The article argues that the dotcom crash was caused by companies lacking structural moats, not a failure of the internet itself, and applies this lesson to Dario Amodei's strategy for building an AI monopoly.
- Successful dotcom-era companies like Google and Microsoft possessed strong structural moats in niches such as search, hardware, or enterprise software that benefited from scale.
- Companies without real moats collapsed when investors demanded returns, contrasting with the current AI landscape where Dario aims to extract rent and data through massive spending on data centers and subsidies.
- The author contends that LLM technology faces a plateau once human-produced data is exhausted and offers limited improvement for real-world problems compared to coding or math.
- Local open-source models are becoming sufficient for profitable use cases, making expensive cloud API calls like GPT-5.5 or Opus economically unviable due to lack of ROI.
The author suggests that Dario's plan to burn trillions of dollars to create a necessary AI utility is flawed because the underlying technology does not provide the exclusive advantages required to sustain such a business model.