿美元，并将事业旗帜插在了巴黎。

据《财富》杂志报道，AMI实验室以35亿美元估值完成了10.3亿美元种子轮融资——这是欧洲史上最大规模的种子轮融资，全球范围内仅次于去年六月Thinking Machines实验室20亿美元的融资纪录。投资方名单堪称战略布局的明星阵容：国泰创投、Greycroft、贝索斯探险基金、英伟达、三星、淡马锡和丰田风投都支持这位图灵奖得主的愿景。杨立昆最初寻求5亿欧元融资，但市场需求远超预期，使得AMI能够筛选符合其产业路线的投资者。创始团队几乎全是Meta FAIR实验室的旧部——研究员谢赛宁和迈克尔·拉巴特，加上Meta副总裁洛朗·索利担任首席运营官——健康AI初创公司Nabla的亚历克斯·勒布伦则出任首席执行官。这不仅是一次融资，更是对超大规模平台生态体系之外世界顶尖AI研究团队的有意重组。

赞助内容
成为AI顾问，为客户实现"AI投资回报"
AI正在改变每个工作场所——但高管们唯恐自己的公司成为"AI零回报"的企业之一。
这正是您的用武之地。通过"AI创新实践"中经过验证的快速实现AI项目回报的方法，您可以建立六位数收入的咨询业务。
点击此处申请加入AI咨询项目→

新闻动态
当AI教父押注非文本领域，欧洲签下巨额支票
世界模型 vs. LLM正统理论
支撑其估值的技术赌注在于：AMI正在构建世界模型而非大语言模型，瞄准物理智能而非文本生成。SiliconANGLE报道称，AMI将采用杨立昆的JEPA架构，该架构通过学习预测世界的抽象表征，同时忽略传感器数据中不可预测的细节——好比理解皮球的弹跳物理原理而无需模拟每个像素。这种方法专为机器人、硬件设计和医疗保健领域打造，这些领域对因果律和物理规律的理解比生成合理文本更为重要。杨立昆告诉路透社，Ray-Ban Meta智能眼镜"可能是短期内的潜在应用场景之一"，并确认AMI已就此项技术部署与Meta展开洽谈。与此同时，STAT披露Nabla正通过CEO勒布伦获得AMI模型的早期使用权，两家公司虽未达成正式股权协议但已开展紧密合作。

世界模型在理论上并非新概念，但Futurum集团分析师质疑时代是否终于成熟，并指出AMI精心构建的投资联盟指向工业应用而非消费级聊天场景。Built In报道称杨立昆将制造商、汽车、航空航天和制药作为近期目标行业——这些领域需要能理解真实世界的AI，而非仅预测下一个词元。尽管侧重长期研究，HPCwire指出杨立昆计划快速发布首批模型并开源技术，在商业冒险中仍坚守其开放研究原则。

欧洲的AI冠军战略
杨立昆将AMI定位为硅谷难以复制的存在：一个明确非中美所属的前沿AI实验室，总部设于巴黎并依托欧洲深厚的AI人才库。时机选择具有战略意义——2025年AI投资占欧洲风投总额的35.5%，据《法国科技期刊》预测该比例将在2026年突破50%，使AMI既成为欧洲AI浪潮的受益者也是加速器。Crunchbase确认本轮融资不仅是欧洲最大种子轮，也是该地区史上最大AI交易之一，与Mistral的20亿美元和Nscale的20亿美元融资并列。地缘政治潜台词清晰可见：随着中美AI竞争白热化，杨立昆为欧洲政策制定者和投资者提供了兼具可信技术领导力与本土主权的第三极。

欢迎分享您对这种前景广阔的AI新路径的看法！
亚历克西斯

英文来源：

Yann LeCun just left Meta, raised a billion dollars in four months, and planted a flag in Paris.
AMI Labs closed a $1.03 billion seed round at a $3.5 billion valuation — Europe's largest seed ever, second globally only to Thinking Machines Lab's $2 billion raise last June, according to Fortune. The syndicate reads like a who's-who of strategic intent: Cathay, Greycroft, Bezos Expeditions, NVIDIA, Samsung, Temasek, and Toyota Ventures all backed the Turing Award winner's vision. LeCun initially sought €500 million but demand exceeded expectations, allowing AMI to cherry-pick investors aligned with its industrial roadmap. The founding team is almost entirely Meta FAIR alumni — researchers Saining Xie and Michael Rabbat, plus Meta VP Laurent Solly as COO — with Alex LeBrun from health AI startup Nabla as CEO. This isn't just a fundraise; it's a deliberate reconstruction of one of the world's best AI research teams outside the hyperscaler ecosystem.
Sponsor
Become an AI consultant and deliver ‘ROI with AI’ to your clients
AI is transforming every workplace – but executives are terrified of becoming one of the companies that gets “no ROI on AI.”
That’s where you come in, and how you can build a 6-figure consultancy with Innovating with AI’s proven methods for delivering fast ROI on AI projects.
Click here to request access to The AI Consultancy Project →
In the News
When the godfather of AI bets against text, Europe writes a check
World Models vs. The LLM Orthodoxy
Here's the technical bet that justifies the valuation: AMI is building world models instead of large language models, targeting physical intelligence over text generation. SiliconANGLE reports AMI will use LeCun's JEPA architecture, which learns to predict abstract representations of the world while ignoring unpredictable details in sensor data — think learning the physics of a bouncing ball without modeling every pixel. This approach is purpose-built for robotics, hardware design, and healthcare, domains where understanding causality and physics matters more than generating plausible text. LeCun told Reuters that Ray-Ban Meta smart glasses are "probably one of the shorter term potential applications" and confirmed AMI is already in talks with Meta about deploying the technology there. Meanwhile, STAT uncovered that Nabla is getting early access to AMI's models through CEO LeBrun, with companies already working closely despite no formal equity deal yet.
World models aren't new theoretically, but Futurum Group analysts ask whether the era has finally arrived, noting AMI's carefully constructed investor syndicate includes strategic players pointing to industrial applications rather than consumer chat. Built In reports LeCun is targeting manufacturers, automakers, aerospace, and pharma as near-term industries — sectors that need AI to understand the real world, not just predict the next token. Despite the long-term research focus, HPCwire notes LeCun plans to release first models quickly and open-source technology, staying true to his open research principles even while building a commercial venture.
Europe's AI Champion Play
LeCun is positioning AMI as something Silicon Valley can't easily replicate: a frontier AI lab that is explicitly neither Chinese nor American, headquartered in Paris and leveraging Europe's deep AI talent base. The timing is strategic. AI represented 35.5% of all European VC in 2025, and French Tech Journal reports that figure is predicted to top 50% in 2026, making AMI both beneficiary and accelerant of Europe's AI moment. Crunchbase confirms the round as not just Europe's largest seed but among the region's biggest AI deals ever, alongside Mistral's $2 billion and Nscale's $2 billion raises. The geopolitical subtext is clear: as US-China AI competition intensifies, LeCun is offering European policymakers and investors a third pole, one that combines credible technical leadership with local sovereignty.
Let me know your thoughts on this new promising approach to AI!
Alexis