https://www.bloomberg.com/news/articles/2025-09-25/david-einhorn-sees-tremendous-capital-losses-from-ai-spending

“David Einhorn cautioned that the unprecedented amount of spending on artificial intelligence infrastructure may destroy vast amounts of capital,”

Basically this is nonsensical, the infrastructure will always exists and not a one-time use like consumer products. His perspective is too dreamy or groundless. This is never about creating a consumer product that maybe nobody is going to buy, compute power will always earn its own investments back overtime and that demand never decline historically.  “destroy vast amounts of capital” this clearly shows that he lacks understanding of how cloud computing (AWS, Oracle) derives their revenue from and their huge operating margins a far from being “destroyed".

Due to all the fearmongering perspectives that have little to no projections, most are solely based on “Dotcom era” and “Bubble” perspective is very much akin to those who fear running because he had fell before. Such irrational fears tend to overlook factors of the fall, but look solely at “velocity”. It is similar to "you are running so fast, so you will fall because you fell before.”

We all know rationally, it doesn’t work that way. Running fast doesn’t logically linked to falling. If not, there is no Olympics because everyone is fearful of falling.

Those that look at OpenAI spending and Nvidia investment from a negative perspective tend to exaggerate, and like to based on today’s use cases to make future projections. That is akin to making predictions in the industrial revolution that electricity is only going to be used for light bulbs.

The spending on compute power is never going to be a waste nor go away. We have already saw how some of the bitcoin mining companies have shifted from mining to powering AI. Most of the analysts somehow have a skewed perspective that compute power can only to used for 1 thing? Such is a case for Coreweave (previously a bitcoin mining computer).

We can see how the shift is going into robotics (physical AI). Cloud computing and AI computing will only need more compute and never less. We have already seen such evolution before in cloud computing. More advanced models will only need more compute never less.

Infrastructure spending and compute demand have historically never been on the decline. The datacenter that they built today might not even able to handle the computer demand 3-5 years later (assuming that they don’t upgrade anything).