In an AI World, MSFT & AMZN Are Shorts
If we are going to drink the tonic and assume that everything will be automated by AI tomorrow (which is what company valuations suggest), then we ought to short MSFT & AMZN.
The rationale is as follows: Azure is the engine that drives MSFT. AWS is the engine that drives Amazon. Neither Azure nor AWS make world-class chips that are competitive with NVIDIA (NVDA). Only Google (GOOGL) has meaningful AI traction with its TPU chips, most notably, Anthropic’s Claude models leverage Google’s TPUs.
Therefore, in a “AI world”, if you are one of the cloud hyper-scalers (AWS, Azure, Google, maybe Oracle (ORCL)), you need to be able to offer world-class compute to your customers. This means that AWS and Azure are completely dependent on third party GPUs such as those from NVIDIA and AMD (AMD).
Leaning on chip manufacturers is only going to take you so far. I would argue that when it comes to Cloud Services, it is chip quality that is the key differentiator when it comes to AI processes. The software piece is not as complicated to do well.
Thus, what is to stop NVIDA and perhaps AMD and other chip makers from investing heavily in proprietary cloud businesses? NVIDIA doesn’t need Azure, AWS and other cloud service providers as customers.
Google on the other hand is much better positioned.