The training of AI models is expensive. The running of those models is becoming cheaper and cheaper, as evidenced by DeepSeek's breakthrough:

While DeepSeek’s savings on training might look like a big deal today, it is the savings on inference, which has to happen at runtime, will be the key. It will drive major demand as the whole planet moves over the coming years to take advantage of this new power. - https://crazystupidtech.com/archive/what-deepseek-means-for-everyone/

What does this mean for OpenAI and Anthropic? As Benedict Evans lays out in "The Deep Research Problem", they lack a moat. The rise of open source models and the decreasing cost of running models means they face tremendous pressure.

Who are the winners here? The cloud providers. I believe we will see Anthropic and OpenAI continue to lead in consumer facing AI applications. But in the corporate world? Folks will not be able to resist the cost savings of running their own internal models.

There are some open questions here that I need to refine:

• What counts as a cloud provider? What about companies (like Google) that double dip?

• What counts as an AI company? Is Meta considered one?