Amid all the hype about artificial intelligence, many investors are leaping before they look. However, an important debate to consider is whether AI companies can build durable businesses with moats — or whether they’re commodity-like companies.
At the 2026 Morningstar Investment Conference, Malik Ahmed Khan of Morningstar said they believe economic moats do exist in AI and that these companies are building durable competitive advantages. He expects AI companies to be able to generate excess returns and put up the robust profitability required to support their lofty valuations. However, he also took the time to present the bear case so that investors can consider both sides of the equation.
How LLMs work
Large language models are set up in four stages:
- Data ingestion and processing
- Pre-training or core model building
- Post-training and alignment
- Inference and deployment
The training stage includes building the factory. The companies gather raw text data from billions of documents on the internet, massive GPU clusters process the data, and neural network layers enable the model to learn patterns and relationships. Finally, the finished blueprint is released.
Login required to continue reading.
Setup a free account to get access to this article (no credit card required).
View Full Article
