Nvidia Agrees to Buy Open Source Platform Hugging Face For $12.9 Billion

“Aug 26 (Reuters) - Nvidia has agreed to buy AI platform Hugging Face for $12.9 billion, The Information reported on Wednesday, citing a person with knowledge of the deal.
The chipmaker and New York-based Hugging Face, which hosts open-source large language models and datasets, did not immediately respond to Reuters requests for comment outside regular business hours.
Nvidia was among the backers, alongside Salesforce and Alphabet's Google, that backed Hugging Face in a $235 million funding round in 2023 that valued the company at $4.5 billion.

The Financial Times reported in January that Hugging Face last year rejected a $500 million investment offer from Nvidia that would have valued it at $7 billion.

Nvidia on Wednesday forecast a 70% jump in revenue next fiscal year, underscoring unabated demand for AI computing. Nvidia also said it has $18 billion committed to equity investments for the rest of 2027 fiscal year.
(Reporting by Carlos Méndez in Mexico City; Editing by Rashmi Aich and Subhranshu Sahu)“

Source: https://lufkindailynews.com/news_re...cle_befd777c-9a61-5feb-8723-289b1e68c1f5.html

is this Nvidia's answer to Deep Seek ?

Marees said:

is this Nvidia's answer to Deep Seek ?

Click to expand...

to a hundred deep seeks

Oh great =\

I don't see how this could be good. For 12.9 billion, they are going to want ROI, so I am interested in what that model (pun intended) they plan to use to recoup that.

Given that Nvidia really likes proprietary moats, this concerns me.

sc5mu93 said:

I don't see how this could be good.

Click to expand...

Nvidia is pushing for openSource and open competition on the LLM-software stack a lot, because the alternative tend to be Google running google models on google hardware, OpenAI running OpenAI on OpenAI hardware and so on.

That why they are almost the only open source LLM (not just openWeight) of any major player, they push neocloud that have zero special affiliation outside just running model for any clients.

Because and has long they can make money on the hardware, they will love an software side that can run on anything and they stay confident they will offer the best solution to run those.

As long hardware margin are high and that this world stay open that will continue to be their strategy, they will push for open LLM models, open software stacks they helped created and boost (TensorRT, triton, Nemo), independant clouds, local personal server and so on.

The cheapest the stack above it the more people will use, the more they will buy compute, the more open and the less Amazon model running on amazon chips, the better for the general run everything hardware solution they make.

One day, all the major integrator will have "good enough" inference hardware to consider ditching Nvidia we can imagine, some Nvidia will start competing with them, probably at first by becoming a seller of compute instead of hardware, making their own supercomputer they loan...

And probably start making their own LLM if all the best one do not run on Nvidia anymore and becoming a competitor there, if that happen, one think for sure they will protect themselve, plan and be ready for both level.

Wait nVidia is buying the company that OpenAI's AI "went rogue" and breached?

Oh for fuck's sake. Great, the company that never met a proprietary standard it didn't like from PhysX, GSync, RTX-related everything, DLSS, and more, (and for awhile forced its customers to not be able to use their own 'superior' GPUs along with the cheaper, equal-or-better performance, open FreeSync monitors despite no technical blockers, to just name one issue) is now going to be grabbing hold of the largest place that handles a lot of the LoRAs, other weights, model distribution etc..and generally allows access to open, self-hosted AI/LLM tech.. While there is the potential they'l just sit back and collect the check, but I'm waiting to see that anything that allows "Open, but also lets it work on non-Nvidia hardware" to end up somehow being hobbled, or DMCA'd or something else. I have zero trust in Nvidia to allow true openness to flourish as opposed to "Well its technically open, but it only REALLY works on our hardware or with our other tech". They've had multiple opportunities to push for true openness over the years and there isn't even a Linux driver that's at at parity with AMD's as of yet despite the massive finances NV has available.

Consider me extremely skeptical that this isn't going to be things that favor NV hardware or using NV created APIs or standards and if there's any serious competition elsewhere, be it hardware or software related, it won't be given the equal chance. its all well and good to sit on top and talk about openness when it only really drives people towards your ecosystem, as opposed to really give them the chance to make use of some stellar open tech that could actually let them hop out of the pool they control.

RanceJustice said:

While there is the potential they'l just sit back and collect the check,

Click to expand...

that seem impossible, this is a strategic acquisition for sure and HuggingFace make little revenue for the moment.

One thing is certain, HuggingFace cloud providing of LLM service and others will continue to run mostly on Nvidia hardware, but they will continue for a while to offer a vast array of solution and try to appear a bit neutral and their big strength in that space now is how much they appear multi-vendor and extremely open.

Here open source model is so well aligned with Nvidia business model that there is a chance that this is a giant injection in the open source ecosystem to counterbalance the closest weight and the non open source only open weight outthere.

I am sure lot of fear when microsoft bought Github, about none occurred and GitHUb still fully open to easily run your stuff on AWS or google cloud.

One big possible advantage here, it give Nvidia a deep and early knowledge to a lot of openSource projects... they will be able to optimise their driver-hardware for them and vice-versa it will continue to be super easy and natural for openSourcer to optimise on Nvidia.

sc5mu93 said:

I don't see how this could be good. For 12.9 billion, they are going to want ROI, so I am interested in what that model (pun intended) they plan to use to recoup that.

Given that Nvidia really likes proprietary moats, this concerns me.

Click to expand...

You own NVidia stock?

If not, Jensen probably has money oozing out all his pores and needs somewhere to put it all? Or just drunk with "power."