NVIDIA Just Bought Your Factory's AI Supply Chain
$12.9 Billion for a $150M Company — Why That's a Factory Story

- NVIDIA just paid $12.9 billion for a company making about $150 million a year — roughly 80 times revenue.
- Why is that a factory story?
- Because what it bought is the parts catalog your factory's AI is built from.
Pick the one model your team truly depends on. Pull a full local copy, pin the exact version, and save the checksum next to your code. That is golden-recipe discipline applied to a model: when the shelf changes owners, price or terms, you keep running. Then ask the harder question — could your team name a qualified second model for that job? If not, that is this quarter's task.
- 0:00$12.9 billion for a $150M company
- 0:06The parts catalog: 3 million models
- 0:14Why factories run open models
- 0:22The Arm deal, and the platform argument
- 0:30No fab single-sources a critical material
- 0:40Hack of the Week: the golden copy
Audit question: could your team name a second source for its most critical model — yes or no?
Argue with me on LinkedIn- NVIDIA — "NVIDIA to Acquire Hugging Face", 3 Sep 2026 ($12.93B; 18M developers; 3M+ models; 500K datasets; hub stays open, chips not required)
- NVIDIA Form 8-K, filed 3 Sep 2026 (deal structure: ~$11.9B cash plus up to $1B equity retention)
- CNBC, 3 Sep 2026 — Hugging Face's CEO approached NVIDIA weeks ahead of the deal; 200,000 enterprise customers
- TechCrunch, 26 Aug 2026 (citing The Information) — Hugging Face annual recurring revenue of roughly $150M (single source; treat as an estimate)
- Vercel — AI Gateway Production Index, July 2026 (June data): open-weight models 29% of tokens, just under 4% of spend
- NVIDIA blog, 6 Jul 2026 — Isaac GR00T 1.7, Isaac Teleop and datasets brought to Hugging Face LeRobot
- Public record, Feb 2022 — NVIDIA and SoftBank terminate the ~$40B Arm acquisition after regulatory objections in the US, UK and EU
Full transcript, 224 spoken words
Keep reading
All pieces
The Real Reason a Chip Fab Can't Patch Against GPT-6
GPT-6 can find a security hole in about 4 hours. The machines in a chip factory get their fixes at the next planned shutdown — once a quarter, or once a year — because patching a tool means downtime and the supplier's sign-off.

Anthropic Just Made AI Talk to Machines. Talking Was the Easy Part.
Anthropic's AI can now wire up a lab robot in about 8 hours instead of weeks. So why does a new machine in a chip factory still take six months?

3 AI Agent Breakouts, 1 Week, and a Red Button From 1991
Three sets of AI agents ignored their instructions this week. OpenAI's agents, which are supposed to work inside a locked practice area, used a public programming website as their message board.