AI Infrastructure

With AI, we are in the same position as we were with the internet in 1995 (and maybe the PC in 1980).

We have this revolutionary technology, and only a small select group of people around the world are instinctively drawn to it (200 million out of 8 billion). These people are using it to get work done, while the rest are somewhat ambivalent about it.

Many investors think they are investing in technology (such as the clusters of data centers training foundational models), but what they are really investing in so far is augmenting those 200 million David Allens (author of Getting Things Done).

Without a distinct product category harnessing foundational models for significant, standalone functions, the value will primarily go to three types of entities:

1. Independent problem-solvers like these 200 million David Allens, who apply this technology for their customers.

2. Legacy companies that incorporate it into their existing frameworks.

3. OpenAI, Anthropic, and other foundational model producers (Meta, Amazon, Mistral, Cohere).

Many of you have seen that I've started to become skeptical about the investment case that many investors are making towards generative AI. This is a contrarian take that has gotten me some interesting feedback so far.

What are they investing in when they build these billions of dollars worth of chip infrastructure and data centers?

This is the first technology revolution I've seen where there doesn't seem to be any importance placed on the case that "efficiency of GPUs may significantly change."

Every previous generation of technology has been based on efficiency being king. This revolution has completely thrown that out the window, and investors are investing huge amounts into this infrastructure that makes bitcoin mining seem like a lamp in terms of energy draw.

Here are some facts:

1. Meta plans to invest between $35 billion and $40 billion in AI infrastructure in 2024.

2. Microsoft has announced multiple significant investments, including $3.3 billion in Wisconsin to build AI data centers and a manufacturing-focused AI Co-Innovation Lab, and an additional $2.9 billion investment in Japan for cloud and AI infrastructure over the next two years.

These giant investments are being made into technology that may get vastly more efficient, as Kip Mock states in this clip. I reached out to Kip to interview him about Valar Atomic's plan to create synthetic hydrocarbons next to a nuclear power plant but ended up getting some very interesting insights into the bear case of the AI hype (why I love doing podcasts).

If you like this clip, I suggest listening to the full episode on Crazy Wisdom (iTunes, Spotify, and YouTube).