Discussion about this post

User's avatar
Vits Karel's avatar

Interesting articles/analysis of the AI build-out.

Just wondering, if it's such a risky investment now in these high-tech hyperscalers, why does a guy like Warren Buffett initiated a buying spree in Google (Alphabet), in 2025 ?

Mark Pizzini's avatar

Very interesting thesis. One thing to consider: there is no rule that says a hyperscaler needs to keep buying the latest chip. They will only do that if they have the demand for the additional compute and the new chip offers a better return than the old one. As demand slows, they will spend less and less on new chips. They are currently playing catch-up, and that won’t last forever. Depending on the pricing they may be able to get a better return from buying an older generation chip. A second thing to consider is the lead times to add capacity and the granularity of capacity. With railroads, the lead times were relatively long and the amounts of capacity tended to be relatively lumpy. For example, if I have a line from NY to Philly and I want to add a line from Philly to DC, it’s all or nothing and I cannot get paid until it’s 100% complete. The biggest concern I have is how long it will take for: 1) existing companies to generate returns 2) emerging AI businesses take to reach Amazon and Google-like scale. The returns from source 1 will be needed to bridge the gap until the returns from source 2 are big enough. The longer that bridge, the more tenuous things get for the hyperscalers

1 more comment...

Ready for more?