NVIDIA acquires Hugging Face: Jensen Huang puts $12.93 billion on the table to control the main AI modeling platform
NVIDIA has just closed a definitive agreement to buy Hugging Face. The company will disburse some 12.93 billion dollars in cash to complete the operation. Without a doubt, this is one of those bombshell news that, once again, shakes such quicksands as those of the AI sector. To understand the magnitude of the operation and what it means for the development of this technology, you have to understand well what Hugging Face is.
We are, greatly simplifying it, before the GitHub of AI models. It is a nerve center where the main open models come together and which, according to NVIDIA, are used by some 19 million developers and 200,000 organizations. Therefore, the hardware giant specialized in artificial intelligence is absorbing the AI software giant.
What Jensen Huang is comparing
The integration of the Hugging Face catalog, which houses more than 3 million models and 500,000 data sets, represents a decisive step in the company’s diversification within the sector that brings it the most profits. The firm chaired by Jensen Huang already controlled the physical part. It did so with its graphics for data centers and the CUDA ecosystem. Also with the arrival of RTX Spark, a platform designed, above all, for the execution of models locally. Despite this dominance, Huang was missing a key piece.
Now, the acquisition of Hugging Face puts the multinational in a strategic position to influence how companies download, evaluate and execute their models. This also provides you with data, key information to understand how the market is moving.
In the statement announcing the acquisition, Jensen Huang insisted that Hugging Face will continue to operate as an open and agnostic service with respect to the underlying infrastructure. The manager has stated that It will not be mandatory to use NVIDIA processors to build or deploy projects on the platform. According to the company, the portal will maintain multi-cloud compatibility and respect interoperability with hardware from different semiconductor manufacturers.
Despite official statements, operating on a platform considered neutral in terms of AI may raise doubts among competitors in the semiconductor sector. Hugging Face’s status as an independent evaluator and showcase for open models from Meta, Google or Mistral changes context when it is integrated into the NVIDIA structure. Support for rival technologies such as AMD ROCm or Intel Gaudi could gradually be diluted. They are assumptions, but not at all crazy.
Hugging Face needed a giant to back her up
NVIDIA is a technology giant. And, as we said, Hugging Face is also one, but more because of its ability to bring together developers of AI models and not so much because of its history or its infrastructure. The costs of bandwidth and storage were increasingly higher, especially as its popularity skyrocketed.
In the letter signed by Huang himself, the following is mentioned: “NVIDIA’s infrastructure, engineering and global reach can help improve the reliability of the platform, securitymodel evaluation, inference, and deployment capabilities, while preserving the open ecosystem that made Hugging Face such a foundational project.”
Therefore, NVIDIA is not only committed to sustaining the Hugging Face infrastructure, while maintaining it as a neutral space. It also talks about security, a sensitive topic considering the incident that occurred just a few weeks ago in which several AI agents entered its systems autonomously.
As usual, the formalization of the agreement will be subject to the approval of the competition authorities in the United States, the United Kingdom and the European Union. Antitrust scrutiny will analyze whether this integration could restrict equal access to development tools or limit competition in the global market. We will carefully follow the entire process.
