Investors debate whether AI models will become cheap and similar

In short: Big tech is spending huge sums on AI, but some researchers and executives say AI models may become cheap and hard to keep unique.

What's going on

Big technology companies are expected to spend about $800 billion this year on AI data centers, which are large buildings full of computers that run AI systems. The business case depends on whether each company’s AI model can do something meaningfully better or different than rivals.

A post on X from someone described as an OpenAI researcher argued that as communication improves, knowledge spreads faster. The post also suggested that AI tools can help people build more AI, which could make it harder for any one company to hold a lasting advantage. In simple terms, it is like a recipe that keeps getting copied and improved until everyone can cook the same dish.

Anthropic CEO Dario Amodei raised a similar concern in a February interview. He described a world where a small number of firms can afford the costs of building AI and where models are still different from each other. But he also pointed to a risk: if AI systems help build new AI systems, the “moat” (a protective barrier that keeps competitors out) could shrink, and many more players could catch up.

What to watch

If AI models become more similar, prices could fall and profits could drop, even as spending on data centers stays high. Watch for signs that smaller companies can match top models without huge budgets, and for how firms try to stay distinct through features, safety controls, or exclusive data.

Source: Financial Times