Now Playing
NVIDIA Acquires Hugging Face: The End of Neutral AI Infrastructure
Audio is available on Spreaker — see link below.
NVIDIA's $12.9B acquisition of Hugging Face signals the collapse of open-model independence as consolidation sweeps every layer of the AI stack. Today's briefing covers the infrastructure land-grab, ChatGPT's $1B ad business, Anthropic's physical AI play, and the cost collapse reshaping model economics.
Audio is available on Spreaker — see link below.
NVIDIA just bought Hugging Face for twelve point nine billion dollars. That's not a partnership.
Here's what the timing tells us. Hugging Face's annualized revenue jumped from one hundred million to one hundred fifty million in the first half of this year, and its subscriber base doubled.
Infrastructure consolidation isn't just happening at the model layer. SLB is acquiring Kelvion, a thermal management company, for four point one billion dollars.
While infrastructure consolidates, the application layer is building a second revenue engine. ChatGPT crossed one billion dollars in annualized ad revenue in two hundred days.
The cost picture is shifting fast. Qwen released a one hundred twenty-five billion parameter multimodal model at one-ninth the prior training cost, using a Mixture-of-Experts architecture with only six billion active parameters at inference time.
Thomson Reuters is making a different kind of bet. The company spent forty million dollars over two years building a custom legal AI on top of Alibaba's Qwen, trained on decades of proprietary legal content.
The signal across all of this is consolidation at every layer: chips, cooling, model distribution, and now physical AI deployment. The measured read is that the independent middle is being squeezed.
Chapter summary auto-generated from the verified script. Listen to the full episode for the complete content.