Like many things where what constitutes base data changes and hence graphs change pretty radically, this can be cleanly explained by diagnostic drift. Here's a detailed breakdown: https://www.cremieux.xyz/p/how-to-end-the-autism-epidemic
While it's true that parents and schools have a weird relationship with ASD, with some parents doing anything to refuse the label no matter how apt, at other times with officials and parents going out of their way to do things like label kids with symptoms of severe trauma as having ASD, along with a long tail of higher SES parents who see claiming various special ed needs (not just ASD ones) as a way to get more resources for their children, particularly in states where they can get additional vouchers or payments with a label, these things don't really change a ground truth: severe autism is not easy to hide or easy to fake, and cases are increasing.
IMO diagnostic drift is maybe a fine thing hem and haw about when it comes to mild cases.
But it's basically a form a bike shedding because the severe cases are so incontrovertible and so much more common than they used to be. You can't claim that kids with severe cases are okay and just out to get money because they're so obviously and clearly not well. And schools have gone from maybe having none to a couple per grade to needing whole classrooms or even schools to safely handle high severity autism over the last 50 years.
You could maybe instead claim that "when achieving any semblance of normalcy is impossible, society shouldn't spend so much effort," but that is usually not a well-received message because it sounds like the forced institutionalization or incarceration of the mentally ill or locking them in bedrooms to be forgotten.
Similarly, there are some situations where the repercussions of severe early childhood trauma get diagnosed as autism, but these are also situations where you'd still have a massive service need, so there's no cost reduction to be had, just a proportionally small chunk of misdiagnosis.
I will ask someone more knowledgeable than me about this when I have a chance, but I think it became common to look at the severe cohort under the label of "profound autism," a few years ago, but I think has a tendency to be listed as a % of overall ASD cases rather than a separate rate.
But also noted that they find a lot of the research lacking - they view research, even that using the "profound" label, as not doing a sufficient job of mapping severity bands: the DSM now has three levels of severity, but research going back decades is hard to map to the evolving buckets. And they view the profound label as probably too narrow for fully encompassing kids who needs high levels of support in the educational setting.
Friend also says the way they would probably look at it in their school district is by label and support level: ASD, and then whether they need para pro support, or require a self contained classroom (IE a specialized classroom environment), and in terms of the ASD labels needing para pro or self contained setting going up in quantity even while district overall enrollment declines. And those support levels pretty much guarantee that either kiddo has severe behaviors or parents have lawyers, and the vast majority of parents don't have lawyers.
They attribute some of this to non-exceptional-education kids being drawn off to charter schools (which despite theoretical obligations basically do a good job of not providing real services for special needs kids to get them back to publics rather than reducing profits), but not remotely all of it.
Separately, one of my relatives has kiddos with what I'd colloquially call severe autism (like I don't see their kids ever living independently) and their district tried to move a big chunk of their elementary high support kids to a middle school because they "ran out of space" for elementary self contained / autism classrooms.
