Table of Contents
Imagine the platform that has become essential for millions of AI developers falling under the control of the world’s largest chip manufacturer. This is the scenario unfolding with the announced acquisition of Hugging Face by Nvidia, a move that could profoundly redefine the balance of the open-source ecosystem in artificial intelligence.
According to several economic media outlets, including The Information, Bloomberg, and TechCrunch, Nvidia has reportedly reached an agreement to acquire Hugging Face for $12.9 billion, approximately 13 billion. This amount values the company at an exceptional multiple, close to 86 times its estimated annual revenue of around $150 million.
This valuation marks a spectacular leap compared to Hugging Face’s last known fundraising: $235 million raised in 2023, which then valued the startup at $4.5 billion, in a round led by Salesforce Ventures with participation from GV (Alphabet), IBM Ventures, and already Nvidia. Notably, Hugging Face had previously declined an investment offer of $500 million from Nvidia, which would have valued the company at $7 billion, partly to avoid dependence on a dominant investor.
At this stage, neither Nvidia nor Hugging Face has officially confirmed the transaction. Business Insider mentions advanced discussions without a signed agreement, while The Information claims that an agreement has already been reached between the two parties.
Often compared to a GitHub for artificial intelligence, Hugging Face hosts nearly 3 million public models as well as countless datasets and open-source tools, used daily by millions of developers. However, a handful of these models concentrate the majority of traffic and downloads, making it a prime observation point for technological trends in the sector.
For Nvidia, the interest is primarily strategic. Most models downloaded from the platform must then be run on computing infrastructures, predominantly equipped with Nvidia GPUs, whether on-site or in the cloud. By establishing itself at this key stage of the chain, often referred to as the “developer funnel,” Nvidia ensures that an increasing share of computing demand is channeled towards its own chips and its CUDA software ecosystem.
The operation would also mark Nvidia’s return to cloud computing, an activity the group had put on hold about a year ago with its DGX Cloud offering. Hugging Face already offers paid hosting services allowing developers to run or fine-tune models without managing their own GPU infrastructure. By acquiring the platform, Nvidia would thus acquire an already operational computing power rental activity, without having to relaunch its own service from scratch.
This dimension also has a financial scope. Nvidia has committed to guaranteeing part of the cost of tens of billions of dollars in cloud contracts concluded by its clients. In the event of unused capacity, the group could find itself having to offload it, which Hugging Face’s clientele could precisely absorb.
The acquisition comes as Clément Delangue, co-founder and CEO of Hugging Face, multiplies public positions in favor of open models. He notably explained that his company used a version modified by Nvidia of a Chinese open-source model to recover from a recent cyberattack. Delangue also co-signed, alongside Nvidia’s CEO Jensen Huang and about twenty other executives, a letter urging US authorities to support open models rather than restrict them, in a context where China is increasingly dominant in this segment.
It is precisely this neutrality, always claimed by Hugging Face, that worries some observers. The platform owes a large part of its success to having gathered contributions from all the major players in the sector: Nvidia, AMD, Intel, the main cloud providers, as well as countless research labs and independent developers.
Some analysts believe that Nvidia’s financial and technical resources could instead accelerate the adoption of open models, by offering enhanced support to the ecosystem. Others fear, on the contrary, a gradual shift of the platform towards Nvidia’s technologies and hardware, to the detriment of competing solutions, which would undermine its position as an impartial reference player.
The scale of the operation and its strategic importance for the entire sector suggest a thorough examination by competition authorities, in the United States, the European Union, and probably the United Kingdom. Regulators should notably question the risk that Nvidia directs Hugging Face’s previously neutral hosting towards its own hardware, to the detriment of AMD, Intel, or chips designed internally by major tech groups.
Nvidia has already experienced friction with antitrust authorities in the past. In February 2025, the group filed a lawsuit against European regulators regarding the review of its acquisition of Run:ai, arguing that the Italian referral procedure exceeded the limits set for the control of small acquisitions. Even earlier, the planned acquisition of Arm for $40 billion was abandoned in 2022 under pressure from international antitrust authorities. A $12.9 billion transaction involving such a central infrastructure as Hugging Face could revive this type of regulatory scrutiny.
For organizations that currently rely on Hugging Face in their artificial intelligence projects, this acquisition requires particular vigilance. It is notably recommended to assess the level of dependence on the platform, anticipate potential changes in governance and usage conditions, and implement portability strategies to limit risks related to potential alignment with a single provider.
This operation is also part of a broader trend of consolidating the open layers of the AI ecosystem (models, routing, development tools) by major industrial players, similar to Stripe’s acquisition of OpenRouter. The common goal is to secure strategic positions in the infrastructure before enterprise demand truly explodes.
This consolidation movement coincides with the emergence of new open-source licenses dedicated to artificial intelligence, such as OpenMDW-1.1 (Open Model, Data and Weights), supported by the Linux Foundation since May 2026. These licenses aim to clarify the rules of use, distribution, and modification of open models, at a time when the question of the neutrality of the platforms hosting them becomes central for the entire AI development community.