If I was to build a software startup today I’d pick GCP and DataBricks. Wall Street doesn’t understand that the way you monetize AI is not advertising normie queries, it’s onboarding enterprises onto your cloud platform via the frontier model you offer.

Google is the only one of the big 3 public cloud platforms that has their own top tier frontier model, the other two license models built by startups they don’t entirely control. They also are the cheapest per token and are focusing the most on latency which is underestimated.

Currently 1/3 of all queries are computer science questions. IMO this fraction just goes up from here, especially if you add in queries embedded in a software application that aren’t directly classified as “computer science”.

Also, this is not the dot com bubble. You need a massive amount of capex to actually build this out, and you need the legit talent to execute. Google has the major distribution with GCP, they have the top tier AI talent, and they have an infinite cash machine. They will win.