This reminds me of a project I worked on for a grad school data science course here in Canada. We tried to map this "digital divide" using public data.
Turns out, it's really tough to do accurately. The main reason is that the public datasets are a mess. For example, the internet availability data is in neat hexagons, while the census demographic data is in weird, irregular shapes that don't line up. Trying to merge them is a nightmare and you lose a ton of detail.
So our main takeaway, rather than just being a pretty map, was that our public data is too broken to even see the problem clearly.
I wrote up our experience here if anyone's curious: https://zeinh.ca/projects/mapping-digital-divide/