Moat

A company makes money because it offers something customers want. Once that works well, imitators show up and offer the same thing more cheaply. A moat is anything that stops these imitators. That can be a well-known brand, a patent, an especially low price made possible by huge volumes, or simply the hassle that switching means for customers. The term comes from investing and was made popular by the US investor Warren Buffett. The image is the water-filled ditch around a medieval castle: it doesn’t make the castle stronger, it just makes attacking it unappealing.

Why investors ask about this with AI companies

A lead without a moat isn’t worth much. Whoever has the best product today might not have it a year from now. But investors don’t pay for a stock based on today’s profits — they pay for many years of profits to come. That’s why the decisive question isn’t who’s ahead right now, but who will still be ahead in ten years.

This question is especially open in AI. Large language models — the programs behind chatbots like ChatGPT — are often technically similar to one another. When one provider releases a better model, the others often catch up within a few months. Freely available models, whose blueprint anyone can download, shrink this gap even further. A purely technical lead is therefore a weak moat.

Other areas look different. The chipmaker Nvidia doesn’t just sell hardware — it has spent years supplying a software environment that countless researchers have built their programs on top of. Switching to a competitor would mean rewriting much of that. It’s exactly these conversion costs that form the moat.

What a moat is actually made of

Economists distinguish a few typical sources. The network effect is the strongest: a service becomes more valuable the more people use it. A social network without friends is useless, so everyone stays where everyone already is. Alongside this are switching costs — the effort that changing providers causes. A company that runs its entire accounting through one piece of software won’t switch over a ten percent price difference.

A third source is economies of scale. Whoever produces very large volumes spreads fixed costs over more units and can offer lower prices than any newcomer. On top of that come legal protections like patents, and finally brands, for which customers pay more out of habit or trust.

A common mistake is confusing size with a moat. Nokia was once the largest phone maker in the world and still lost the market within a few years. Market share only protects you if it triggers one of the mechanisms mentioned above. Likewise, a good reputation alone isn’t a moat if customers can switch effortlessly at any time.

The term in earnings reports and headlines

In business news, the word usually shows up when a lead is wobbling. When the Chinese company DeepSeek unveiled a powerful model at very low cost in early 2025, tech stocks dropped sharply. Headlines immediately asked whether American providers' moat was smaller than thought.

Executives also use the term themselves, for instance on quarterly earnings calls. A leaked internal memo at Google in 2023 was titled “We Have No Moat” and was quoted around the world. When you read the word, it’s almost always about the same question: how long can today’s profit be defended?

For you as a reader, the term is above all a checking tool. When an article celebrates a company, you can ask what’s actually stopping a competitor from doing the same thing. If no convincing answer comes to mind, the lead is probably temporary.