英伟达以129亿美元收购HuggingFace
英伟达以129亿美元收购开源模型平台HuggingFace。HuggingFace将成为全球最大算力提供商英伟达的一部分。该交易涉及芯片厂商对AI软件生态的整合，可能影响未来AI芯片的供应与生态合作。
原文深度解读
本文报道英伟达以129亿美元收购开源AI模型平台HuggingFace。文章涉及英伟达、HuggingFace及其联合创始人Clem Delangue等主体，时间背景为2026年。关键数字包括129亿美元收购价、HuggingFace约1800万用户及1亿开发者目标。该交易体现芯片厂商对AI软件生态的整合，可能影响AI芯片的供应与生态合作。
- 英伟达以129亿美元收购HuggingFace，后者由三位法国企业家于2016年创立，是开源AI模型、数据集和软件库的主要平台。
- 英伟达AI软件副总裁Justin Boitano表示，英伟达致力于保持HuggingFace的开放和中立，使其与所有云提供商和硬件公司互操作。
- Boitano称，英伟达特定的运行时将与vLLM和SGLang等开源运行时共存，确保HuggingFace仓库中的开放模型能在所有硬件上优化。
- HuggingFace联合创始人兼CEO Clem Delangue表示，收购后所有生态参与者仍可使用开源基础构建产品，并强调其平台的开源和中立性。
- Delangue否认收购反映欧洲科技困境，并希望收购能对欧洲科技生态产生积极影响，计划纳入更多欧洲实验室的开放模型。
- Delangue称，英伟达的支持将使HuggingFace更快实现1亿开发者目标，目前平台约有1800万用户。
供应链影响
- 英伟达对HuggingFace的收购可能增强其对AI软件生态的影响力，进而影响AI芯片的软件兼容性和生态合作方向。
- HuggingFace保持硬件中立可能使其他芯片厂商（如AMD、Intel）继续获得开源模型支持，但英伟达的深度参与可能改变未来优化资源的分配。
- 收购可能加速AI模型在英伟达硬件上的部署，从而影响数据中心对AI芯片的需求结构，具体取决于生态开放程度。
- HuggingFace用户增长（目标1亿）可能带动对算力（包括英伟达芯片）的需求，但影响程度取决于平台使用模式。
- 欧洲开源模型（如Mistral）的纳入可能促进欧洲AI生态，但可能影响全球AI芯片市场的竞争格局，取决于各国政策与市场反应。
系统解读，仅作为判断线索，不构成备货、出货、涨价或投资建议。
英伟达以129亿美元收购HuggingFace。这家以表情符号命名的初创公司由三位法国企业家于2016年创立，已成为开源AI模型、数据集及相关软件库的首选平台。
在宣布交易的新闻发布会上，英伟达AI软件副总裁Justin Boitano面临许多关于HuggingFace产品中立性的问题。
“我们致力于帮助建设这个社区，管理并培育这个社区，”Boitano说。“我们希望继续与全球所有AI开发者合作，确保他们拥有一个开放、中立的平台，能与世界上每个云提供商、每个硬件公司互操作。”
英伟达相信维持一个由封闭模型提供商和开放模型组成的健康生态系统，他补充说，开放模型使公司能够构建自己的领域特定智能。英伟达是HuggingFace迄今为止最大的贡献者。
作者：Heather Campbell，Siemens EDA PAVE360市场总监，2026年9月7日
作者：Colin Duggan，BG Networks CEO兼联合创始人，2026年9月3日
作者：Teng Tang Yang，UMC产品营销部高级部门总监，2026年9月3日
Boitano表示，保持HuggingFace的硬件中立符合英伟达的利益。
“Hugging Face团队建立了一个了不起的社区，[…] 而做到这一点的唯一方式是建立开发者社区的信任，并让开发者能够从世界上任何他们想运行的平台访问模型和数据集，”他说。“我们有动力和动机保持这种开放和中立，确保开发者可以在任何他们想工作的地方工作，访问并将最好的开放模型连接到每个云提供商、每个新云（neocloud）、每个数据中心，并在每个设备上可用。”
Boitano表示，英伟达特定的运行时将继续与vLLM和SGLang等开源运行时共存，这些运行时是开放和中立的，确保HuggingFace仓库中的所有开放模型都能针对所有硬件进行优化。
HuggingFace是开源AI模型和数据集的主要平台，以表情符号命名。
HuggingFace联合创始人兼CEO Clem Delangue表示，收购仍意味着当今生态系统中的所有参与者都可以使用开源基础来构建他们想要的任何东西。
“我们做的绝大多数事情都是开源的——开放模型、开放数据集，这些本质上都是中立的——每个人都可以获取它们，每个人都可以优化它们，如果他们对我们的开源不满意，每个人都可以分叉它，”他说。“几乎从定义上讲，我们所做的一切都是开放和中立的，并赋予每个人构建更多东西的能力。”
Delangue是法国人，他驳斥了一种说法，即收购HuggingFace象征着在欧盟建立科技巨头的困难。
“三位联合创始人都是法国人，但我在创办公司时已经在美国，所以很自然地，我们用美国投资者建立了它，”他说。“我认为这将对欧洲科技生态产生非常积极的影响，因为显然创始人是法国人，团队中有很大一部分是法国人，所以我们希望这次收购后，他们能够向欧洲科技生态进行大量再投资。”
HuggingFace将继续支持来自全球的开放模型和开放数据集，Delangue表示他希望纳入更多来自欧洲实验室的开放模型，如法国的Mistral和德国的Black Forest Labs。
Delangue表示，HuggingFace团队很高兴现在有能力“长远思考”。英伟达的支持将使HuggingFace能够更快地实现平台上1亿开发者的目标，他说，并指出该平台目前约有1800万用户。
“我们认为我们正处于HuggingFace和开源AI整体的拐点，”他说。“我们认为有了资源、支持和英伟达的长期方法，我们将能够基本上做我们过去10年一直在做的事情，即支持开源AI社区，但达到下一个水平，以更大的规模，我们认为这对该领域和世界来说确实是必要的。”
另见：
在欧洲建立AI芯片初创公司需要什么
在本期AI with Sally中，我们与Axelera CEO兼联合创始人Fabrizio Del Maffeo坐下来，庆祝公司成立五周年。我们深入探讨了启动和扩展欧洲半导体初创公司的现实，涵盖了筹集资金、克服风险厌恶和争夺顶尖工程人才等关键挑战。
AI模型、并购、开放模型、软件
英伟达
Sally Ward-Foxton
Sally Ward-Foxton为EETimes.com和EETimes Europe杂志报道AI。Sally过去18年一直在伦敦撰写关于电子行业的文章。她曾为Electronic Design、ECN、Electronic Specifier: Design、Components in Electronics等多家新闻出版物撰稿。她拥有剑桥大学电气与电子工程硕士学位。
在LinkedIn上关注Sally
Nvidia has acquired HuggingFace for $12.9 billion. The startup, which is named after an emoji, was founded by three French entrepreneurs in 2016. It has become the go-to repository for open-source AI models, datasets, and associated software libraries.
