NVIDIA acquires Hugging Face: $12.9B open platform gamble

NVIDIA is acquiring Hugging Face for $12,930,300,000, according to NVIDIA Blog. The deal puts the largest open-model hub under the largest AI compute vendor.

StartupHub data

Companies working on this

Profiles of the companies named in this story, with funding and a one-liner from our database.

Groq

$1.0B

Groq develops a high-performance AI inference chip and compiler for ultra-low latency AI applications.

Cerebras Systems

$23.0B

Industry-leading AI infrastructure for ultra-fast AI inference and training powered by Wafer-Scale Engines.

Etched

$21.0B

Etched is building the hardware for superintelligence by burning the transformer architecture into their chips.

Hugging Face

$4.5B

Hugging Face is the leading AI community and platform for machine learning collaboration, enabling developers to build, share, and deploy models, datasets, and applications.

More than 18 million developers share over 3 million models, 500,000 datasets and 1 million applications on the platform. More than 200,000 companies use it to find and deploy AI.

What NVIDIA says will not change

Hugging Face will stay an open platform for the whole ecosystem. Developers can pick the models, frameworks, clouds and inference providers they want.

NVIDIA compute won't be required to build or deploy through Hugging Face. The platform will keep supporting multi-cloud and multi-accelerator development.

That promise carries weight because NVIDIA is already the largest contributor of open models and data to Hugging Face, with over 500 models and more than 250 datasets released there.

Why distribution beats silicon alone

The price secures distribution, not just weights. Hugging Face is where models get found, evaluated and deployed, and NVIDIA wants its infrastructure to improve reliability, safety and inference while preserving that open funnel.

For chip challengers, the funnel is the moat. If the default place to try a model runs best on NVIDIA, alternatives have to fight harder for adoption.

How inference hardware rivals stand now

StartupHub.ai data shows the inference chip field is crowded and heavily funded, which makes a distribution chokepoint more consequential for challengers like Groq, Cerebras, SambaNova and Etched.

Groq raised $750 million at a $6.9 billion valuation, a sign of investor interest in inference. SambaNova closed a $1 billion financing at an $11 billion valuation.

Cerebras raised $1 billion at a $23 billion valuation, nearly tripling its prior mark. Etched raised $300 million at a $10.3 billion valuation for its inference-optimized chip.

Those valuations depend on winning inference workloads away from NVIDIA. Control of the repository where 200,000 companies shop for models tilts discovery, benchmarking and one-click deployment toward NVIDIA-optimized paths, even if the platform stays technically open.

The pattern is familiar. A year ago, Hugging Face engineers were already automating their own jobs with AI agents on the platform itself, a sign of how central the hub has become to workflow experimentation (/ai-news/artificial-intelligence/2026/hugging-face-engineer-automates-job-with-ai-agents). Owning that workflow matters more than owning one more model family.

What to watch next

Watch whether Hugging Face keeps neutral defaults for model ranking, evaluation and inference routing. Any drift toward NVIDIA-preferred kernels or clouds will push rivals to build parallel hubs or pay for placement.

The deal will also test antitrust scrutiny of AI stack integration. At $12.93 billion, this isn't a talent acqui-hire, it's infrastructure acquisition, and regulators will read it that way.

NVIDIA just bought the front door to open AI. The question is how much rent it charges to walk through it.