Apple Ramps Up R&D Intensity to Pre-iPhone Levels

> That's why I said Moore's law limping along.. it has slowed significantly, and the original transistor doubling every 18~24 months is definitely dead, but transistor processes are still improving, slowly.

What they're really doing is getting more power efficient. The "feature size" (number of nm) is fiction at this point. And the tricks they're using to eek out improvements at this point are, uh, interesting. Like they can't make it better without causing errors to occur so they just do that anyway and cover it up by using more error correction. It's really not obvious how much more of this there is to be found.

> The trade off is always there, but more people are using their phones over their desktop computer these days.

That isn't exactly true. The growth in the number of people with a desktop computer has leveled off, but it hasn't really gone down, and the things people use them for are still the same things. If you're going to write a long document, you're going to want a full-sized keyboard. The GPU needed for modern AAA games isn't going to fit in a phone. Professional activity that requires a lot of computation, like compiling code or editing video (or, going forward, a lot of this AI stuff), is a lot faster on a high-power device with more CPU or GPU cores.

What actually happened is that at least the same number of people still use a desktop computer, but now everybody has a phone, including the people who traditionally had a desktop (and still do). So more people have a phone than have a desktop, but not because fewer people have a desktop.

And it would be hard for some new technology to do the same thing to phones because nearly everybody has a phone. A new device can't have 20% more users than phones do when 97% of people have a phone.

IMO for a place like apple, where they are buying chips in such volumes they define entire production runs anyway and define target design specs for vendors there is probably already a lot of cost they are consuming where adding something like the apple silicon design project is cheaper. The marginal cost of going full apple silicon is lower as a result. Apple has also been very frustrated with their chip vendors, especially intel, in making what they want so I imagine it started as a 'last resort' project after they've exhausted their vendors.

While another smaller place that wanted to do something like apple silicon would have much higher marginal costs because they would need to spend on the things that apple was already spending on implicitly. It became an economy of scale thing.