Theme: Are hyperscalers making good returns and are we in a capex bubble? Tech sector and some stock implications.
Executive summary
One of the current debates in the hyperscaler world (MSFT, GOOG, META, AMZN, ORCL) is whether the massive levels of capex that is being spent is delivering good returns and based on that, are we in a capex bubble or not. To start with context for how much is being spent, we estimate that during the transition to the cloud/pre-AI era, close to $1t -$1.5t was spent over 10-12 years before normalizing vs. the AI hyperscaler build out that so far has been close to $0.8t over the past few years. That implies the AI buildout has likely already reached 50-80% of what the cloud era spent over a much longer period – 3 years for AI vs. 10-12 years for cloud!
The speed of capex spend is breathtaking and unprecedented and can be attached to both a bull case and bear case. Either there is a bubble in capex that could burst in the next 12-24 months, or we are still in the early innings, or both. Return on investment is one parameter in this equation along with ability to finance as another. We will share our view on this later but the main analysis we wanted to look at for this note is what the returns look like so far for the hyperscalers and what are some of the financial model and investment implications.
