AI lifecycle

There have been a few images doing the rounds which shows the intertwined funding relationships of the AI ecosystem. This one from Bloomberg got a lot of attention.

Let me simplify what is happening here. Demand from the end user is red hot. The service providers need to buy GPU chips (mostly from Nvidia) to meet that demand. You know that irritating lag you get when you are waiting for ChatGPT to make a picture? That delay will worsen if new computing power isn't purchased, as thousands of new users are added daily.

The service providers have two major constraints, access to energy and cash. Nvidia can solve one of those problems because it has too much cash, currently sitting on $60 billion and printing more every quarter. So they invest in their clients who can then buy more Nvidia chips.

This will unravel and pop if the demand dries up. Do you believe that will happen? I don't. If the demand continues to grow, Nvidia fixes the cash bottleneck problem and ends up with stakes in software companies that will continue to be their clients for many years and will most likely start making big profits themselves. Seems like money well spent to me.