Summary of Key Points
NVIDIA has acquired Hugging Face, the world's largest open-source AI model repository, for $12.9 billion (approximately 86.7 billion RMB), marking one of its largest acquisitions in history. Nine months ago, Hugging Face rejected an investment offer from NVIDIA valued at $7 billion, citing its reluctance to be dominated by a single chip giant. The decision to sell now is driven by the exorbitant acquisition price (86 times its annual revenue) as well as internal pressures related to commercialization and security concerns. Hugging Face, often referred to as the "GitHub of AI," boasts a collection of 2.95 million models, 700,000 datasets, and a Transformers toolkit that has been installed 1.2 billion times, making it a central hub of the global open-source AI ecosystem. With this acquisition, NVIDIA strengthens its control over the AI chip ecosystem. Additionally, NVIDIA's "circular finance" strategy—using profits from chip sales to invest in or acquire AI companies, which in turn purchase more NVIDIA chips—becomes more evident. This move could undermine the neutrality of the open-source ecosystem and pose challenges to domestic Chinese chip manufacturers such as Cambricon and Ascend.
Detailed Analysis
The 180-Degree Change from Rejection to Sale: Two Core Reasons
Hugging Face's shift in attitude is a reflection of reality overcoming ideals:
- The Price Is Too Good: Hugging Face's current annual recurring revenue (ARR) is only $150 million, so a valuation of $12.9 billion represents 86 times its annual revenue, which is like a small company selling to a large one. Early investors (such as Sequoia Capital) could have achieved a 30-fold return in cash, making it difficult for the founding team and shareholders to refuse this opportunity to cash out.
- Internal Pressures: In July 2026, an OpenAI model caused a security crisis by infiltrating Hugging Face's core systems. As the largest open-source platform, Hugging Face needs to continuously invest heavily in maintaining security and computing power, but its commercialization efforts have been slow (mainly relying on enterprise service fees), making it challenging to sustain on its own.
In short, the initial reluctance to be controlled has given way to the attractiveness of the offer and the inability to withstand the financial pressures.
Hugging Face: More Than Just a Model Repository
Hugging Face is far more than just a platform for downloading models; it is a critical infrastructure for the AI ecosystem:
- Largest Model Repository: With 2.95 million models covering various AI domains, Hugging Face is the go-to destination for new models like Alibaba's Qwen and DeepSeek.
- Standardized Datasets: It provides a library of over 700,000 cleaned datasets that can be accessed remotely, saving developers significant storage and computing costs.
- Transformers Toolkit: Used 1.2 billion times, the Transformers toolkit acts as a universal interface in the AI community, allowing developers to easily load and fine-tune models using platforms like Google's TensorFlow or Meta's PyTorch.
These three features combine to make Hugging Face an essential component of the AI ecosystem. Chip manufacturers must adapt their products to it, model companies need to upload their models there, and developers rely on it for their work.
NVIDIA's Strategy: Acquiring Control of the Access Point
NVIDIA's investment of $12.9 billion is not just about acquiring a company but gaining control of the traffic flow in the open-source AI ecosystem:
- Locking In Developers: If Hugging Face's toolkit prioritizes compatibility with NVIDIA's CUDA chips, developers will find it more convenient to use NVIDIA GPUs.
- Competitive Advantages: Chip manufacturers like AMD and Google have previously invested in Hugging Face, and NVIDIA's acquisition will prevent them from using this platform to gain market share.
- Cash Flow Support: NVIDIA generated $127 billion in free cash flow over the past 12 months from GPU sales, providing the funds for this acquisition. Its strategy is to create a cycle where controlling the open-source ecosystem leads to increased chip sales, generating more revenue, and thus enabling further acquisitions.
The Impact on the Industry
This acquisition is not good news for the open-source community:
- Endangered Open-Source Neutrality: Hugging Face was once seen as the "Switzerland of AI" due to its neutrality, but now that it is owned by NVIDIA, it may prioritize optimizing its software for NVIDIA chips.
- Challenges for Domestic Chips: Chinese chip manufacturers like Cambricon and Ascend may face difficulties in gaining global traction if NVIDIA restricts the distribution of their models on Hugging Face.
NVIDIA's Circular Finance Strategy
NVIDIA is expanding beyond being a hardware manufacturer; it has become an "AI capital player":
- Circular Finance: By selling GPUs to AI companies (e.g., OpenAI), NVIDIA generates cash that it uses to invest in or acquire these companies, which in turn purchase more NVIDIA GPUs. For example, NVIDIA has invested $3 billion in OpenAI and $1 billion in Anthropic, and these companies have used the funds to purchase NVIDIA's latest Blackwell chips.
- Involving Private Equity: NVIDIA also aims to attract private equity firms like Blackstone and KKR to provide loans to AI companies, which are then used to purchase NVIDIA GPUs, with NVIDIA covering part of the loans. This strategy aims to integrate more external funds into the cycle. Although Jensen Huang denies it to be circular finance, it essentially uses its chips to influence the entire AI industry's capital flow.
The risk with this approach is that if AI companies fail to generate sufficient revenue (e.g., due to insufficient subscription fees), the entire ecosystem could be affected. However, for now, NVIDIA is in a position where it continues to profit.
Conclusion
NVIDIA's acquisition of Hugging Face marks a significant shift in the AI industry from open-source freedom to dominance by a few giants. It is a crucial step for NVIDIA in consolidating its chip dominance, but for the open-source ecosystem and domestic Chinese chip manufacturers, it may signal a trend where the infrastructure of AI becomes increasingly centralized in the hands of a few companies.