Nvidia confirmed, on September 3, that it's buying Hugging Face for roughly $13 billion — the platform more than 18 million developers and over 200,000 companies use to discover and publish AI models. It's Nvidia's second-largest acquisition ever, and the first time the company has committed serious money to something that isn't silicon. The detail that matters isn't the price tag — it's what Hugging Face represents today: the most neutral layer of open-source AI infrastructure.
What Changed
According to CNBC and Bloomberg, the deal includes roughly $11.9 billion payable to Hugging Face shareholders and up to $1 billion in retention equity for employees joining Nvidia. Closing is expected in the first half of 2027, subject to regulatory approval. Hugging Face hosts more than 3 million models and functions as the world's main repository and discovery showcase for open-source AI — used both by people building on Nvidia GPUs and by Nvidia's direct competitors. Nvidia said it will keep the platform open, along the same lines as today.
Why It Matters
Hugging Face is often called the "Switzerland of open-weight AI" — precisely because it hasn't favored any particular hardware or model vendor. Analysts at The Register and Yahoo Finance already flag the real risk: under Nvidia's ownership, library optimizations, quantization formats, and serving features could start shipping Nvidia-first, with other backends "catching up" months later — a roadmap bias that, on its own, is exactly the kind of competitive effect antitrust regulators look for. The direct precedent is Nvidia itself: in 2020, the company tried to buy Arm — another neutral layer the entire industry depends on — and the deal collapsed in 2022 after fierce opposition from Google, Microsoft, Qualcomm, and others, over exactly the same neutrality concern. AMD, Intel, and the in-house chip efforts at Google, Amazon, and OpenAI all have a direct interest in Hugging Face staying neutral, and are likely to raise objections during the regulatory review.
The Impact for Brazil
Brazilian companies that use open-weight models through Hugging Face — a real part of the cost-reduction strategy we've covered here with DeepSeek and other open models — should watch this process closely. If the deal moves forward, it's worth reassessing how much your company's model and hardware choices depend on a single discovery-and-distribution platform, and mapping an alternative in case optimizations start subtly favoring one vendor. This isn't a reason to panic now — Nvidia has promised to keep the platform open, and the Arm precedent shows this kind of acquisition may not even close — but it's a real reason for attention.
Entercast's Take
This might be the strongest data point this month for the theme we've followed since August: vendor concentration is a risk most companies underweight when choosing their AI stack, ahead of price or performance. We saw it in SpaceX's acquisition of Cursor, in the geopolitical fight over chip supply chains between Pax Silica and WAICO, and in OpenAI's own chip, built specifically to reduce Nvidia dependency. Now it's Nvidia moving up the stack to control the open-model distribution layer itself. For whoever leads AI adoption at your company, the exercise is simple: for every critical piece of your AI stack, ask how many acquisitions or contract disputes it would take to break it.