This is honestly completely fine, and I don’t understand how people can think this is a conflict of interest? NVIDIA has the most ridiculous monopoly on GPU’s

This is honestly completely fine, and I don’t understand how people can think this is a conflict of interest? NVIDIA has the most ridiculous monopoly on GPU’s right now. Their RTX4090 is really fast and would be great for clusters but NVIDIA software driver locked it from sending data at high speed to other GPU’s that prevents it from being used in clusters. This is not a one off thing, NVIDIA regularly does stuff like this to piss off consumers but we are at the mercy of buying NVIDIA chips and making it a trillion dollar company. There were reports that the H100 (latest and fastest known AI chip from NVIDIA) costs only 3000$ to make, while it is sold for 30,000$. Everyone is desperate to add some competition to this industry. If I were CEO of OpenAI even I would be desperate to create another chip to compete with NVIDIA and end this resource reliance. It’s not obvious that this effort should be started within OpenAI. It may turn out to be a distraction for a startup doing really well, many large companies with much more money to spend have spent huge resources building their own chips (TPU’s, Tesla DOJO, Intel Arc) and still NVIDIA is Uber dominant. I think it makes complete sense to start it outside of OpenAI. It would be a conflict of interest, only if: 1. The chip is of poor quality and Sam Altman forces OpenAI to buy it anyways to enrich himself. I think the possibility of this happening is close to 0, everything I’ve seen and heard of Sam doesn’t make it seem like he’s someone who would do something like this. That said we are right now 2-3 years away optimistically from such a chip being developed, it’s too early to level any conflict of interest allegations! As am ML practitioner I for one strongly welcome Sam Altmans Chip startup, just as I welcome MS building their own chip, Teslas Dojo, Google TPUs and anyone else who wants to challenge NVIDIA