That's a good question with a neat answer: assortive mating can amplify a randomly occurring gene by sexual selection. Read on for the details.

Suppose that there was a gene that made you more likely to be good at white collar thinkwork. You meet your partner at a professional event and have a baby. The baby now has between zero and two copies of the gene. And depending on gene dominance, the probability of at least repeating the ratio of the previous generation is like 75% IIRC.

Repeat, and fold back in. Next generation, you have a couple folks with both mutations, who exhibit an even stronger affinity for quiet thoughtful work (and people). Their kids will have a 100% chance of carrying at least one copy, right? The numbers for the third gen will be even higher.

This is all massively simplified but yeah. Mendel has the answers. The picture is even clearer if you suppose multiple genes whose effects stack, which appears to be the case with the ol 'tism.

And, finally, there is the lived experience of why this happens so much. Speaking as someone on the spec, I could never imagine life partnership with an allistic person, nor can I imagine such a person putting up with my deficiencies and quirks. Moreover, quite literally all my friends are also autspec and would never interface with someone who isn't. So I believe ASD is strongly selected for by people with ASD.

But given the volumes I'd expect we are in the hockey stick armpit of an exponential curve. Like HIV bubbling away in 1950s Africa long before we even had a name for the condition, nineteenth century colonial industrialism probably quietly worked up a stock of the relevant mutations, and the mutants found each other in the grand blender of the 20th century's upheavals.

As it were, silicon valley is one of the precipitants of sex, war, and intellect.