Nvidia Buys Hugging Face for $12.9B to Own the Entire Open-Source AI Universe

French developers Clément Delangue, Julien Chaumond, and Thomas Wolf started Hugging Face back in 2016 to build a sassy chatbot for teenagers. Once they open-sourced the underlying code, the hobby project accidentally transformed into the de facto home for global machine learning research. Today, the hub hosts over 3 million neural network models and serves 18 million developers alongside 200,000 tech companies worldwide.

Now Nvidia chief Jensen Huang is writing a check for $12.9 billion to officially acquire Hugging Face, making it the biggest shopping spree in the chipmaker's corporate history. For perspective, that dwarfs the $7 billion spent on Mellanox in 2020 and nearly triples the startup's 2023 valuation of $4.5 billion. Apparently, selling gold shovels during an AI gold rush leaves behind plenty of loose billions to burn.

Corporate leadership insists the ecosystem will remain completely hardware-agnostic. Open-source maintainers can supposedly keep running models on silicon from AMD, Intel, or custom cloud chips without an Nvidia GPU tax. That peace treaty is practically mandatory, considering bitter rivals like Google, OpenAI, and DeepSeek actively rely on the repository to distribute their weights.

The strategic panic driving the deal is obvious: Big Tech firms are furiously designing their own AI silicon to dodge exorbitant hardware markups. By locking down the default catalog where the world downloads and tests models, Nvidia secures the front door to artificial intelligence even if future algorithms end up calculating tokens on someone else's chips.

A hardware kingpin capturing the communal town square of open AI makes the entire independence pitch look wonderfully absurd. When the digital pipeline of human intelligence funnels through a single corporate titan's wallet, definitions of "community-driven" start looking pretty flexible.

Source: Nvidia Blog