Nvidia has agreed to acquire Hugging Face, the developer hub for open AI models, for roughly $12.9 billion, its second-largest purchase ever. The deal reaches past chips toward the layer where AI is built and shared, a hedge as Nvidia's biggest customers design their own silicon and open-weight models gain ground across the enterprise.

[For more news, click here]

A year ago, Hugging Face turned Nvidia down when they offered roughly $500 million for a stake, a figure that valued the developer platform at about $7 billion, and the startup declined, according to the Financial Times. This week, Nvidia said it would buy Hugging Face outright for about $12.93 billion, nearly double that earlier implied valuation and the second-largest acquisition in the company's history.

Hugging Face is not a household name outside engineering teams, yet it sits close to the center of how modern AI gets made. More than 18 million developers, researchers, and creators use the platform to share upward of 3 million models, and more than 200,000 companies rely on it to discover and deploy AI, according to figures Nvidia cited. Under the terms, Nvidia will pay around $11.9 billion to Hugging Face investors and set aside an equity-based retention program worth up to $1 billion for employees who join the company. The transaction is expected to close in the first half of 2027, pending regulatory approval.

Why a Chipmaker wants the Software Layer

Nvidia controls the vast majority of the market for AI accelerators, and demand for its processors remains enormous. However, Meta, OpenAI, Microsoft, and Anthropic are each developing their own AI chips to cut reliance on Nvidia's costly and supply-constrained hardware. Should that effort succeed over the next several years, the pricing power that made Nvidia one of the most valuable companies on earth could soften.

Owning Hugging Face gives Nvidia a foothold that does not depend on shipping another graphics processor. Investors who follow the company read the move as a step from being a supplier of chips toward being a broader AI platform, positioning the business for a future in which hardware margins may compress even as AI adoption keeps climbing. In essence, it is the difference between selling the equipment and owning the marketplace where the work happens.

The Case for Open Models

Nvidia is also planting a flag in one of the defining arguments in artificial intelligence, the divide between open and closed systems. Closed models from companies such as OpenAI and Anthropic keep their inner workings private and reachable mainly through paid interfaces. Open-weight models, the kind hosted on Hugging Face, let companies download the parameters that govern how a model behaves and run them on their own infrastructure, customizing and inspecting them freely. Many technology leaders believe that keeping models in-house also makes proprietary and sensitive data more secure.

Chief Executive Jensen Huang has argued for the open approach for some time. “Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty,” he wrote about the acquisition, adding that “they allow every developer, startup, university, industry and country to build with, customize and benefit from AI.”

Demand for open-weight models has surged as businesses balk at the steep cost of deploying proprietary systems, and Chinese labs including DeepSeek and Z.ai have emerged as serious players, releasing open models that rival the strongest American systems on tasks from coding to reasoning. For Nvidia, sitting atop the platform where those models are distributed offers something beyond revenue, it provides a direct line of sight into what developers are building, insight and data that could help the company keep pace with leading labs in the United States and China.

Can an Open Platform Stay Neutral

The purchase raises a harder question that Nvidia will have to answer for years. Hugging Face has thrived in part because it is seen as neutral ground, a place where models built for any chip, from any company, can live side by side. Placing it inside the dominant maker of AI hardware complicates that reputation. Huang moved quickly to address the concern, saying that “Hugging Face will remain an open platform for the entire AI ecosystem” and that Nvidia's chips would not be required to build on or deploy through it.

Whether developers and rival hardware makers accept that assurance is another matter, and regulators may weigh in as well. A deal that hands the leading AI chip company control of the field's main model repository invites scrutiny over competition, especially as governments grow more attentive to concentration across the AI supply chain. Nvidia has spent heavily to keep its footprint wide, and this acquisition extends that reach into software and community in a way that antitrust authorities in the United States and abroad are likely to examine before the deal can close.

For American business leaders, the deal sharpens a choice already sitting on many technology roadmaps, whether to build on closed models that are simple to adopt but expensive at scale, or on open-weight systems that demand more engineering yet keep data in-house and costs under control. Nvidia is signaling that it expects the open path to widen through its spend of $13 billion to be in the centre of it.

The company still earns its money selling the most sought-after chips in the world, and that will not change soon. But acquiring the platform where AI is discovered, tested, and shared, Nvidia is preparing for a market in which its influence rests not only on the silicon it ships, but on the ecosystem it owns.

Related Articles

•  Why Fortinet Is Turning to Intel to Build Its Next Cybersecurity Chip

•  VAST Data and CrowdStrike Extend Cybersecurity to Enterprise AI Data

•  ServiceNow and Aramco Digital Partner to Expand AI Across Enterprise Workflows