Are AI companies still buying training, post training data or hiring their own data annotator, synthetic data specialists etc..?

Not sure if this is the right sub-reddit, but I'm wondering if anyone has knowledge in actual trend within AI companies, especially the bigger labs. No doubt they still have agreement to collect and buy new data for training their models, but is the trend going down? Similarly for alignement, post training and fine tuning are they actually buying? Or is there a shift towards internalizing the capabilities? I'm seeing big labs hiring for synthetic data generation, sometimes even data annotations... or the other hand I'm also seeing startups getting traction by focusing more the infra for doing