AI is all the rage these days. But few people have a clear picture of what the AI ecosystem looks like. I have been thinking about it for some time, and feel that a supply chain perspective can do the job. Not long ago, I had the good fortune of listening to one of Barry Ritoltz’s interviews with Tony Kim from Blackrock in which Tony gave an excellent depiction. I thought I would like to share.
I am going to be paragraphing somewhat. Think of a pyramid, or a layer cake. On the bottom there are power and energy, which feed into the next layer up which are computer servers and chips. Those servers sit in data centers somewhere, but today we call them computing cloud. These three layers can be categorically lumped into one big label of infrastructure.
The big bulk part of the pyramid are models and data. For models, you have mostly a handful of companies (Microsoft, Google, OpenAI, xAI, etc..) that are building these foundation models.
AI models need data to train them. Therefore, data providers are an indispensable part of the ecosystem. These include large content providers such as Reuters of news, or Thompson Reuters who has the largest legal datasets.
Finally on the top of the pyramid are applications, from popular consumer applications such as ChapGPT or Perplexity AI, to professional applications in a wide range of industries such as healthcare.
When I fed the above to ChapGPT, it produces the following:
Applications
(ChatGPT, Perplexity AI, etc.)
/ \
/ \
Industry-specific apps Consumer apps
(Healthcare, Finance, Law)
/ \
/ \
--------------------------
| Models & Data |
| (OpenAI, Google, etc.) |
| (Reuters, Databases) |
--------------------------
/ \
/ \
-----------------------------
| Infrastructure (Cloud) |
| (Power, Servers, Chips) |
| (Data Centers) |
-----------------------------Not bad, eh?
I then tried Copilot. Surprise… It generated something nice, but not very useful. It is the picture you see at the top. Well… so much as AI today.