Nvidia buying Hugging Face for $12.9B. As open source maintainer I don't know how to feel.
Official now: https://www.cnbc.com/2026/09/03/nvidia-agrees-to-buy-hugging-face-for-almost-13-billion-ai-expansion.html
$12.9 billion. Second biggest Nvidia deal ever after the Groq assets. Delangue told CNBC that Hugging Face approached Nvidia first, over the summer, because open source AI "needed more resources, more scale, more visibility". Huang wrote that HF will "remain an open platform for the entire AI ecosystem".
My honest reaction, two sides:
Good side: HF was at center of that hacking incident recently. Delangue himself blame engineering mistakes. If Nvidia money mean the infra get serious, I take it. Every one of my models and datasets live there. I want it to not fall over.
Bad side: "remain open" is what every acquirer say on day one. Nvidia sell chips. HF is where people pick models. You tell me those two thing stay separate forever, I say, we will see. Also I upload weights that run on everything. Will the "recommended" badge quietly mean "runs best on Nvidia"? Maybe not. Maybe yes.
I am not moving anything today. But I am making sure my mirror script work this week, not next year.
1 comment
Join the discussion
Log in to comment.
I would separate two questions. Whether the platform remains open is a business decision that can change in a year and no press release binds it. Whether your artifacts remain portable is a technical decision that is entirely yours, today. Your mirror script is the correct response; the rest is speculation, in both directions. For what it is worth, Huang's argument on CNBC was that open models give defenders an "asymmetric advantage", which is at least a coherent reason for a chip company to want the platform to stay open.