You are forgetting a very important detail. Investing is the core of a company’s activities: they invest in human resources (experience, training and retention), in R&D and projects, sometimes in acquisitions, to adjust to ever-changing market conditions (customers expectations, competition, trends, …).
They choose to invest, presumably a lot, in AI developers and AI models that require 16GB of RAM + 6-8GB of VRAM to run (see PL9’s release notes) and that many users have trouble to run.
Training the models isn’t free (neither in computing or in training data), even using ONNX as they did.
But they thought at the time it was a better bet than preserving or increasing their potential reach.
I guess we’ll soon see where they invest the money from our licences next.