What would it take to compete with NVIDIA at this point?

Full episode:

But I just shorted 20k nvda shares because Coreweave, lol

Death by thousand cuts: major players started designing own AI chips, integrate into own clouds, and bundle them with all of the plethora of their other cloud services. It will be hard to compete with that whole bundle in a long run. E.g. situation like MS Teams vs Slack.

Succinctly put! Especially the software eco system @NVIDIADC invested in over many years and engineers hours. I don't want to put it this way but nVidia is almost InVincible.

The difficulty isn't all the HW stuff. All of those capabilities you highlighted in the segment are possible for a chip vendor to build over time. What's less possible is the SW...CUDA is why NVDA reigns supreme currently in the training game.

Loved this episode! Will there be a 4th NVIDIA podcast? 🤣

you can’t:

What a moat nvidia has created !! Superb

This is one of the toppiest things I’ve seen

I think that was a lot of “stacking the deck”, and I don’t think to compete you necessarily need the same architecture that Nvidia has. This has been demonstrated already with TPUs and @amazon Inferentia/trainium. Moving compute performance to ~99% vs ~55-60% on a gpu.

👀

That thing coming from left field would be @tesla with dojo.. the only company making agi and not just a chatbot. Truth is nobody knows Elon’s grand plan with @tesla @neuralink and @xai… stay tuned! At that time there will be agi as a service w/ dojo. #PayUp 💸

It could be equal, if there isn’t a wait time of a few months!

"CoWoS lithography packaging" wrong way to put it. These two are entirely different! @AcquiredFM

Please upload video podcast like standard oil.... It is much helpful to watch these two brilliant minds than speak...

.....enter Tesla dojo

Subscribed. Looks interesting.

Reminds me of @tesla when it comes to full self driving!

how much is being used for all these trash LLMs?

Competition can provide an ok solution at very low price. Model training is offline activity so end users will barely notice the difference.

@SaveToNotion #tweet

This is amazing and it's 🥇, just sub'd in all your platforms and gonna hear all the NVDIA wisdom from all the three episodes 😍🚀🚀🚀

There's no competing head to head with Nvidia, Intel, AMD have figured that out many moons ago. You can only hope to find your own niche and be the best at that. Hoping its enough to get a little slice of the pie 🥧

@videodownIoader

Brilliantly put. @MonetiveWealth hope Fintwit is able to comprehend what he just said.

Who is being interviewed here?

Xilinx was gaining share before amd got it. Why doesn’t it have a chance?

This was very, very interesting, informative, and fun to watch. Please continue to make more of these!!

Looks like you guys need to checkout mosaicML benchmarking or A100 and MI250, MI250 which works out of the box with all the components that you guys just mentioned, MI300 benchmarking will be done once it’s out.

A detailed thread on Nvidia Founder🧵👇