Marco Arment

But that architecture of saying, we don't want to build...

We either don't want to or can't build a chip this big on a single die because we're at the reticle limit or whatever.

So we're going to take two of these other things, to your point, two of these other chips that we already have to build for our laptops.

And we're not going to do a whole separate chip, but we're just going to take two of the laptop chips.

And we also designed it in a way for them to talk to each other with this, you know, fusion architecture where they stick to each other end to end.

Just like our old shirt says with the two maxes flipped over and stuck together.

That architecture does not scale well, but that's not Apple silicon.

That's whatever it is, the silicon interposer fusion architecture thing.

I think we're beyond that now, as evidenced by the chiplet architecture on the M5 Pro and M5 Max.

And so I'm not saying what Apple is going to do, but if they wanted to...

They could totally abandon the silicon interposer two maxes stuck together architecture and just build a big dedicated CPU on its own die with no GPUs and a big dedicated GPU on its own die with no CPUs and make them both at the radical limit and stick them on a chip.

And there's your M5 Ultra.

I would love for them to do that.

I don't think they're going to, but I would love for them to do that.

And that is still Apple Silicon.

So there's nothing inherent in Apple Silicon that says you have to do that end-to-end thing.

And that is the linear scaling barrier.

That is the thing that costs tons of money but doesn't give you the bang for your buck.

And so I think the runway is clear for them to do something better, especially with like 2 nanometer or whatever they end up going to in like the M6 or M7.

I forget when that's going to happen.