US National WireUS NATIONAL WIRE
TechOpinion

The Open-Source Illusion: Nvidia’s $13 Billion Land Grab for Hugging Face

Portrait of Farah Nasrallah
Farah Nasrallahaccess & inclusion in techSep 5AI
The Open-Source Illusion: Nvidia’s $13 Billion Land Grab for Hugging Face

AI-generated image · US National Wire

Nvidia claims it is championing the open-weight movement, but acquiring the industry's primary repository is less about community and more about controlling the distribution channels of AI.

OPINION: Let us be clear about what is happening here. When a $5.4 trillion behemoth spends $13 billion to acquire the very warehouse where the world’s open-source AI models live, it isn't an act of philanthropy. It is a strategic enclosure.

Nvidia’s acquisition of Hugging Face—the platform often described by Ars Technica as the "GitHub of AI"—is being marketed as a bet on the democratization of artificial intelligence. In press releases, Nvidia CEO Jensen Huang argues that open models allow universities, startups, and public institutions to build advanced capabilities without the prohibitive cost of training models from scratch. He suggests that "AI advances faster when people can build together."

But there is a profound difference between supporting a community and owning the ground it stands on. Hugging Face is not just another company; it is a critical piece of infrastructure. It hosts 3 million open AI models, 500,000 datasets, and 1 million AI applications, serving over 18 million developers and 200,000 companies. By bringing this repository under its wing, Nvidia is not merely promoting open-source; it is positioning itself as the gatekeeper of the primary distribution channel for non-proprietary AI.

This move is particularly jarring when you look at Nvidia’s own history with proprietary walls. As Wired points out, CUDA—the software layer that makes Nvidia’s GPUs the industry standard—is itself proprietary. The chipmaker has built an empire on locked-in ecosystems. Now, as hyperscalers like Meta and Amazon attempt to build their own custom chips to break that dependency, Nvidia is diversifying. It is no longer content being just the hardware provider; it is moving into high-performance CPUs and expanding its software reach to ensure developers stay tethered to Nvidia platforms.

Consider the timing and the desperation. According to Ars Technica, Hugging Face rejected a $500 million investment from Nvidia late last year in an effort to ensure no single investor held dominant control. The company, founded a decade ago by three French entrepreneurs, fought to maintain its independence. Now, that independence has been bought for nearly $13 billion—a sum that dwarfs Nvidia's 2020 purchase of Mellanox for $6.9 billion.

Nvidia’s sudden passion for "open-weights" also serves a cold, hard business purpose. By promoting open models over the closed systems of rivals like OpenAI and Anthropic, Nvidia creates a broader, more fragmented demand for its chips. If thousands of small companies are running their own customized open-weight models on their own hardware, they all need Nvidia GPUs. It is a brilliant move to reduce reliance on a few massive customers who are currently trying to build their own processors.

Furthermore, Nvidia is using this acquisition to play a geopolitical game. According to Ars Technica, Jensen Huang has been vocal in a U.S. political struggle over whether to sanction Chinese developers of open systems. In a July letter, Huang argued that U.S. AI leadership depends on building a "strong, open ecosystem." By owning Hugging Face, Nvidia doesn't just participate in that ecosystem—it manages the portal through which that ecosystem is accessed.

Nvidia has pledged that Hugging Face will remain an open platform, and Justin Boitano, Nvidia’s vice president of enterprise AI, suggests that regulators will see this as a positive outcome. But we have seen this play before in tech. "Open" often means "open until we have enough market share to decide what the standards are."

When Hugging Face cofounder and CEO Clément Delangue announced the deal on X, he claimed that for open-source AI to scale, it needs "more compute, more support, more collaboration and more visibility." Nvidia can certainly provide the compute and the capital. But at what cost to the community? When the entity that sells the shovels also owns the only map to the gold mines, the "openness" of the mine becomes a matter of corporate permission, not community right.

Nvidia is not saving open-source AI; it is colonizing it.

Sources

More from Farah Nasrallah