AI companies today fall into 4 main buckets (market map with selected AI companies)

In my view, companies in the AI landscape today fall into four main categories:

• Foundational model labs

• Physical infrastructure

• Applications (vertical & horizontal)

• Platform (SW) infrastructure

The first two buckets share certain characteristics which make them less attractive as investment opportunities: high capital intensity and well-established incumbents. We still may see innovation in these buckets, but there are high barriers to entry that may prevent widespread development. In my opinion, investors should look to the second two buckets for future investment opportunities, which is what this thesis will focus on.

Applications

Observers of the industry generally believe that 2024 will be the year AI matures toward real-world applications. The big question is which spaces are most ripe for disruption by these AI applications. My sense is that we should consider certain verticals which are well positioned to leverage AI at scale. I believe these verticals will have a few key characteristics: first, I think they will be verticals which historically haven’t always been early adopters in tech changes, which means their tech stack may be a bit outdated to begin with. These companies will feel a pressure to involve AI in some sort in order to stay ahead of competitors. A second characteristic, I think, is that they will likely be verticals which have a lot of unstructured or unused data to extract value from. These two characteristics are very prominent in healthcare, legal, and financial institutions, among others. 

Platform Infrastructure

Many companies will seek to develop AI workflows in house, either to improve internal workflows or augment external products for their customers. This means that companies will need to build infrastructure to carry out their ML operations, from data through training and deployment. Things like feature storage, data pipelines, test cases and safety/privacy are all required when trying to develop models that can scale across an organization. Incumbent tech giants have already invested in building some of these tools in-house to support their AI workflows (e.g. ads for Meta). However, many other companies trying to catch up in this AI wave may not have the resources to do so, opening the door for platform infrastructure companies to sell workflow tools. It’s still to be seen whether plug and play modular tools or full toolchain companies will win out, but it’s worth keeping an eye on startups in this bucket as infrastructure spend within companies can often be very high. 

All in all, there is an abundance of innovation in the AI space currently, but by looking at key buckets within the space, we can make sense of potential areas to focus on. This is a work in progress, and certainly isn’t exhaustive, but hopefully it gives a sense of how I’m thinking about the landscape. 

Last Edited: 03/18/2024