FT says slower AI progress could cut costs for investors

In short: A Financial Times analysis says that pacing AI development could lower the massive costs of building new AI models and that investors might benefit.

What's going on

Some leaders at major AI companies have been talking about slowing the pace of AI progress. The idea is to avoid creating very powerful “frontier” AI models, meaning the newest and strongest systems, before society has clear safety checks.

Anthropic CEO Dario Amodei recently argued that AI progress should be “paced.” He suggested governments should coordinate standards and that companies should give outside experts access to review risks. The piece notes that US politicians may not be eager to step in, but companies and investors may still have reasons to slow down.

One reason is money. Training new AI models requires huge amounts of computing power, which is like renting an enormous fleet of supercomputers for long periods. The column points to OpenAI plans that could total $750 billion in computing capacity by 2030. It argues that if spending is delayed, the cost looks smaller in today’s dollars, because the money goes out later.

The column also says a slowdown might not hurt sales as much as some people think. Even if the newest models can charge higher prices, many customers may already rely on cheaper options. AI companies have also been competing on price.

What to watch

A key risk is competition. If top companies slow down, rivals, including open-weight models that people can download and modify, may catch up. At the same time, the column argues there is still a lot of business to be made by turning today’s AI into practical tools inside everyday software.

Source: Financial Times