Untitled

This acquisition and the Stripe acquisition of OpenRouter are puzzling. Both HF and OR, from a technological standpoint, don't appear to have much IP that cannot be easily replicated. I wonder how much it's worth to Nvidia just to obtain a list of who's downloading which models and for what purpose. Similarly with the Stripe acquisition of OpenRouter. In the past, it would have been valuable to obtain an audience or market share or even a productive team. However, in this new age of the surveillance state ... telemetry, orchestration, and financial

Don't see this happening. Both Nemo and TensorRT from Nvidia are heavily invested in low precision formats and higher inference throughput as well as low memory use. It's more about what happens in relation to GGUF, MLX and non CUDA-native frameworks/formats.

I really, really doubt that, and I'm a bit surprised (not too surprised, HN has gotten really cynical) that this is the top comment. I think the future nvidia most doesn't want is a small number of closed labs that run away with it, and gain power to steer a large part of the hardware spending towards nvidia's competitors, as leverage in pricing negotiations for big hardware buys. Their ideal would seem to be lots of competition in the model space, driving lots of innovation in lots of areas, at low margins for the model makers, driving use and demand for hardware way up. They want the portion of the economy working on this to go up, and that's not going to happen if there's just a few companies that can afford to build models.

In that case, somebody will open a new platform to upload the quantized models and hf will lose market share

Too base cynicism. I'd be willing to bet you my $200 to your $100 that doesn't happen.