AI demand exploded (320x in one year). But energy, memory, and costs are creating three walls at once. Those who could read the numbers moved already. For everyone else, the bill is coming.
“Amazon alone will invest $200 billion in infrastructure in 2026, more than the annual budget of entire nations. It also signed an agreement with Anthropic: $33 billion total in equity, $100 billion in AWS services over 10 years with 5 gigawatts guaranteed. In other words: Amazon isn’t betting on Anthropic, it’s building it.”
really did your research with this article. it shows! awesome stuff. great read all around.
The news about 40% of data centers being shut down was very popular lately, and I saw a lot of people saying, "The bubble is bursting; they were building things they don't need."
With the data in hand, the reality is very different!
The article is right: transformers take 5 years, HBM covers 60% of demand through 2027, and the "AI included in your plan" model is dead. Real constraints. Real money.
But there's a second layer. The frontier is capacity-constrained. The long tail isn't.
A fine-tuned 7B running on a single GPU for ~$30 doesn't need Stargate. It doesn't need Three Mile Island. It runs today on hardware that already exists. Most enterprise problems - classification, extraction, structured RAG, voice matching - never needed Opus in the first place.
So both are true: frontier AI is hitting a physical ceiling, and most teams relying on it never needed to be there.
“Amazon alone will invest $200 billion in infrastructure in 2026, more than the annual budget of entire nations. It also signed an agreement with Anthropic: $33 billion total in equity, $100 billion in AWS services over 10 years with 5 gigawatts guaranteed. In other words: Amazon isn’t betting on Anthropic, it’s building it.”
really did your research with this article. it shows! awesome stuff. great read all around.
The news about 40% of data centers being shut down was very popular lately, and I saw a lot of people saying, "The bubble is bursting; they were building things they don't need."
With the data in hand, the reality is very different!
yeah right? makes a huge difference.
My guess is that people are in fact hoping that AI is a bubble because they are afraid of what is coming :)
same. i think there is FUD, but realistically all the ai power users get it. these tools aren’t going away, they are getting exponentially better.
Two things are true at once.
The article is right: transformers take 5 years, HBM covers 60% of demand through 2027, and the "AI included in your plan" model is dead. Real constraints. Real money.
But there's a second layer. The frontier is capacity-constrained. The long tail isn't.
A fine-tuned 7B running on a single GPU for ~$30 doesn't need Stargate. It doesn't need Three Mile Island. It runs today on hardware that already exists. Most enterprise problems - classification, extraction, structured RAG, voice matching - never needed Opus in the first place.
So both are true: frontier AI is hitting a physical ceiling, and most teams relying on it never needed to be there.
This is true. For lot of my tasks I currently need the frontier models, but for smaller and simple tasks the local models are very good