An important element of this is that there is strong and self-reinforcing geographical clustering effects around high-value skills. Engineers do not generically make $300k-$1M, it is engineers with specific domain expertise. I've had teams all over the world, the market is more efficient than I think some people expect. Focusing on wage arbitrage ignores that companies are optimizing for ROI, not cost per se.
I think it is also important to recognize that the engineers top American tech companies hire are often not American, they just happen to be in America. It is a geographical bias, not an American-as-nationality bias.
Engineers with unusually high skill in any particular area are geographically diffuse. You can't hire a "team" in some distant inexpensive country, there might only be one or two such engineers in the entire country and they are difficult to find. The moment a company invests in creating a concentration of domain skill where none really exists, that clustering of expertise has a natural gravity not only for engineers with those skills but for other companies that want to hire those skills. Why search for a particular engineering expertise in Bolivia when you know there are a hundred top engineers in some specific city? Even if the wages are higher, the risk is lower, hiring velocity higher, and the engineers more experienced, which has a large real ROI on its own. So people with those skills gravitate toward that concentration because that is where the money and jobs are. Wash, rinse, repeat.
There are many large cities in the world with a reputation for having a proper tech presence, albeit on a lower tier, where it is nearly impossible to hire certain expertise and experience even for Silicon Valley money.
You see the effects of this throughout tech, and not just software, where specific skill sets are oddly concentrated in one or two regions, including ones that are not particularly known for being tech or where the company that originally created the concentration became irrelevant decades ago. This expertise flywheel is how new globally leading tech hubs are created in practice e.g. Taiwan for silicon fabrication or Seattle for cloud.
This is the tech hub bootstrapping problem. You have to heavily invest in a valuable technical domain that is currently being ignored by existing tech hubs. Taking advantage of these opportunities requires vision and an appetite for risk, since it is a contrarian position almost by definition. I think this explains much in the European context, where you have large concentrations of excellent engineers but minimal investment in genuinely bold visions.