Nvidia is paying $12.9B for Hugging Face, a company doing $150M a year in revenue. That's roughly 86x revenue for a business whose core assets, open models and open datasets, are things anyone can copy for free.
We've talked about three vectors of defensibility in the AI era at Dynamo since 2016. Domain expertise. Proprietary data. Distribution, aka customer base. Hugging Face barely has the second one. Its models and datasets are open by design, which is why a whole cottage industry of "private fortress" alternatives exists just to give enterprises what Hugging Face can't. It doesn't have much of the first one either. It's horizontal infrastructure, closer to a public square than a specialist. What it does have is 18M developers, 3M models, and 200K companies showing up every day. That's the asset worth $13B.
Two years ago Nvidia offered $500M for a stake at a $7B valuation and got turned down. This week it paid nearly double that valuation to own the whole company outright. The technology hadn't moved much in that window. What moved is how much it's worth to sit at the front door developers already use, especially with OpenAI and Anthropic both building proprietary chips to cut Nvidia out of their stack.
Same pattern as every infrastructure cycle before this one. AWS won because it became the default place developers went to provision compute, not because it had the best hardware. GitHub followed the same script. It became the place code lived and reputations got built. Nobody picked it for being the strongest version control tool on the market. Distribution compounds once enough of a market's attention lands in one place, because leaving costs more than staying.
The part that doesn't resolve cleanly is the revenue gap. $150M against a $13B price tag means Nvidia isn't underwriting today's business. It's underwriting the chokepoint before someone else buys it, betting the monetization comes later. That's a defensive move made under pressure. It's a real gamble, not proof that distribution pays for itself.
For founders, the takeaway isn't to build a developer platform and hope someone acquires it. It's narrower than that. If your product sits at a point in the stack that drives recurring usage, being the default matters more than being the best. Data can be commoditized and domain expertise can be replicated, but the place people already show up and build habits around is much harder to dislodge.
Watch what Nvidia does with Hugging Face's roadmap over the next year. That's when we'll know if this was worth it.
Congrats to the Hugging Face team!