I think it's less a case of saturation and more a case of companies realizing that most data scientists don't actually deliver value commensurate to the salaries they were asking. Most companies simply don't have enough data, or don't have hygienic enough data, or don't have the engineering heft to build a reliable data pipeline, so the data scientists often find themselves set up to fail. I've seen that happen a few times.

But, it's not like the fundamentals of the field are wrong. Predictive modeling is still really useful. It's just larger firms are the only ones capable of realizing that value.