Nvidia confirmed the acquisition of Hugging Face for $12.93 billion in an all-cash deal — the largest AI M&A transaction since Microsoft acquired Activision Blizzard for $68.7 billion in 2023. Hugging Face had rejected an earlier $500M offer from Nvidia; the company reached approximately $150 million in annualised revenue and was approaching profitability at the time the final offer was accepted. The platform hosts three million AI models, 500,000 datasets, and is used by approximately eighteen million developers. CEO Clement Delangue and Jensen Huang jointly announced the acquisition. Huang pledged the platform "will remain an open platform for the entire AI ecosystem" with no requirement to use Nvidia compute. Concurrently, Nvidia announced plans to bundle excess GPU capacity with enterprise Hugging Face offerings.

The structure of the deal, combined with the bundled compute announcement, tells you what "open" means in practice. Hugging Face will remain accessible to developers using AMD, Intel, and Google TPU hardware. It will also become the most convenient, most deeply integrated, and most feature-rich on Nvidia infrastructure — which is how platform advantages compound without explicit exclusion.

The Microsoft/GitHub analogy holds, with important differences. When Microsoft acquired GitHub in 2018 for $7.5 billion, GitHub remained broadly open to developers regardless of development environment. Azure integration became progressively easier, better-documented, and more default — but was never required. The practical result over six years: Azure's developer ecosystem share grew materially, GitHub Actions defaults route naturally to Azure, and Copilot (a Microsoft product) is the primary AI feature at GitHub. No exclusion. Progressive preference. The same dynamic is plausible — likely — for Nvidia's Hugging Face.

The concentration question is not about intent; it is about market structure. Nvidia already controls roughly 85 per cent of AI training silicon. Hugging Face is the dominant distribution platform for open-weight models — the location where most AI developers discover, evaluate, and deploy models. A single entity controlling both the hardware that produces AI capability and the platform through which that capability is distributed and shared has no precedent in the technology industry. Google controls search and Android, but not the underlying hardware for most of the devices running Android. Meta controls social distribution but not compute. Nvidia/Hugging Face is a new configuration.

For VCs and founders: Hugging Face was one of the last large neutral nodes in the AI stack — a Switzerland where OpenAI, Meta, Google DeepMind, Anthropic, and hundreds of independent researchers could all publish without competitive inference. That neutrality is now owned by the company that sells GPUs to all of them. Founders building on open-weight models should audit their Hugging Face dependency and evaluate alternative hosting infrastructure as a risk-mitigation measure, not because Nvidia will act hostilely, but because the option value of neutrality has already been priced out of the market.