Nvidia agreed to buy open-source AI platform Hugging Face for $12.9 billion, the chipmaker’s second-largest acquisition, as it expands beyond hardware into the software layer of the AI stack.
The deal, first reported by The Information on August 26 and confirmed in a CNBC interview with Hugging Face CEO Clem Delangue on September 3, marks a significant shift for Nvidia. The company known primarily for GPU chips is now investing heavily in the platforms where AI models are trained, stored, and distributed.
Delangue told CNBC he approached Nvidia CEO Jensen Huang weeks before the deal was announced, rather than being approached by Nvidia. The timing matters: it suggests Hugging Face sought Nvidia’s resources proactively, rather than being swept up in a broader acquisition wave. Huang said in a blog post that the deal “will expand access to AI for developers and institutions worldwide.”
A platform, not just a model hub
Hugging Face operates what has become the default repository for open-source AI models, datasets, and tools. The platform hosts more than 1.5 million models and has become central to how researchers and developers share and deploy AI systems. Its Transformers library is among the most widely used tools in machine learning, referenced in thousands of research papers and production deployments across companies of all sizes.
For Nvidia, the acquisition provides direct access to the ecosystem where AI models get distributed. It also gives Nvidia cloud infrastructure capabilities, a market it has been trying to re-enter after selling its cloud business years ago. Owning the platform where models live could strengthen Nvidia’s position as the default hardware choice for AI training and inference across both cloud and on-premises deployments.
The deal also covers Hugging Face’s growing enterprise business. The company has been building commercial features for organizations that need to host private models, manage access controls, and deploy AI systems in regulated environments. That revenue stream, while still early, gives Nvidia a foothold in the enterprise AI platform market that hyperscalers like Amazon, Google, and Microsoft already dominate.
The price tells a story
The $12.9 billion price tag represents a nearly threefold increase from Hugging Face’s last known valuation of $4.5 billion in 2023, when Salesforce Ventures led a $235 million funding round with participation from Alphabet’s GV, IBM Ventures, and Nvidia itself.
Hugging Face turned down a $500 million investment from Nvidia late last year that would have valued the company at $7 billion, according to the Financial Times. The eventual acquisition price of $12.9 billion suggests both companies saw more value in full integration than a minority stake. For investors who participated in earlier rounds, the deal delivers a strong return on paper.
The valuation jump also reflects how much the AI infrastructure landscape has changed in three years. Hugging Face’s model count, user base, and enterprise revenue have all grown substantially since 2023, and the competitive pressure from proprietary platforms like OpenAI and Anthropic has made an independent Hugging Face harder to sustain at scale without a deeper-pocketed backer.
Context: Nvidia’s acquisition strategy
The Hugging Face deal is Nvidia’s second-biggest acquisition after the $6.9 billion purchase of Mellanox Technologies in 2020, which gave Nvidia high-speed networking capabilities for data centers. It comes as the broader AI industry consolidates around a handful of infrastructure providers. Stripe recently acquired OpenRouter, another AI infrastructure startup, and major cloud providers have been building or buying their own model-hosting capabilities.
The deal also arrives amid an ongoing debate about open-weight AI models. Huang and 24 other companies, including Hugging Face, signed a letter to the U.S. government earlier this year urging officials to support open models rather than restrict them. Owning Hugging Face positions Nvidia on the side of that argument with significant financial backing and operational control over the platform’s future direction.
What happens next
The acquisition still requires regulatory approval. Nvidia and Hugging Face will need to navigate antitrust review, particularly given Nvidia’s dominant position in AI chips and Hugging Face’s central role in open-source model distribution. The deal is expected to close in late 2026 or early 2027, depending on regulatory timelines and any conditions imposed by reviewing authorities.
For the open-source AI community, the immediate question is whether Nvidia will preserve Hugging Face’s platform-neutral approach or steer it toward closer integration with Nvidia’s hardware ecosystem. Delangue has built Hugging Face’s brand on supporting models across all hardware platforms, and any shift toward Nvidia exclusivity could alienate parts of the developer community that depend on the platform’s neutrality for their workflows.
The broader question is what this means for AI competition. Nvidia already controls the majority of AI training hardware. Adding the dominant open-source model platform gives it influence over both the chips that train models and the repository where those models get shared. Whether that concentration benefits or constrains the AI ecosystem will depend on how Nvidia chooses to manage the conflict between its hardware business and Hugging Face’s open-source ethos.
The two companies had already been collaborating closely before the acquisition. In July, Nvidia and Hugging Face expanded their LeRobot integration to bring new AI models to the open robotics community, giving developers tools to train and deploy robot policies on Nvidia hardware. That partnership gave both sides a preview of how deeply their platforms could interoperate.
Competitors are watching closely. AMD has been building its own ROCm software ecosystem to challenge Nvidia’s CUDA dominance, and an Nvidia-owned Hugging Face could make it harder for AMD to gain traction in model optimization and deployment workflows. Intel, meanwhile, has been investing in open-source AI tooling through its Habana Labs division, and the Hugging Face deal could narrow the channels through which Intel reaches AI developers.
discussion