Why the Best AI Companies Are Open
Hugging Face's quiet $4.5B bet on openness reveals a strategic principle most executives have wrong: the platform that captures 1% of a 1000x larger pie still wins.
In May 2022, a small company in Paris closed a Series C round valuing it at $2 billion. The product was free. Anyone could download it. Anyone could fork it. By August 2023, the same company raised another $235 million at a $4.5 billion valuation. By August 2024, it was profitable.
The company is Hugging Face. The product is open-source AI infrastructure. And the strategic principle behind its rise is one that most executives still get wrong.
The 1% That Beats 100%
Clément Delangue, Hugging Face’s CEO, explained the logic in a single sentence: “With open source, you don’t have to capture 100% of the value you create. The platform creates 1000x more value being open than if it were closed. Even capturing 1% of this value has the potential to be more successful than the traditional proprietary approach.”
This is not a rounding error. It’s a different theory of how value compounds.
A closed platform — OpenAI, Anthropic, the major proprietary players — has to capture as much value as possible from each user, because every dollar leaving the platform is a dollar lost. The math gets demanding fast: a closed platform with 1 million users at $20/month makes $240M/year, but it has to defend that revenue against every competitor that arrives.
An open platform with 100 million users captures only a small fraction in revenue, but it captures something more durable: the ecosystem. Every model published, every dataset shared, every demo built strengthens the platform’s gravity. Closed competitors can’t replicate that, because the contributors are not employees.
By October 2024, Hugging Face hosted over three million models, datasets, and applications, with new models created every 10 seconds. No closed company can ship at that velocity, because no closed company has 24,000 contributors building for free.
What This Means for Your Strategy
If you’re an executive evaluating where to make AI investments, the open-vs-closed decision shows up at three levels:
Which platforms you build on. Closed APIs are convenient until they’re not. Pricing changes, models deprecate, terms shift. Open infrastructure — even if it costs more to operate — preserves optionality. Hugging Face’s enterprise customers (Bloomberg, Intel, Grammarly, Pfizer) didn’t pick it because it was cheapest. They picked it because they couldn’t be locked out.
Which models you fine-tune. Delangue’s observation in early 2023, repeated more forcefully by 2024: “Smaller, specialized models are better.” The future of enterprise AI isn’t one model to rule them all. It’s hundreds of fine-tuned variants, each optimized for a specific workflow. That kind of specialization is impossible on closed platforms — you can’t fine-tune what you can’t access.
Which knowledge you capture. When your AI workflows live on closed platforms, the institutional knowledge embedded in your prompts, fine-tunes, and evaluations belongs to the vendor. When they live on open infrastructure, that knowledge is a strategic asset you own. Most organizations don’t notice the difference until they try to switch vendors.
The Counterargument
The case for closed platforms is real: they’re easier to start with, they handle the operational complexity, they ship the most capable frontier models first. For a team of five trying to validate an idea, calling an API is the right answer.
The case turns once you’ve validated. Past that point, every month you stay on closed infrastructure is a month you’re renting your moat from someone else.
What Hugging Face Knew Early
The decision to open-source Transformers in April 2018 was not, at the time, a strategic masterstroke. Delangue described it as “a way to pay it forward for the community.” The team didn’t predict the network effects. They believed in openness as an ethical posture, and the strategic returns followed.
The lesson for executives is uncomfortable: the highest-leverage AI strategy decisions don’t always look like strategy. They look like values choices that compound for a decade. Hugging Face decided early that the platform would always be free for the contributors who made it valuable. Eight years later, that decision is what makes a $4.5B valuation defensible.
The companies that win the next decade of AI won’t be the ones that locked the most value behind paywalls. They’ll be the ones that built the platforms others built on top of.