There are at least two reasons: it's still expensive, and we still don't really know what to do with it.
The "$1000 genome" is a bit of a myth, or rather PR hype from Illumina. Sure, they proved it could be done if you juke the numbers the right way. But I believe a clinical-grade whole-genome sequence still costs several times that much (see https://bmchealthservres.biomedcentral.com/articles/10.1186/...). Even at $1000/genome, sequencing the entire US population would still be $330 billion, for unclear benefit.
I'd love to be corrected, but the science just isn't there to show us what to do with WGS data except in limited applications like cancer or "rare diseases". It's a bit of a chicken-and-egg problem in that sense, and it's being worked on. There are pharma companies and public consortia sequencing hundreds of thousands of genomes, and mining them alongside medical records and other phenotypic data. So the value may come eventually, but it will probably be less about preventing illness so much as curing disease in a much more targeted way. And even then, you won't need to have your whole genome sequenced just to know if you should take drug A or drug B -- a cheap targeted test will suffice.
First, context: These are small scale clinical/experimental settings which clearly do not take advantage of economies of scale. Labor costs are thus greatly magnified. This effect is also obscured by their costing breakdown, which often does not separate reagent cost from labor.
Unless I'm misreading this, the pricing that they define often includes 2 sets of sequencing runs, one for the baseline and one for the tumor samples. That puts the existing price at about 4000 USD for 2 samples or a 2000 USD/sample cost.
The expansive definition of "sequencing cost" to envelope downstream analysis and storage here is also a mistake IMO. I would rather narrow down on specifically the cost to get the raw sequence data out, excluding clinical and bioinformatics analysis. This would include (reasonable) sample extraction, library preparation, and the actual sequencing process itself.
There are definitely significant cost barriers to sequencing, but we shouldn't discount savings that come with scale.
Of course, all this math goes out the window if we were to look at exome sequencing instead which while less complete, still gives a large amount of information at a fraction of the cost.
Also, I would challenge your suggestion that "The expansive definition of 'sequencing cost' to envelope downstream analysis and storage here is also a mistake." I would argue that it's the opposite: delivering valid and useful sequencing results to a broad population would require even more resources directed toward analyzing the results and also interpreting them for the patients.
(To be clear here in europe medical records are held by healthcare services, but they are protected. They are not private property tho)