Why the Next Trillion Dollar AI Company Won't Be a Frontier Model

"AI is the new electricity.", Andrew Ng (CoFounder of Google Brain)

This week, I'm tackling one topic that I know you'll benefit from:

Why the Next Trillion Dollar AI Company Won't Be a Frontier Model

(And How This Can Get You A Job)

After spending hours preparing for this week's newsletter, I'm quite certain the next trillion dollar AI company won't be built off of another frontier AI model, but rather it will be built on the infrastructure that businesses use to host and deploy them. Listen closely, because I'm also going to tell you how this can land you a VERY comfortable tech job.

To emphasize my point, you all have likely heard about the recent events with Claude Fable being released and then being near immediately banned because it was deemed a security threat. Well, interestingly, Claude Fable was reportedly costing Anthropic $600 an hour, so the ban ended up saving Claude MILLIONS of dollars. However, in this same space of time, companies like Databricks that make the infrastructure for these models have remained profitable.

Databricks, for example, has a valuation of over $60 billion dollars that is driven by enterprise demand as opposed to the consumer hype. Everyone talks about Fable, but the companies that make these money burning models possible are hardly mentioned in popular discourse.

Now I'm going to explain what this means for you:

The Skill Employers EXPECT: You should learn at least one AI enterprise platform deeply. (AI Foundry, Azure, DataBricks Mosaic AI, AWS Bedrock, or Google Vertex AI for some examples) These platforms let you build, evaluate, and scale AI agents. Companies are using these every day.

The Skill That Separates You

You should learn Context Engineering

This means knowing:

1. What information and tools the model should have access to
2. When to retrieve knowledge instead of relying on memory
3. How to measure success of the model. (Most important)

The Project To Learn These Skills This Week

Build a secure internal knowledge assistant. It should store company documents in a vector database, add role-based access (ex: HR can't access Finance), log every response, and display the source of every answer from the retrieval. This is a simple outline because the project should challenge you to think of the features that aren't specified and train you to ask the right questions.

If you want to sharpen your machine learning skills even more, I also selected a challenge problem for you this week:

If you learned something from this newsletter, make sure to forward it to a friend.