Software in the AI Era: Model Providers Are Expanding Their Layer, and Large Players May Keep Getting Larger

The development of the software industry in the AI era People have generally compared software development in the AI era with the successive arrivals of the internet and mobile internet. Software applications developed on infrastructure-as-a-service (IaaS) and platform-as-a-service (PaaS) foundations such as AWS, Azure, and GCP. That led to a proliferation of software-as-a-service (SaaS) developers and sharing-economy businesses such as Uber and Airbnb. The expectation was that once OpenAI, Anthropic, Google, and others had established large-model services, software developers could build a multitude of applications on top. But from OpenAI through Anthropic, now in the spotlight, my intuition is that the dominant model providers seem to be making the model layer ever more expansive while continually launching new applications. They appear to want to take on almost every application with meaningful scale. That leaves an extremely thin, highly undifferentiated, and easily replaced layer of value for small software developers. This is my personal impression, and it may be wrong. If it proves right, however, the proliferation of SaaS in the internet era may look completely different from software in the AI era. Large players may become even larger.