AI was one of the topics that dominated the zeitgeist in 2025. There’s so much happening at any given moment and there’s so much more written about it that it’s hard to figure out how to make sense of it all.
That’s especially the case given the tremendous hype around this technology.
I find it helpful to think of AI in terms of three kinds of bets.
The first is working on a foundation model. This involves a select group of labs – for now, that’s Anthropic, OpenAI, Google, Meta, and a few others. The bet here is “superintelligence” – which I think is just a fancy term for an incredibly dependable AI assistant or agent that every consumer will use to navigate the internet and their digital life.
There’s potentially a tremendous amount of money to be made here. This is already evidenced by the billions of dollars of subscription revenue flowing into these systems. Just imagine what happens when you add advertising revenue into the mix and you can see how lucrative this could be for a lab that figures this out.
But ultimately, the game here is providing incredible intelligence in everyone’s pocket and really owning that market. That’s what every lab is racing toward.
Winning consumer attention is challenging and does tend to have winner-take-all dynamics. So, you also have these labs going after verticals (e.g., Anthropic for coding) as a way to hedge their bets. That brings us to the second category.
The second is Applied AI. This is going to be the vast majority of every other company that is building technology for various verticals/industries or functions. Here, the bet is simple – can you use AI to dramatically disrupt/change how things work in that particular industry or function?
Essentially, this is going to create new categories of winners and losers. New category winners who will get there by completely disrupting existing workflows. There are many industries/functions/verticals with archaic, human-heavy workflows that can all be reimagined – many for the better.
And, again, as you can imagine, there’s a lot of wealth creation that can occur here – proportional to the breadth and depth of the disruption.
The final area is AI adjacent companies. These are companies that provide tools or platforms. NVIDIA is an example of an AI adjacent company. So are the cloud providers – Amazon, Microsoft, and Google Cloud – along with fast-growing data and AI tool providers like Databricks and Snowflake.
In all of these, the bet is that as AI use continues, more and more workflows will need these tools, and these tools will essentially take a percentage of the AI economy.
I call these out in these three buckets just because this is the bet you’re making when you invest in one of these companies or when you decide to work at one of these companies. And it’s helpful to be clear about what you’re betting on.
For example, I know someone who was choosing between working at a lab or an applied AI company. It became a lot easier for this friend to figure out what they’d be interested in once the central bet was clarified.
This is not to say that these bets will all work. There are a collection of other factors – whether the energy needed for all this will be built out in time, whether AI will actually disrupt workflows in the timeframe being bet on,, and so on.
They are called bets for a reason.
Best to go in with clarity of thought and eyes wide open.
A Learning a Day 