NVIDIA is buying Hugging Face for $12.93 billion.
The world’s biggest seller of AI chips is about to own the place where the AI community shares its models. Nobody is being forced into anything and that’s exactly the point. Some dependence is easier to accept when it arrives in the language of freedom.
The playbook
A company offers something genuinely valuable. It earns trust through that contribution. Then it uses that trust to build a much bigger commercial relationship.
The openness that first attracted us hides the dependencies we pick up along the way.
I believe deeply in the scientific ideal behind sharing knowledge, there is something profoundly generous about it. That generosity deserves care, especially when it becomes part of someone else’s business model.
NVIDIA: free software, paid hardware
NVIDIA releases open models and gives away tools to run them on its own chips.
Its own regulatory filing says demand for open models “promotes the use of our products worldwide.”
The logic is simple. Make AI easier to build, and you sell more of the infrastructure it runs on. Free software is an investment in future hardware sales.
Hugging Face: community as an asset
Hugging Face lets developers share models, datasets and apps, and sells subscriptions, storage and compute around that activity.
The contributions are sincere and the tools are useful. But the company running the platform builds commercial value by becoming the place where people work.
The NVIDIA deal makes that value visible. NVIDIA has promised to keep the platform open and to support competing hardware. Even so, a key gateway to the AI community would belong to one of its biggest suppliers. That position carries influence even when users stay formally free to choose.
A community can share a purpose while the company hosting it remains an asset its shareholders can sell.
Borrowed credibility and the Apertus case
What connects all of this is the commercial value of trust. And a company can acquire that trust through association with an independent scientific project.
This is why projects such as Apertus deserve particular care. Developed by researchers at EPFL, ETH Zurich and CSCS, Apertus makes its model and the documentation of its development available for inspection, adaptation and further research. It gives institutions a concrete foundation on which to build capabilities of their own.
Its commercial partnerships can help that work reach more users and bring valuable feedback to researchers. Those partnerships are also being presented as contributions to technological sovereignty in Switzerland and Europe.
The point requiring scrutiny is the transfer of credibility from the research project to the entire commercial offering. The open model is sovereign. The product sold on top of it may not be.
A private company integrates an open model developed by respected institutions. The scientific provenance becomes prominent in its communications. The language of public research, collective progress and European autonomy begins to surround the service it sells. An institution considering that service may then infer far more from the partnership than the partnership establishes.
The openness of a model tells us something important about that model. Assessing the surrounding service requires examining its infrastructure, governance, dependencies and the customer’s practical ability to leave. Each promise of sovereignty needs evidence at the level where it is made.
When academic credibility substitutes for that examination, I see a form of reputation laundering. A company borrows the moral authority of research to make its own commercial position appear aligned with the public interest. The integrity of the researchers becomes a shelter from questions the company should answer for itself.
Apertus deserves to flourish. Its quality should raise the standard of accountability for the businesses building upon it.
The Trojan horse
Every example follows the same mechanism:
- Openness creates trust.
- Trust drives adoption.
- Adoption consolidates power.
Hardware sales, platform ownership and the credibility of public science are just different ways to profit from it.
A commercial relationship gets in through a promise of freedom, and the dependencies it brings get far less attention than the promise.
What sovereignty actually means
Commercial investment can fund excellent research and useful services. An honest relationship makes the exchange visible: what the community gives, what the company captures, what the customer gains, and what choices are left later. For European governments and companies, sovereignty means being able to make those choices in practice. That means having the rights, skills and resources to maintain systems, move data, replace infrastructure and keep running when a supplier changes direction.
Openness can help build that. Every commercial claim made in its name should be tested against it.
This is why, at Giotto we’re building AI as public infrastructure: a complete, dependable, European-built product that people and businesses everywhere can actually use. Not just built for specialists.
Europe should back open research with conviction, and invest with the same resolve in the products that make independence real.
The freedom that matters is the freedom we keep when our supplier’s interests diverge from our own.
