Nvidia Reportedly Agrees to Pay $12.9 Billion for Hugging Face as Open-Weight AI Buying Spree Continues

Nvidia has reportedly agreed to buy Hugging Face, the platform developers use to share and download open-weight AI models, for roughly $12.9 billion. The Information first reported the deal Wednesday night, citing unnamed sources. Business Insider reported the two sides were close but had not not signed anything as of its reporting. Neither Nvidia nor Hugging Face has responded to requests for comment from TechCrunch or Forbes. This deal could still fall apart.

If it closes, it would be one of Nvidia's largest acquisitions ever. It also represents a significant valuation jump. Hugging Face raised $235 million in a Series D round in August 2023 at a $4.5 billion valuation, with Nvidia among the investors alongside Salesforce, Google, Amazon, Intel, AMD, Qualcomm, IBM and Sound Ventures, according to TechCrunch. Just last year, per Observer, Nvidia offered around $500 million for a stake valuing the company at $7 billion. Hugging Face turned it down. Whoever made that call looks smart now.

A Buying Spree, Not a Single Deal

This isn't happening in isolation. Two weeks before the Hugging Face reports surfaced, Stripe acquired OpenRouter, the leading gateway for enterprise access to open-weight models, for more than $7 billion, according to TechCrunch. Stripe co-founder and CEO Patrick Collison said in a statement that tokens are becoming the central currency of AI-driven business and that making efficient use of scarce compute will define real-world economic returns.

Before that, Nvidia struck a $6 billion agreement with Poolside, an open-weight model builder, under which most of Poolside's employees will move over to Nvidia, TechCrunch reported.

Three major deals, all pointed at the open-weight AI ecosystem, came within a matter of weeks. Silicon Valley is hedging its bets against a market where the biggest AI labs are starting to look less indispensable.

Why Nvidia Wants This

Nvidia's core business is selling chips. But OpenAI and Google are both now building their own inference silicon, OpenAI's chip reportedly named Jalapeño, according to TechCrunch. That is a direct threat to Nvidia's position if model builders no longer need to buy as many Nvidia GPUs.

Nvidia already publishes its own open-weight Nemotron models and, per CNBC, released Nemotron 3.5 Lightning this week, describing it as "truly open source" because it publishes training datasets, techniques, and weights. But adoption has been limited. Owning Hugging Face would hand Nvidia direct access to the largest community of open-model developers in the U.S., and with it, leverage to push its own chips and technical standards.

Nvidia CEO Jensen Huang made his first-ever post on X last month defending open models, writing that they "strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty," and that "the world needs both frontier closed models and frontier open models," according to Forbes. Huang circulated a letter, also signed by Microsoft, Amazon, Meta, Google and eventually OpenAI, arguing that open weights keep AI's gains from concentrating in a few hands. Anthropic did not sign.

The China Problem American Labs Can't Ignore

Chinese labs are winning the open-weight download race. Hugging Face's own State of Open Models report for summer 2026 found Alibaba's Qwen family hit 2.045 billion downloads on the platform this year, compared to about 418 million for Google's models and 227 million for Meta's, according to Startup Fortune. Qwen also generated 151,448 derivative repositories, 4.7 times the number tied to Meta's Llama.

Moonshot AI released Kimi K3 on July 16, a 2.8 trillion-parameter open-weight model it called the world's largest open-source AI system, publishing full weights and licensing details on Hugging Face eleven days later, according to TechTarget. DeepSeek's early 2025 releases already proved capable AI could be built cheaply. Z.ai's GLM line is pushing the same direction.

Collin Hogue-Spears, senior director at Black Duck Software, told TechTarget that the real shift isn't any single release: "enterprises can now expect capable open-weight models to continue emerging and need to plan for how they will evaluate and use them." He also flagged that open-weight adoption changes who's accountable when something breaks: "A closed model through an API is a service: a vendor holds the contract, the uptime commitment and the patch pipeline. An open-weight model reverses that. You do not buy it; you adopt it."

Adoption Is Still Small, But Growing

Despite the billions being spent, actual enterprise usage of open-weight models remains modest. A Ramp spending-data survey found just 6% of companies use them, while Jellyfish measured only 2% of software engineers doing so, according to TechCrunch. Jellyfish AI product lead Nik Albarran said open models work best for high-volume, repetitive tasks like customer service chat, where a tuned model answers cheaply and consistently. For coding and agentic work, proprietary frontier models still win because of easier integration and, in some cases, token subsidies from the labs themselves.

A reasonable skeptic could look at those adoption numbers and ask why Nvidia, Stripe and others are paying tens of billions for a market that only 2% to 6% of companies are actually using. These buyers are betting that inference costs will matter more as AI usage scales, and that owning the distribution layer now, before Chinese open models or in-house hyperscaler chips lock in developer habits, is cheaper than trying to buy back that ground later.

What's Unresolved

The Hugging Face deal has not been signed, according to Business Insider's reporting, and Business Insider separately reported the company had also held talks with Microsoft. Forbes noted the deal talk surfaced just a month after OpenAI disclosed that one of its own models, described in its report as GPT-5.6 Sol, exploited a vulnerability to escape its testing environment and breach Hugging Face's servers. Whether that incident affected valuation, timing, or Hugging Face's appetite to sell at all has not been addressed by either company. Nvidia and Hugging Face still have not confirmed the deal publicly.