
So, in order to decipher all this (NVIDIA is buying Hugging Face for 13$B), we’ll need to answer three questions:
In the begging there was Github. Github became the de facto open source code repository. On the AI world, LLM (Large Language Model) is the most popular AI method, it’s not the only one, but services like ChatGPT & Claude, are LLM based.
Some LLMs are what’s called Open Weight (the equivalent of Open Source code). As such, these open weight LLMs are freely available to the public, for download & use. Where do you host/offer these LLM for public consumption? In a site called Hugging Face. Needless to say, after ChatGPT explosion, Hugging Face came along for the ride. Cool? Ok, let’s continue.
While OpenAI & Anthropic are one of the many NVIDIA’s customers, it’s in NVIDIA’s interest they’re not the only ones. In fact, by democratizing LLM consumption (Tokenization & Inference, using industry jargon, because, of course!), NVIDIA can guarantee a safe customer base, like it did with video games. Quick note: When graphic cards were all the rage, AI became a nice secondary use case for this type of hardware. This is the reason why NVIDIA has invested in AI for more than a decade! As a secondary use case for its products, the thing is, the use cases positions inverted. AI became first, gaming a distant second.
NVIDIA CEO supports the open models.
As tech traditions continue, Hugging Face it’s -apparently- not profitable. So why pay/offer such an enormous amount of money for a non-profitable business?
So there you have it, you’re now able to decipher AI business jargon 💪🏽