Nvidia reportedly plans to acquire Hugging Face for $12.9 billion, exceeding its revenue by over 80x.
According to The Information, citing a person familiar with the deal, Nvidia has agreed to acquire Hugging Face for $12.9 billion. If confirmed, the acquisition would strengthen Nvidia’s open-model strategy, create an additional channel for selling AI hardware, and help defend its dominant position in the hardware market as Anthropic, Google, OpenAI, and other major hyperscalers develop their own custom accelerators.
Nvidia currently sells hundreds of billions of dollars worth of AI hardware annually. While the ubiquity of its CUDA software stack and leading hardware performance are the primary drivers of its market success, another critical factor is that many AI models are trained on and optimized for Nvidia architecture. Consequently, the greater the number of models built on Nvidia hardware, the stronger the long-term demand for its products.
Hugging Face is an AI development platform best known for the Hugging Face Hub, a GitHub-style repository where researchers and developers publish, discover, download, and collaborate on AI models, datasets, and applications. The company also develops widely adopted software such as the Transformers library and provides tools and cloud services for training, optimizing, and deploying models across various types of AI hardware.
Beyond hosting models and datasets, Hugging Face offers software that helps developers optimize and deploy AI workloads across different CPUs, GPUs, and AI accelerators. Its Inference Endpoints service enables customers to run models on managed infrastructure hosted by AWS, Google Cloud, and Microsoft Azure. Users can select their preferred provider, region, hardware type, and instance size. Notably, the underlying hardware offered by Amazon, Google, and Microsoft is not exclusively Nvidia. Depending on the provider, Hugging Face currently supports configurations including AWS Inferentia, AMD Instinct, Google TPUs, Intel CPUs, and Nvidia accelerators.
The combination of OpenAI models, extensive datasets, popular applications, and cross-hardware optimization capabilities makes Hugging Face strategically vital to Nvidia. On one hand, Nvidia could theoretically make the platform rely exclusively on its hardware, though this would likely trigger significant community backlash and is unlikely in the near term. On the other hand, Nvidia wants open models to remain competitive against proprietary offerings from companies like Anthropic and OpenAI, which may eventually optimize their systems for non-Nvidia hardware. To counter this, Nvidia has been developing its own Nemotron open models and has committed tens of billions of dollars to the initiative. Furthermore, as Hugging Face expands, so does the overall adoption of AI hardware—and Nvidia’s share within it.
Despite its rapid growth, Hugging Face’s revenue remains modest relative to the proposed acquisition price. The decade-old company recently reached approximately $150 million in annualized revenue, up from roughly $100 million several months prior, valuing the deal at around 80 times forward revenue—a figure that underscores the transaction’s strategic rather than purely financial rationale. Negotiations reportedly began after Hugging Face received acquisition interest from other parties.
CEO and co-founder Clem Delangue said in June that paying subscribers doubled during the first half of 2026, and recently stated the company was close to profitability. This surge in demand has been partly fueled by improving open models from Chinese developers including Z.ai, Moonshot, and DeepSeek.
Should the takeover proceed, it will represent another step in Nvidia’s increasingly aggressive expansion across the AI ecosystem, spanning hardware, models, and software. Last week, Nvidia agreed to pay $6 billion to license development technology from open-model developer Poolside and offered employment to more than 100 of its staff. The company has also acquired Groq, Enfabrica, Essential AI, Illumex, and Kumo AI, among others.