In a press conference announcing the deal, Nvidia VP of AI software Justin Boitano was faced with many questions about the neutrality of HuggingFace’s offering.
“We’re committed to helping build this community, stewarding, and fostering this community, Boitano said. “We want to continue to work with all AI developers globally to make sure that they’ve got an open and neutral platform that interoperates with every cloud provider, every hardware company in the world.”
Nvidia believes in maintaining a healthy ecosystem of closed model providers and open models, he added, noting that open models give companies the ability to build their own domain-specific intelligence. Nvidia is by far HuggingFace’s biggest contributor.
By Heather Campbell, Marketing Director for PAVE360, Siemens EDA 09.07.2026
By By Colin Duggan, CEO and co-founder, BG Networks 09.03.2026
By Teng Tang Yang, Senior Division Director of Product Marketing Division, UMC 09.03.2026
Boitano said that it’s in Nvidia’s interests to keep HuggingFace hardware-neutral.
“The Hugging Face team has built an amazing community, […] and the only way you do that is by building the trust of the developer community and giving developers access to run the models and access the datasets from any platform in the world that they want to run it on,” he said. “We are incentivized and motivated to keep this open and neutral and ensure that developers can work wherever they want to work, accessing and connecting the best open models into and across every cloud provider, every neocloud, and into every data center and make availability on every device.”
Nvidia-specific runtimes will continue to exist alongside open-source runtimes such as vLLM and SGLang, which are open and neutral, ensuring all open models in the HuggingFace repository can be optimized for all hardware, Boitano said.
HuggingFace, the leading repository for open-source AI models and datasets, is named after an emoji.
HuggingFace co-founder and CEO Clem Delangue said that the acquisition would still mean all the players in today’s ecosystem can use open-source foundations to build whatever they want.
“The vast majority of what we do is open source—open models, open datasets that are by definition neutral—everyone can take them, everyone can optimize them, everyone can fork our open source if they’re not happy about it,” he said. “Almost by definition, everything that we do is open and neutral and gives everyone the capabilities to build more.”
Delangue, who is French, shot down a suggestion that the acquisition of HuggingFace was emblematic of the difficulties of building tech giants in the EU.
“All three co-founders are French, but I was already in the U.S. when we started the company, so quite naturally, we built it with American investors,” he said. “I think it’s going to create really positive effects on the European tech ecosystem because obviously the founders are French, a big part of the team is French, and so we hope that after this acquisition, they’ll be able to reinvest quite a lot in the European tech ecosystem.”
HuggingFace will continue to support open models and open datasets from around the globe, and Delangue said he hopes to include more open models out of European labs such as Mistral in France and Black Forest Labs in Germany.
Delangue said that the HuggingFace team is excited to now have the ability to “think long-term.” Nvidia’s support will enable HuggingFace to reach its goal of 100 million builders on the platform much more quickly, he said, noting that the platform has around 18 million users today.
“We think we’re at the inflection point for HuggingFace and for open-source AI in general,” he said. “We think with the resources, with the support, with the long-term approach of Nvidia, we’ll be able to do basically what we’ve been doing for the past 10 years, meaning, supporting the open-source AI community, but to the next level at a much bigger scale, which we think is really needed for the field and for the world.”
See also:
What It Takes to Build an AI Chip Startup in Europe
In this episode of AI with Sally, we sit down with Fabrizio Del Maffeo, CEO and co-founder of Axelera, to celebrate the company’s fifth anniversary. We dive into the realities of launching and scaling a European semiconductor startup, covering key challenges like raising capital, overcoming risk aversion, and competing for top engineering talent.
AI MODELS, MERGERS & ACQUISITIONS, OPEN MODELS, SOFTWARE
NVIDIA
Sally Ward-Foxton
Sally Ward-Foxton covers AI for EETimes.com and EETimes Europe magazine. Sally has spent the last 18 years writing about the electronics industry from London. She has written for Electronic Design, ECN, Electronic Specifier: Design, Components in Electronics, and many more news publications. She holds a Masters' degree in Electrical and Electronic Engineering from the University of Cambridge.
Follow Sally on LinkedIn