Why are all these data companies buying internal data?

Well other than the new wave of trying to mine useful saleable insights using new AI tools, there has always been a demand for insider trading signal sources. If you can find data that predicts stock market movements consistently, even if those movements are slow, then you can beat the bond rates and obtain consistent returns (barring any stock market crash).

A friend of mine owns a few home health care businesses in the bay area. He said he's approached weekly and offered anywhere from $500k-2M for his data.

Ai training is the orifice through which everything is subsumed.

To train AI models on it? It's right in the headline.

If only I'd known that I should have been collecting as much minutiae as possible all these years...

They scrapped the entire internet and need more

Wait until the market matures enough for underpaid or disgruntled employees being able to sell data easily.