Guozhen AIGlobal AI field notes and model intelligence

Realtime AI News

NVIDIA reportedly agrees to acquire Hugging Face for $12.9 billion

The Information reports that NVIDIA has agreed to acquire Hugging Face for $12.9 billion, a deal that is not yet formally finalized, after Microsoft's own talks with the open-source AI platform fell through. The platform hosts more than 3 million models and 1 million datasets, making the acquisition a bid to anchor NVIDIA at the center of open-source AI.

Published

NVIDIA has agreed to acquire Hugging Face for $12.9 billion, The Information reports, according to coverage from Chinese outlet QbitAI — a deal that has not yet reached a formal agreement and still carries uncertainty. Microsoft had also met with Hugging Face about an acquisition, but those talks failed, the reports say. Rumors of a major buyer circulated over the past weekend before the purchaser's identity surfaced. TechCrunch had separately reported that NVIDIA was closing in on the acquisition. NVIDIA already knows the company well: it was an investor in Hugging Face's $235 million Series D in 2023, a round led by Salesforce Ventures with Google, Amazon, Intel, and AMD participating, at a post-money valuation of $4.5 billion. The reported $12.9 billion price would roughly triple that valuation in about three years. The deal also has a twist: according to the Financial Times, Hugging Face turned down a $500 million investment from NVIDIA late last year at a $7 billion valuation, saying it did not want a dominant shareholder steering company decisions — only to now reportedly sell outright. Two months ago, the company's annual recurring revenue was only around $100 million. Founded in 2016 as a chatbot app, Hugging Face pivoted into open-source model hosting and became known as the GitHub of AI, hosting more than 3 million open models and more than 1 million datasets and serving as the de facto home for open-source AI teams worldwide. For NVIDIA, the logic is strategic: closed-source labs including OpenAI, Google, Amazon, and Anthropic are building their own chips to reduce dependence on NVIDIA. A thriving open-source ecosystem offers an alternative to closed models and keeps training and inference workloads on NVIDIA hardware, the same playbook Google used with free Android to pull ecosystems onto its hardware. Cloud and balance-sheet math also matter: The Information notes NVIDIA scaled back DGX Cloud roughly a year ago, and Hugging Face already helps developers run models on rented compute. With NVIDIA committed to absorbing hundreds of billions in cloud contract costs for customers, owning Hugging Face would let it resell unused capacity to the platform's customers — and its latest quarterly revenue of $96.2 billion makes the price tag a manageable one. What to watch: whether the deal reaches a formal agreement and clears regulatory review, and how the open-source community reacts to the platform changing hands — a move that would cement NVIDIA's position as the leading open-source AI company in the U.S.

Why it matters

NVIDIA would cement control over the open-source AI distribution layer, strengthening its GPU ecosystem against chip-building rivals and giving it an outlet for unused cloud capacity.

NVIDIAHugging FaceM&A
Back to AI Daily

Nearby Updates

All

08/28, 10:00

OpenAI and Thailand's MHESI launch eight-week accelerator for 10 local AI startups

OpenAI and Thailand's Ministry of Higher Education, Science, Research and Innovation (MHESI) have launched an eight-week accelerator program for local AI startups. The first cohort of 10 companies spans health, wellness, and education, with a focus on turning AI prototypes into trusted products.

08/28, 11:03

RayNeo iO smart glasses off to a hot start: 34g all-day AI wearable from ¥1,996

RayNeo, a leading consumer AR brand, published first-sale results for its iO smart glasses on August 27, reporting a strong start for the 34-gram, all-day AI wearable priced from 1,996 yuan. The company says the hot sales are accelerating the mass adoption of smart glasses.

08/28, 12:06

Zhipu launches GLM-5.3-Flash with SenseTime domestic compute support

Zhipu's lightweight GLM-5.3-Flash model has officially launched, with SenseTime's computing platform providing domestic compute support. Domestic heterogeneous computing is beginning to carry scaled serving for a leading Chinese foundation model, a step toward making frontier AI widely affordable.

08/28, 12:09

SenseTime and HiDream.ai run video-generation workloads on domestic chips at scale

SenseTime's big-device platform and HiDream.ai have completed a full domestic-compute adaptation for video generation, reaching a 93% multi-card parallel speedup for DiT models on domestic chips. HiDream's image models now support large-scale online traffic and short-video creation, with zero-cost migration across more than 10 heterogeneous chip types.