There are several suppositions that you're making that I'm not sure are warranted, the first being that supporting multiple platforms will be "expensive".

Will it? Or, for the most part, will it just be flipping a switch in Xcode to tell LLVM to build a universal binary? If you're doing iOS development you already do it daily when an app is built to run on x86 and the simulator and when it's recompiled to ARM when it's sent to a device. The process is transparent.

As to "experience shows"... well, I lived and worked through both of those transitions. They weren't that rough, and Apple's toolset has improved since then.

Finally, as to benchmarks, I had some of them in the article and it appears that you and I are reading them completely differently. the A12z, for example, is running close to a 13" MBP IN A TABLET. That's without a notebooks power budget and without a notebooks much more extensive cooling capacity (e.g. fans).

That chip is almost 2 generations old, the 13 bests it and the 14 is supposed to kill it. The 14 is supposed to be able to 4K video recording at 240 FPS, whereas the best the A13 can do is 4K at 60. That's 4x the performance.

That is, of course, with custom silicon and where I thought it was funny that you mentioned machine learning tools... as Apple's notebooks will now have the full support of the Neural Engine behind it. And the article mentions just how well a mere tablet manages 4K video editing.

I don't think Apple would have made the jump if they weren't prepared to blow Intel out of the water, performance-wise.

The flip side to your arguments is what if that 12-core Apple Silicon chip screams AND gets 12 hours of battery life, and developers flock to it accordingly?

And, personally, I think the later case is much more likely. At any rate, we'll see in just a couple of months, won't we?