NVIDIA is in negotiations to acquire Hugging Face, a move that would extend its influence into AI model hosting and distribution, raising questions about industry neutrality and control over the AI stack amid ongoing hardware and software innovations.
NVIDIA is reportedly in talks to buy Hugging Face for $12.9 billion, a move that would extend its influence far beyond chips and into one of the AI industry’s most important software and distribution layers. According to The Information, the companies had not confirmed a deal at the time of writing, so the transaction should still be treated as a report rather than a completed acquisition. TechCrunch has also reported the discussions. Hugging Face is widely used to host model weights, datasets, demos and developer libraries, making it a central point of discovery for open AI systems.
If completed, the purchase would give NVIDIA a stronger grip on the full AI stack. It already dominates accelerator hardware and the CUDA software ecosystem; adding Hugging Face would bring control over the place where many developers find, compare and download models. That raises immediate questions about neutrality. Hugging Face currently works across NVIDIA’s rivals, including AMD, Intel and major cloud providers, and those companies would have to judge whether they were comfortable continuing to build on a platform owned by their biggest competitor.
The strategic logic is broader than revenue. Open models remain a useful counterweight to closed AI labs, and the more those models are distributed through a platform that sits at the centre of developer attention, the more valuable that platform becomes. NVIDIA would also gain more leverage over the ecosystem in which models are evaluated and deployed, at a time when the industry is shifting from isolated benchmarks to real-world inference and production use.
That wider race is also visible in OpenAI’s newly disclosed Jalapeño, its first custom inference chip. Tom’s Hardware reported that the processor was developed with Broadcom in an unusually fast nine-month cycle and is designed to improve performance per watt for modern language models and agentic workloads. OpenAI says the chip is aimed at inference rather than training, underlining how much of the industry’s spending is now moving towards serving models efficiently at scale. The parallel is clear: while OpenAI is trying to reduce dependence on outside hardware, NVIDIA is seeking deeper control over the software layer that shapes demand for that hardware.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
